CN102704215A - Automatic cutting method of embroidery cloth based on combination of DST file parsing and machine vision - Google Patents

Abstract

Translated fromChinese

本发明公开一种基于DST文件解析与机器视觉结合的绣布自动切割方法，在毛绒玩具制造业，毛绒玩具的外形和图案是采用刺绣机器进行加工，然后采用激光切割机将整幅绣片进行逐一切割。本发明方法首先通过将DST文件（刺绣文件）解析为图像文件，选择其中的切割单元并提取其信息，获得其单元的内部参数。然后根据照相机拍照得到的切割样本的图像，对其进行相关的处理，将之前得到的切割单元信息与处理后的样本匹配，最终确定切割线位置并产生PLT文件，由切割机进行切割。不同于其它基于机器视觉的激光切割方法，本方法只需要拍照一次就可以完成整幅刺绣样本的切割，本方法具有工作效率高、残次率低的特点，对复杂刺绣样本有很强的切割能力。

The invention discloses an embroidered cloth automatic cutting method based on the combination of DST file analysis and machine vision. In the plush toy manufacturing industry, the shape and pattern of the plush toy are processed by embroidery machines, and then the entire embroidery is cut by a laser cutting machine. slices are cut one by one. The method of the invention first parses the DST file (embroidery file) into an image file, selects a cutting unit therein and extracts its information, and obtains the internal parameters of the unit. Then, according to the image of the cutting sample taken by the camera, relevant processing is performed on it, the cutting unit information obtained before is matched with the processed sample, and the position of the cutting line is finally determined and a PLT file is generated, which is cut by the cutting machine. Different from other laser cutting methods based on machine vision, this method only needs to take a photo once to complete the cutting of the entire embroidery sample. This method has the characteristics of high work efficiency and low defect rate, and has a strong cutting ability for complex embroidery samples. ability.

Description

Translated fromChinese

基于DST文件解析与机器视觉结合的绣布自动切割方法An automatic cutting method for embroidered cloth based on the combination of DST file analysis and machine vision

技术领域technical field

本发明涉及一种绣布自动切割方法，基于DST文件解析与机器视觉的结合。The invention relates to an automatic cutting method for embroidered cloth, which is based on the combination of DST file analysis and machine vision.

背景技术Background technique

目前的绣布激光切割系统主要有两种：一种是先在白纸上面切割一次，然后将绣布样本放在切割面上，将绣布上的花型与白纸上切割出的轮廓对齐后再切割一次；另一种是在激光切割头上安装相机，每切割一个花型单元就要提取样本并处理一次，即每次只能切割一个花型单元。由于故有切割系统存在工作效率低，残次品比重高等问题，因此迫切需要一种新的切割方式，用以取代现行方式。机器视觉的飞速发展为我们解决上述问题提供了一种新的思路。There are two main types of embroidered cloth laser cutting systems: one is to cut once on the white paper, then place the embroidered cloth sample on the cutting surface, and align the pattern on the embroidered cloth with the outline cut out on the white paper Cut again; the other is to install a camera on the laser cutting head, and each time a pattern unit is cut, a sample must be taken and processed once, that is, only one pattern unit can be cut at a time. Due to the problems of low working efficiency and high proportion of defective products in the existing cutting system, a new cutting method is urgently needed to replace the current method. The rapid development of machine vision provides us with a new way of thinking to solve the above problems.

机器视觉就是利用机器代替人眼来做各种测量和判断。它是计算机学科的一个重要分支，它综合了光学、机械、电子、计算机软硬件等方面的技术，涉及到计算机、图像处理、模式识别、人工智能、信号处理、光机电一体化等多个领域。在生产线上，人来做此类测量和判断会因疲劳、个人之间的差异等产生误差和错误，但是机器却会不知疲倦地、稳定地进行下去。一般来说，机器视觉系统包括了照明系统、镜头、摄像系统和图像处理系统。对于每一个应用，我们都需要考虑系统的运行速度和图像的处理速度、检测目标的尺寸还是检测目标有无缺陷、视场需要多大、分辨率需要多高、对比度需要多大等。典型的机器视觉系统可以分为：图像采集部分、图像处理部分和运动控制部分。Machine vision is the use of machines instead of human eyes to do various measurements and judgments. It is an important branch of computer science, which integrates technologies in optics, mechanics, electronics, computer software and hardware, etc. . In the production line, human beings make such measurements and judgments, and there will be errors and mistakes due to fatigue, differences between individuals, etc., but machines will continue tirelessly and steadily. Generally speaking, a machine vision system includes a lighting system, a lens, a camera system and an image processing system. For each application, we need to consider the operating speed of the system and the processing speed of the image, the size of the detection target or whether there is a defect in the detection target, how large the field of view needs to be, how high the resolution needs to be, and how much the contrast ratio needs to be. A typical machine vision system can be divided into: image acquisition part, image processing part and motion control part.

一个完整的机器视觉系统的主要工作过程如下：The main working process of a complete machine vision system is as follows:

1、照相机根据指令进行拍照，其曝光时间预先设定；1. The camera takes pictures according to the instructions, and the exposure time is preset;

2、图像采集部分接收摄像机数字化后的数字图像数据，并将其存放于处理器或计算机内存中；2. The image acquisition part receives the digital image data digitized by the camera and stores it in the processor or computer memory;

3、处理器对图像进行处理、分析、识别、获得测量结果或逻辑控制值；3. The processor processes, analyzes, and recognizes the image, and obtains measurement results or logic control values;

4、处理结果控制流水线的动作、定位、纠正运动的误差等。4. Process the results to control the action, positioning, and correction of motion errors of the assembly line.

发明内容Contents of the invention

发明目的：为了克服现有技术中存在的工作效率低下，材料利用率低，成本较高等问题。本发明通过运用机器视觉的方法，解决了原有切割系统科技含量与工作效率低下、残次品比重高的缺点。通过运用机器视觉的方法，降低了人在切割过程中的作用，依靠相机与切割机的高精度，可以大大提高产品的成品率、降低成本、提高工作效率。Purpose of the invention: In order to overcome the problems of low work efficiency, low material utilization rate and high cost in the prior art. The invention solves the shortcomings of the original cutting system, such as low scientific and technological content, low work efficiency and high proportion of defective products, by using the method of machine vision. By using the method of machine vision, the role of human beings in the cutting process is reduced. Relying on the high precision of cameras and cutting machines, the yield of products can be greatly improved, costs can be reduced, and work efficiency can be improved.

技术方案：为解决上述技术问题，本发明提供一种基于DST文件解析与机器视觉结合的绣布自动切割方法，其特征在于该方法用到一种激光切割机，该激光切割机切割面正上方架设有相机，所述绣布自动切割方法包括如下步骤：Technical solution: In order to solve the above technical problems, the present invention provides an automatic cutting method for embroidered cloth based on the combination of DST file analysis and machine vision, which is characterized in that the method uses a laser cutting machine, and the cutting surface of the laser cutting machine is directly above A camera is set up, and the automatic cutting method of embroidered cloth includes the following steps:

(1)DST文件解析，获得原始刺绣文件图像信息和内部特征信息，为图像匹配做准备工作；(1) DST file analysis, obtaining original embroidery file image information and internal feature information, and preparing for image matching;

(2)样本提取与处理，通过相机拍照获得样本图像信息，并对样本图像进行畸变矫正、二值化处理以及提取样本内部特征操作，获得样本的特征信息；(2) Sample extraction and processing. The sample image information is obtained by taking photos with the camera, and the sample image is subjected to distortion correction, binarization processing, and internal feature extraction of the sample to obtain the feature information of the sample;

(3)匹配，将得到的DST文件与样本的特征信息进行匹配，匹配单元质心并根据最小二乘法进行质心匹配与切割线校正；(3) matching, the obtained DST file is matched with the feature information of the sample, the unit centroid is matched and the centroid is matched and the cutting line is corrected according to the least square method;

(4)DST映射，将匹配后的切割线坐标通过坐标变换映射到切割面坐标系下用于切割，然后输出激光切割机能够执行的PLT文件，由激光切割机进行切割。(4) DST mapping, the coordinates of the matched cutting line are mapped to the coordinate system of the cutting surface for cutting through coordinate transformation, and then output the PLT file that the laser cutting machine can execute, and the laser cutting machine performs cutting.

本发明通过DST解析算法将用于加工毛绒玩具的绣布刺绣文件(DST文件)信息输入到处理系统进行处理，获得匹配样本，然后通过相机提取绣布样本图像，并运用DST解析获得的信息以及一定的处理算法检测绣布，获得其中各个切割单元的位置，将处理结果反馈到激光切割机执行切割操作，对提高切割精度、降低工作复杂度、减小工作量有很重要的意义。The present invention uses the DST analysis algorithm to input the information of the embroidered cloth embroidery file (DST file) used for processing plush toys into the processing system for processing to obtain matching samples, then extracts the embroidered cloth sample image through the camera, and uses DST to analyze the obtained information And a certain processing algorithm detects the embroidered cloth, obtains the position of each cutting unit in it, and feeds back the processing results to the laser cutting machine to perform cutting operations, which is of great significance for improving cutting accuracy, reducing work complexity, and reducing workload.

本发明利用机器视觉代替人工的方法，利用照相机分辨率高、工作稳定的性能，通过一定的处理程序，运用先进的图像处理算法可以大大提高切割效率，降低切割机对于人的依赖程度，提高劳动生产率，由于本发明的高精度，可以有效提高布料的利用率，降低切割的残次品比重。不同于其它基于机器视觉的激光切割方法，本方法只需要拍照一次就可以完成整幅刺绣样本的切割，具有工作效率高、残次率低的特点。对复杂刺绣样本有很强的切割能力。The invention uses machine vision instead of manual methods, utilizes the performance of high resolution and stable work of the camera, through a certain processing program, and uses advanced image processing algorithms to greatly improve the cutting efficiency, reduce the dependence of the cutting machine on people, and improve labor efficiency. Productivity: Due to the high precision of the present invention, the utilization rate of cloth can be effectively improved, and the proportion of defective cut products can be reduced. Different from other laser cutting methods based on machine vision, this method only needs to take a picture once to complete the cutting of the entire embroidery sample, which has the characteristics of high work efficiency and low defect rate. It has a strong cutting ability for complex embroidery samples.

附图说明Description of drawings

图1机器视觉切割流程图Figure 1 Flow chart of machine vision cutting

图2DST文件显示画面；Figure 2 DST file display screen;

图3对应DST文件刺绣后的绣布；Figure 3 corresponds to the embroidered cloth after DST file embroidery;

图4DST文件内部独立单元提取；Figure 4 Extraction of independent units inside the DST file;

图5边框提取流程图；Fig. 5 frame extraction flow chart;

图6DST单元和边框相交；Figure 6DST unit intersects with the frame;

图79点漫水填充法；Figure 79 point flood filling method;

图8用于获得变换坐标系的切割后的圆图像；Fig. 8 is used to obtain the cut circle image of the transformed coordinate system;

图9MFC操作界面；Figure 9 MFC operation interface;

图10激光切割机示意图；Figure 10 schematic diagram of laser cutting machine;

图11图像匹配后在MFC界面显示出的图像；Figure 11 The image displayed on the MFC interface after image matching;

图12切割线定位整体图像；Figure 12 Cutting line positioning overall image;

图13切割线定位局部图像；Figure 13 Cutting line positioning partial image;

图14激光切割机切割过程；Figure 14 Laser cutting machine cutting process;

图15切割后图像。Figure 15 Image after cutting.

具体实施方式Detailed ways

理论基础：图像识别。图像识别的第一步是提取有效的图像特征，在这里，我们主要引入矩特征量和轮廓特征量，矩特征量包括Hu矩、归一化转动惯量(NMI)和仿射不变矩，而轮廓特征量包括轮廓离散化参数和奇异值分解。这5个特征量都具有良好的平移、旋转和比例不变性，在图像目标识别中得到了广泛的应用。Theoretical Basis: Image Recognition. The first step in image recognition is to extract effective image features. Here, we mainly introduce moment features and contour features. The moment features include Hu moment, normalized moment of inertia (NMI) and affine invariant moment. The contour feature quantity includes contour discretization parameters and singular value decomposition. These five feature quantities all have good translation, rotation and scale invariance, and have been widely used in image target recognition.

下面结合附图1对本发明作更进一步的说明。Below in conjunction with accompanying drawing 1 the present invention will be further described.

1DST文件解析1DST file analysis

由于电脑设计的绣花文件为DST格式，不能直接获得其内部特征与图像信息，因此本发明的第一步是DST文件解析，对DST文件进行解析，获得其图像信息与内部特征。DST文件是由头文件与花样文件体组成，两部分用于存放不同的信息，具体解析过程如下。Because the computer-designed embroidery file is in DST format, its internal features and image information cannot be directly obtained. Therefore, the first step of the present invention is to analyze the DST file to obtain its image information and internal features. The DST file is composed of a header file and a pattern file body. The two parts are used to store different information. The specific analysis process is as follows.

1.1DST头文件组成1.1 DST header file composition

花样文件头主要用来存放一些关于花样的说明信息，DST文件有一个512字节的文件头，前128字节以文本文件的形式存储了一些文件信息，这些信息包括信息标识和信息值，信息值使用十进制码，未使用的部分由空格字符(ASCII码值0x20)代替，每一段由字符“\”(ASCII码值0x0D)隔离，具体各段意义参见表1。The pattern file header is mainly used to store some description information about the pattern. The DST file has a 512-byte file header, and the first 128 bytes store some file information in the form of a text file. These information include information identification and information value, information The value uses decimal code, and the unused part is replaced by a space character (ASCII code value 0x20), and each segment is separated by the character "\" (ASCII code value 0x0D). See Table 1 for the meaning of each segment.

表1  DST格式花样文件头结构Table 1 DST format pattern file header structure

1.2花样文件体结构1.2 Pattern file body structure

花样文件里存放的是描述花样全部针迹动作的数据，既包括了每条针迹的位移，也包括了绣花机的剪线、换色等动作的控制。DST、格式的花样文件体是从第513个字节开始到文件结尾，且每3个字节用来表示一针，如表2所示。The pattern file stores the data describing all the stitching actions of the pattern, including the displacement of each stitch, and the control of the embroidery machine's thread trimming, color changing and other actions. The pattern file body in DST format is from the 513th byte to the end of the file, and every 3 bytes are used to represent a stitch, as shown in Table 2.

A₇A₆A₅A₄A₃A₂A₁A₀A₇ A₆ A₅ A₄ A₃ A₂ A₁ A₀  B₇B₆B₅B₄B₃B₂B₁B₀B₇ B₆ B₅ B₄ B₃ B₂ B₁ B₀  C₇C₆C₅C₄C₃C₂C₁C₀C₇ C₆ C₅ C₄ C₃ C₂ C₁ C₀

表2  花样文件中每三个字节表示一针Table 2 Every three bytes in the pattern file represent a stitch

本文约定水平向右为X正向，垂直向下为Y正向。我们令Δx，Δy，分别表示一条针迹在X、Y方向上的位移，即当前针相对于前一针绣框在X、Y方向上移动的相对步数。DST是一种三进制的编码形式。Δx，Δy需要参照表3所示的权值计算出来。This article agrees that horizontally to the right is the positive X direction, and vertically downward is the positive Y direction. We let Δx and Δy represent the displacement of a stitch in the X and Y directions respectively, that is, the relative steps of the current needle moving in the X and Y directions relative to the embroidery frame of the previous stitch. DST is a ternary encoding form. Δx and Δy need to be calculated with reference to the weight values shown in Table 3.

表3.DST格式的一针中每位对应的权值Table 3. The weight corresponding to each bit in a needle in DST format

计算公式如下：Calculated as follows:

Δx＝A₃*(-9)+A₂*9+A₁*(-1)+A₀*1+B₃*(-27)+B₂*27+B₁*(-3)+B₀*3+C₃*(-81)+C₂*81Δy＝A₇*1+A₆*(-1)+A₅*9+A₄*(-9)+B₇*3+B₆*(-3)+B₅*27+B₄*(-27)+C₅*81+C₄*(-81)DST中C₇、C₆、C₁、C₀这四位的组合构成其功能码，如表4所示。Δx＝A₃ *(-9)+A₂ *9+A₁ *(-1)+A₀ *1+B₃ *(-27)+B₂ *27+B₁ *(-3)+B₀ *3+C₃ *(-81)+C₂ *81Δy＝A₇ *1+A₆ *(-1)+A₅ *9+A₄ *(-9)+B₇ *3+B₆ *(-3)+B₅ *27+B₄ *(-27)+C₅ *81+C₄ *(-81) The combination of C₇ , C₆ , C₁ , and C₀ in DST Its function code, as shown in Table 4.

  功能 Function取值value  缝制sewingC₇＝C₆＝0，C₁＝C₀＝1C₇ =C₆ =0, C₁ =C₀ =1  跳针jump stitchC₇＝C₁＝C₀＝1，C₆＝0C₇ =C₁ =C₀ =1, C₆ =0  换色change colorC₇＝C₆＝C₁＝C₀＝1，表示该针的其它位为0C₇ ＝C₆ ＝C₁ ＝C₀ ＝1, indicating that the other bits of this needle are 0  剪线 Trimming若干个连续的跳针码Several consecutive skipping codes  结束 FinishC₇＝C₆＝C₅＝C₄＝C₁＝C₀＝1，表示该针的其它位为0C₇ ＝C₆ ＝C₅ ＝C₄ ＝C₁ ＝C₀ ＝1, it means that other bits of this needle are 0

表4DST的功能码Table 4 Function codes of DST

在DST格式文件末尾还有一个结束字符0xlA，另外，空针为特殊的缝制针，值均等于0。There is also an end character 0xlA at the end of the DST format file. In addition, the empty needle is a special sewing needle, and its value is equal to 0.

根据上述过程，将DST文件的所有数据信息进行解析，解析完成后，创建一幅空白图像，在该图像上根据解析得到的信息，在此空白图像上勾画出来，从而获得DST文件的图像信息。According to the above process, all the data information of the DST file is analyzed. After the analysis is completed, a blank image is created, and the information obtained by the analysis is drawn on the blank image to obtain the image information of the DST file.

1.3整版DST文件中的单元提取1.3 Cell extraction in full-page DST files

图2是一幅典型的毛绒玩具制造业的DST文件，图3为其对应的刺绣布料。这类DST文件的典型特征是：Figure 2 is a typical DST file of the plush toy manufacturing industry, and Figure 3 is the corresponding embroidery fabric. Typical characteristics of such DST files are:

1、DST文件中有一个大的外边框，用于确定绣布的大小；1. There is a large outer border in the DST file, which is used to determine the size of the embroidery cloth;

2、内部有若干独立单元，每一个单元即为制作毛绒玩具的绣片；2. There are several independent units inside, and each unit is an embroidery piece for making plush toys;

3、独立单元都有各自的边框线，这条边框线实际是该单元的切割线；3. Each independent unit has its own border line, which is actually the cutting line of the unit;

4、独立单元之间为跳针，没有线连接，或者只有一根线连接；4. There is a jumper between the independent units, there is no wire connection, or there is only one wire connection;

5、为了节省空间，内部单元之间往往按照左右对称，或者上下对称进行排列；5. In order to save space, the internal units are often arranged according to left-right symmetry, or up-and-down symmetry;

6、独立单元内部元素不超过切割线。6. The internal elements of the independent unit do not exceed the cutting line.

根据这些特征，我们采用了如图4所示的方法来提取DST文件中的独立单元。According to these characteristics, we adopted the method shown in Figure 4 to extract the independent units in the DST file.

1.4DST解析1.4 DST analysis

理论上DST文件的解析按照DST文件的头文件组成和编码方式进行处理就可以了，但是在实际的DST文件中，文件数据往往会出现个别错误，典型问题就是头文件中表示整幅DST大小的+X，-X，+Y，-Y数据往往错误，所以需要对DST中的点进行排序计算，获得实际的DST图像大小。In theory, the analysis of DST files can be processed according to the header file composition and encoding method of the DST file, but in the actual DST file, there are often individual errors in the file data. The typical problem is that the header file indicates the entire DST size. +X, -X, +Y, -Y data are often wrong, so it is necessary to sort and calculate the points in DST to obtain the actual DST image size.

首先，按照DST文件编码方式获得每针的偏移坐标，即Δx，Δy值，然后通过逐步相加，获得整幅DST每一针的坐标。由于DST文件中坐标零点为起始点，为了后续图像处理方便，需要把坐标进行处理，将左上方第一点设为坐标零点，向右为x轴正方向，向下为y轴正方向。在解析的同时，记录下每一针对应的刺绣操作状态。Firstly, obtain the offset coordinates of each stitch according to the DST file encoding method, that is, Δx, Δy values, and then obtain the coordinates of each stitch of the entire DST through stepwise addition. Since the coordinate zero point in the DST file is the starting point, for the convenience of subsequent image processing, the coordinates need to be processed, and the first point on the upper left is set as the coordinate zero point, the right direction is the positive direction of the x-axis, and the downward direction is the positive direction of the y-axis. While analyzing, record the embroidery operation status corresponding to each stitch.

1.4.1外边框的判断1.4.1 Judgment of outer frame

一般来说整幅绣布都是有外边框的，绣布外框主要是用于大体确认DST文件的尺寸，但是我们在实际调研中发现有的产品不需要外边框。因此，DST格式分析的第一步，就是判断是否有外边框。判断外边框的流程，如图5。Generally speaking, the entire embroidered cloth has an outer frame, and the outer frame of the embroidered cloth is mainly used to roughly confirm the size of the DST file, but we found that some products do not need an outer frame during actual research. Therefore, the first step in DST format analysis is to determine whether there is an outer border. The process of judging the outer border is shown in Figure 5.

在实际的DST绣片中为了节省布料，有些DST单元可能与外边框相交，如图6。所以边框的尺寸不一定和DST版面大小相等。外边框都是由绣花机一次绣成的，因此外边框是一个连续的闭合数据段，而DST文件中连续数据段非常多，因此需要对收集到的DST连续数据段进行校验，判断其是否为外边框。判断条件：In order to save fabric in actual DST embroidery pieces, some DST units may intersect with the outer frame, as shown in Figure 6. Therefore, the size of the border is not necessarily equal to the size of the DST layout. The outer frame is embroidered by the embroidery machine at one time, so the outer frame is a continuous closed data segment, and there are many continuous data segments in the DST file, so it is necessary to verify the collected DST continuous data segment to determine whether it is for the outer border. Analyzing conditions:

1.四个端点和DST版面大小相关；1. The four endpoints are related to the DST layout size;

2.各端点之间的点均在一条直线上。2. The points between the endpoints are all on a straight line.

1.4.2DST单元边框的判断1.4.2 Judgment of DST cell frame

毛绒玩具DST文件中的DST单元有如下特征：The DST unit in the plush toy DST file has the following characteristics:

1、DST单元边框闭合；1. The frame of the DST unit is closed;

2、单个DST单元边框刺绣过程中不换线；2. Do not change the thread during the border embroidery process of a single DST unit;

3、两相邻边框若距离较近可能不换线，一般都会换线；3. If the distance between two adjacent borders is relatively close, the line may not be changed, and the line will generally be changed;

4、DST单元排布不确定，目前相关领域还没有自动排版软件，都是设计人员自己随机排布，随意性较大；4. The arrangement of DST units is uncertain. At present, there is no automatic typesetting software in related fields, and the designers arrange them randomly, which is relatively random;

5、若有外边框，则DST单元的边框只被外边框包围。若无外边框，则没有其它闭合数据段包围DST单元边框。5. If there is an outer frame, the frame of the DST unit is only surrounded by the outer frame. If there is no outer border, then there are no other closed data segments surrounding the DST cell border.

根据以上特征，首先需要确定所有的闭合数据段，然后判断某个闭合数据段是否是单元边框。根据特征即可判断一个闭合数据段是否是DST单元边框。在实际操作中，DST单元边框往往是不规则图形，判断内部是否有元素较为困难，因此我们都采用外接矩形的方法行进处理。According to the above characteristics, it is first necessary to determine all closed data segments, and then determine whether a certain closed data segment is a cell border. Whether a closed data segment is a DST cell frame can be judged according to the feature. In actual operation, the borders of DST units are often irregular shapes, and it is difficult to judge whether there are elements inside, so we all use the method of circumscribed rectangles for processing.

1.4.3DST内部元素确定1.4.3DST internal element determination

毛绒玩具DST文件中DST单元内部元素特征：Features of the internal elements of the DST unit in the plush toy DST file:

1、内部单元分多次绣完。出于减少换色次数的考虑，一般都是将整个版面中相同的颜色都绣完之后才会换色，所以一般来讲内部单元要多次才能绣完，所以内部元素较为分散。1. The internal unit is embroidered several times. In order to reduce the number of color changes, the color is generally changed after the same color in the entire layout is embroidered, so generally speaking, the internal units need to be embroidered many times, so the internal elements are relatively scattered.

2、内部元素不会超过单元边框。2. Internal elements will not exceed the cell border.

根据以上特征，我们的处理方法主要是将除开外边框(若存在)、单元边框之外的连续数据和已有的DST单元边框外接矩形进行匹配，即可确定某个DST单元边框中有哪些元素。According to the above characteristics, our processing method is mainly to match the continuous data except the outer border (if it exists) and the unit border with the existing bounding rectangle of the DST unit border, so as to determine which elements are in a certain DST unit border .

1.5提取DST单元内部信息1.5 Extract internal information of DST unit

根据以上DST文件解析得到的DST单元，提取其内部参数，用于下一步的图像匹配。通过计算获得内部特征图、质心、特征向量等参数。由于DST单元内部有可能存在过多的特征，因此在提取DST单元内部信息的时候可以设置为自动提取信息，也可以由人工提取的方式，即用鼠标选择感兴趣的区域。本发明倾向于后者，匹配的效果更好一些。According to the DST unit obtained by parsing the above DST file, its internal parameters are extracted for the next step of image matching. Parameters such as internal feature maps, centroids, and feature vectors are obtained by calculation. Since there may be too many features inside the DST unit, when extracting the internal information of the DST unit, it can be set to automatically extract the information, or it can be manually extracted, that is, to select the area of interest with the mouse. The present invention tends to the latter, and the matching effect is better.

2样本提取与处理2 Sample extraction and processing

视觉检测最重要一步就是图像采集与处理，通过高分辨率的相机可以实现精确的视觉测量，本系统采用Cannon 550d相机，本相机具有1800万像素，分辨率为5184*3456，镜头为定焦50mm镜头。通过高分辨率的照相机可以较为全面的提取样本信息，为样本图像的处理奠定了基础。本系统具有较强的适用性，可以应用于其它类似的系统中。The most important step in visual inspection is image acquisition and processing. Accurate visual measurement can be achieved through a high-resolution camera. This system uses a Cannon 550d camera. This camera has 18 million pixels, a resolution of 5184*3456, and a fixed-focus lens of 50mm. lens. The sample information can be extracted more comprehensively through the high-resolution camera, which lays the foundation for the processing of the sample image. This system has strong applicability and can be applied to other similar systems.

2.1系统标定2.1 System Calibration

由于要处理的绣布最大尺寸为60cm*40cm，考虑到相机的镜头参数，将相机架设于切割面正上方1.4m处。一旦相机安装到位，将镜头垂直对准下方的切割机工作面，调整相机至最佳焦距，然后锁死，不再改变相机参数和位置。相机的有效视野约为62.98cm*42cm。Since the maximum size of the embroidered cloth to be processed is 60cm*40cm, considering the lens parameters of the camera, the camera is set up 1.4m directly above the cutting surface. Once the camera is installed in place, align the lens vertically to the working surface of the cutting machine below, adjust the camera to the best focal length, and then lock it, and do not change the camera parameters and position. The effective field of view of the camera is about 62.98cm*42cm.

根据张正友法进行相机标定，标定后可以得到相机的畸变参数、然后再利用最小二乘圆检测由相机坐标系到切割机坐标系的映射坐标等信息。从而完成如下功能：According to the Zhang Zhengyou method, the camera is calibrated. After calibration, the distortion parameters of the camera can be obtained, and then the least square circle is used to detect the mapping coordinates from the camera coordinate system to the cutting machine coordinate system. Thereby completing the following functions:

1、通过畸变参数进行预矫正，尽可能的恢复真实的绣布样本图像；1. Perform pre-correction through distortion parameters to restore the real embroidered cloth sample image as much as possible;

2、进行相机坐标系到物理坐标系(机器坐标系)下的映射，从而指导切割机正确切割。2. Carry out the mapping from the camera coordinate system to the physical coordinate system (machine coordinate system), so as to guide the cutting machine to cut correctly.

在第一次使用时需要进行标定，用于获得相机的畸变参数、内参矩阵与相机坐标系到切割机坐标系的映射关系。标定过程为：Calibration is required at the first use to obtain the camera’s distortion parameters, internal parameter matrix, and the mapping relationship between the camera coordinate system and the cutting machine coordinate system. The calibration process is:

首先通过“张正友棋盘格标定法”获得相机的镜头畸变参数以及内参矩阵。然后保存参数，以后每获得一幅图像后，首先进行畸变矫正，以尽可能恢复原始图像的信息。Firstly, the lens distortion parameters and internal parameter matrix of the camera are obtained through the "Zhang Zhengyou checkerboard calibration method". Then the parameters are saved, and after each image is obtained, the distortion correction is performed first, so as to restore the information of the original image as much as possible.

接下来，采用“基于最小二乘法圆检测的坐标系标定法”获得坐标系变换矩阵，步骤如下：Next, use the "coordinate system calibration method based on least squares method circle detection" to obtain the coordinate system transformation matrix, the steps are as follows:

(1)首先设计好用于切割标定圆的PLT文件，如下图7所示，在一张70cm*55cm的白纸上切割出6*9的不同半径排列的圆；(1) First design the PLT file for cutting the calibration circle, as shown in Figure 7 below, cut out 6*9 circles arranged in different radii on a piece of white paper of 70cm*55cm;

(2)从PLT文件中获取切割机坐标系下精确的圆心坐标，P_ij＝(x_ij，y_ij)，其中x_ij、y_ij分别表示第i行第j列的圆的圆心在切割机坐标系上的x轴坐标和y轴坐标；(2) Obtain the precise circle center coordinates in the cutting machine coordinate system from the PLT file, P_ij = (x_ij , y_ij ), where x_ij and y_ij represent the center of the circle in the i-th row and j-column respectively on the cutting machine x-axis coordinates and y-axis coordinates on the coordinate system;

(3)利用相机拍摄一张清晰的标定圆图像，然后利用已有的畸变参数，进行图像矫正；(3) Use the camera to take a clear calibration circle image, and then use the existing distortion parameters to correct the image;

(4)采用漫水填充法对矫正后的图像进行二值化处理，将图像中的圆变成白色，背景变为黑色；(4) Binarize the rectified image by using flood filling method, turning the circle in the image into white and the background into black;

(5)对二值化后图像进行轮廓提取，检测出每个标定圆的轮廓点序列C_ij；(5) Contour extraction is performed on the image after binarization, and the contour point sequence C_ij of each calibration circle is detected;

(6)对每个轮廓点序列进行RANSAC圆拟合，获得每个圆的圆心坐标Q_ij＝(x′_ij，y′_ij)，其中x_ij、y′_ij分别表示第i行第j列的圆的圆心在相机坐标系上的x轴坐标和y轴坐标；(6) Perform RANSAC circle fitting on each contour point sequence, and obtain the center coordinates Q_ij =(x′_ij , y′_ij ) of each circle, where x_ij and y′_ij represent row i and column j respectively The x-axis coordinates and y-axis coordinates of the center of the circle on the camera coordinate system;

(7)将切割机坐标系下的圆心坐标按照从左到右，从上往下的形式排列为齐次坐标矩阵P，将相机坐标系下圆心坐标也按照对应的次序表示为齐次坐标矩阵Q；(7) Arrange the coordinates of the center of the circle under the cutting machine coordinate system as a homogeneous coordinate matrix P from left to right and from top to bottom, and express the coordinates of the center of the circle under the camera coordinate system as a homogeneous coordinate matrix in accordance with the corresponding order Q;

$\mathrm{<span><span>P</span>P</span>}<span><span>=</span>=</span>\left[\begin{array}{c}{\mathrm{<span><span>x</span>x</span>}}_{\mathrm{<span><span>11</span>11</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>y</span>the\; y</span>}}_{\mathrm{<span><span>11</span>11</span>}}<span><span>,</span>,</span>\mathrm{<span><span>1</span>1</span>}\\ {\mathrm{<span><span>x</span>x</span>}}_{\mathrm{<span><span>12</span>12</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>y</span>the\; y</span>}}_{\mathrm{<span><span>12</span>12</span>}}<span><span>,</span>,</span>\mathrm{<span><span>1</span>1</span>}\\ <span><span>.</span>.</span><span><span>.</span>.</span><span><span>.</span>.</span>\\ {\mathrm{<span><span>x</span>x</span>}}_{\mathrm{<span><span>19</span>19</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>y</span>the\; y</span>}}_{\mathrm{<span><span>19</span>19</span>}}<span><span>,</span>,</span>\mathrm{<span><span>1</span>1</span>}\\ {\mathrm{<span><span>x</span>x</span>}}_{\mathrm{<span><span>21</span>twenty\; one</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>y</span>the\; y</span>}}_{\mathrm{<span><span>21</span>twenty\; one</span>}}<span><span>,</span>,</span>\mathrm{<span><span>1</span>1</span>}\\ <span><span>.</span>.</span><span><span>.</span>.</span><span><span>.</span>.</span>\\ <span><span>.</span>.</span><span><span>.</span>.</span><span><span>.</span>.</span>\end{array}\right]$$\mathrm{<span><span>Q</span>Q</span>}<span><span>=</span>=</span>\left[\begin{array}{c}{{\mathrm{<span><span>x</span>x</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>11</span>11</span>}}<span><span>,</span>,</span>{{\mathrm{<span><span>y</span>the\; y</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>11</span>11</span>}}<span><span>,</span>,</span>\mathrm{<span><span>1</span>1</span>}\\ {{\mathrm{<span><span>x</span>x</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>12</span>12</span>}}<span><span>,</span>,</span>{{\mathrm{<span><span>y</span>the\; y</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>12</span>12</span>}}<span><span>,</span>,</span>\mathrm{<span><span>1</span>1</span>}\\ <span><span>.</span>.</span><span><span>.</span>.</span><span><span>.</span>.</span>\\ {{\mathrm{<span><span>x</span>x</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>19</span>19</span>}}<span><span>,</span>,</span>{{\mathrm{<span><span>y</span>the\; y</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>19</span>19</span>}}<span><span>,</span>,</span>\mathrm{<span><span>1</span>1</span>}\\ {{\mathrm{<span><span>x</span>x</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>21</span>twenty\; one</span>}}<span><span>,</span>,</span>{{\mathrm{<span><span>y</span>the\; y</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>21</span>twenty\; one</span>}}<span><span>,</span>,</span>\mathrm{<span><span>1</span>1</span>}\\ <span><span>.</span>.</span><span><span>.</span>.</span><span><span>.</span>.</span>\\ <span><span>.</span>.</span><span><span>.</span>.</span><span><span>.</span>.</span>\end{array}\right]$

(8)采用最小二乘法，计算出从照相机坐标系映射到切割机坐标系的变换矩阵M，使之满足下式：(8) Calculate the transformation matrix M mapped from the camera coordinate system to the cutting machine coordinate system by using the least square method, so that it satisfies the following formula:

Q*M＝PQ*M=P

(9)保存矩阵M，将其作为以后的坐标系映射参数。(9) Save the matrix M and use it as the coordinate system mapping parameter in the future.

2.2样本提取2.2 Sample extraction

样本提取的过程是对样本进行拍照，然后执行畸变矫正的过程。每次放入绣布成品的时候需要保证样本完全在照相机的视野内，防止出现样本偏离相机视野的情况。在切割机的工作面上印有工作区的边界线，每次放置绣布时只要左上角大致对齐边界线即可保证绣布位置在有效检测范围内。The process of sample extraction is the process of taking pictures of the samples and then performing distortion correction. Every time you put in the finished embroidered cloth, you need to ensure that the sample is completely within the field of view of the camera to prevent the sample from deviating from the field of view of the camera. The boundary line of the working area is printed on the working surface of the cutting machine. When placing the embroidered cloth each time, as long as the upper left corner is roughly aligned with the boundary line, the position of the embroidered cloth is within the effective detection range.

2.3样本处理2.3 Sample processing

图3所示为一张采用全自动电脑刺绣机完成的绣布，绣布上布满了很多个花型单元(刺绣样品文件中的最小单元)，这些花型单元的排布位置基本与DST设计文件一致，因此存在一个从DST文件到绣布样本的透视变换矩阵，将DST坐标系变换到绣布物理坐标系。电脑刺绣机按照DST文件内包含的刺绣数据，执行缝制、跳针、换色、剪线等刺绣动作。通常电脑刺绣机的精度为0.1mm，在忽略机器系统误差的条件下，可以认为DST文件和实际的绣布成品，是一对一的全比例映射。当绣布被人为地放置在切割机的切割面上时，存在一个从DST文件到绣布样本图像的透视变换矩阵，将DST坐标系变换到相机中的绣布坐标系。但是实际使用中，绣布是柔软易变形的，导致这种变换是存在误差的。尤其是当绣布局部翘起时，会有明显的变形。因此需要采用最小二乘法，计算出最优变换矩阵。Figure 3 shows a piece of embroidered cloth completed by a fully automatic computer embroidery machine. The embroidered cloth is covered with many pattern units (the smallest unit in the embroidery sample file). The arrangement positions of these pattern units are basically the same as those of DST The design files are consistent, so there is a perspective transformation matrix from the DST file to the embroidered cloth sample, which transforms the DST coordinate system to the embroidered cloth physical coordinate system. According to the embroidery data contained in the DST file, the computerized embroidery machine performs embroidery operations such as sewing, stitch skipping, color changing, and thread trimming. Usually the accuracy of computerized embroidery machines is 0.1mm. Under the condition of ignoring the machine system error, it can be considered that the DST file and the actual finished embroidery cloth are one-to-one full-scale mapping. When the embroidered cloth is artificially placed on the cutting surface of the cutting machine, there is a perspective transformation matrix from the DST file to the embroidered cloth sample image, which transforms the DST coordinate system into the embroidered cloth coordinate system in the camera. But in actual use, the embroidered cloth is soft and easy to deform, which leads to errors in this transformation. Especially when the embroidery layout is partially lifted, there will be obvious deformation. Therefore, it is necessary to use the least squares method to calculate the optimal transformation matrix.

通常一个花型单元包含了一个闭合的切割线以及内部多个花型元素。这些花型元素的排布和形状都是刺绣机器依据DST文件绣出来的，因此，可以根据DST文件中获得花型单元的位置，在绣布上大致找到对应的刺绣花型单元。Usually a pattern unit contains a closed cutting line and multiple pattern elements inside. The arrangement and shape of these pattern elements are all embroidered by the embroidery machine according to the DST file. Therefore, the position of the pattern unit can be obtained according to the DST file, and the corresponding embroidery pattern unit can be roughly found on the embroidery cloth.

样本处理的基本步骤为：The basic steps of sample processing are:

1、单元定位，依据DST单元排布位置以及相机的线性变换，估计出花型单元的大致位置，定位出一个矩形框。定位过程中，根据DST解析得到的花型单元的大小，选取一个可以包括花型单元的矩形框，然后根据由相机到切割机的坐标系变换得出变换到切割机上的矩形框。1. Unit positioning, according to the arrangement position of the DST unit and the linear transformation of the camera, estimate the approximate position of the pattern unit, and locate a rectangular frame. In the positioning process, according to the size of the pattern unit obtained by DST analysis, a rectangular frame that can include the pattern unit is selected, and then the rectangular frame transformed to the cutting machine is obtained according to the coordinate system transformation from the camera to the cutting machine.

2、二值化，将花型单元矩形框内的彩色图像二值化，设置绣布背景色变为黑色，其他元素变为白色，保留花型单元内部元素的轮廓特征。本部分用了改进型的漫水填充法，用于提高漫水填充的效果。通过在刺绣样本中随机选择多个点作为样本点，通过判断这些随机选择的点是否为背景点，如果是则保留当作种子点，如果不是则忽略，改进后增大了系统的鲁棒性，漫水填充的效果更好。2. Binarization, binarize the color image in the rectangular frame of the pattern unit, set the background color of the embroidery cloth to black, and change other elements to white, and retain the outline features of the elements inside the pattern unit. This part uses the improved flood filling method to improve the effect of flood filling. By randomly selecting multiple points in the embroidery sample as sample points, by judging whether these randomly selected points are background points, if they are, they will be kept as seed points, if not, they will be ignored, and the robustness of the system will be increased after improvement. , the effect of flood filling is better.

3、计算轮廓特征向量，计算二值化边框的Hu矩与质心。3. Calculate the contour feature vector, and calculate the Hu moment and centroid of the binarized frame.

2.3.1单元定位2.3.1 Unit positioning

根据DST解析结果，可以得到每个DST单元的大小以及对应的边界框。DST文件与相机采集到的绣布样本之间也存在着一种线性变换，因此可以根据从DST文件中解析到的单元位置估算出绣布中的各个单元位置。由于相机是固定不动的，而且DST刺绣文件与对应的绣布可以看作是基本一致的。只是DST文件中的坐标与对应的相机坐标之间存在着平移、旋转、缩放三种变换。According to the DST parsing results, the size of each DST unit and the corresponding bounding box can be obtained. There is also a linear transformation between the DST file and the embroidered cloth sample collected by the camera, so the position of each unit in the embroidered cloth can be estimated based on the unit position parsed from the DST file. Since the camera is fixed, and the DST embroidery file and the corresponding embroidery cloth can be regarded as basically the same. It's just that there are translation, rotation, and scaling transformations between the coordinates in the DST file and the corresponding camera coordinates.

绣布的摆放位置初步决定了平移因子T和旋转因子R。由于绣布样本的摆放受到一定程度的约束，即绣布必须放入相机的有效视野中，因此可以估计出平移因子T′。通过相机的视野大小可以初步估计绣布的缩放系数S′。由绣布的四个角可以估计出旋转因子R′，具体步骤如下：The placement of the embroidered cloth preliminarily determines the translation factor T and rotation factor R. Since the placement of the embroidered cloth samples is restricted to a certain extent, that is, the embroidered cloth must be placed in the effective field of view of the camera, so the translation factor T′ can be estimated. The zoom factor S' of the embroidered cloth can be preliminarily estimated through the field of view of the camera. The rotation factor R′ can be estimated from the four corners of the embroidered cloth, and the specific steps are as follows:

(1)根据绣布摆放的位置约束：由于绣布摆放时需要考虑相机的视野问题，本系统在切割机上标有有效区域与直角坐标。将绣布大体按照要求放入切割机，则绣布旋转角度一般小于±10°。又由DST文件解析结果可以获得4个顶点单元的质心齐次坐标V_k＝(x_k，y_k，1)，k＝1，2，3，4，以及对应的矩形单元边框R_k＝W_k×H_k。顶点单元按照顺时针方向编号为1～4。(1) According to the position constraints of embroidered cloth placement: Since the field of view of the camera needs to be considered when placing embroidered cloth, this system marks the effective area and rectangular coordinates on the cutting machine. Put the embroidered cloth into the cutting machine roughly according to the requirements, and the rotation angle of the embroidered cloth is generally less than ±10°. The centroid homogeneous coordinates V_k = (x_k , y_k , 1), k=1, 2, 3, 4 of the four vertex units can be obtained from the analysis results of the DST file, and the corresponding rectangular unit frame R_k =W_k × H_k . Vertex units are numbered from 1 to 4 in a clockwise direction.

其中参数：x_k为DST解析后得到图像的第k个顶点单元质心的x轴坐标；y_k为DST解析后得到图像的第k个顶点单元质心的y轴坐标；R_k为DST解析后得到图像的第k个顶点单元的边框；W_k为其边框宽度；H_k为其边框长度。The parameters: x_k is the x-axis coordinate of the centroid of the k-th vertex unit of the image obtained after DST analysis; y_k is the y-axis coordinate of the mass center of the k-th vertex unit of the image obtained after DST analysis; R_k is obtained after DST analysis The border of the kth vertex unit of the image; W_k is its border width; H_k is its border length.

(2)采用估算出的平移和缩放变换系数，由下式得到样本图像中4个顶点单元的初始质心齐次坐标：(2) Using the estimated translation and scaling transformation coefficients, the initial centroid homogeneous coordinates of the four vertex units in the sample image are obtained by the following formula:

V′_k＝(x′_k，y′_k，1)，k＝1，2，3，4V'_k = (x'_k , y'_k , 1), k=1, 2, 3, 4

图像中4个顶点单元的边框：Border of 4 vertex cells in the image:

R′_k＝W′_k×H′_k＝S′W_k×S′H_kR'_k ＝W'_k ×H'_k ＝S'W_k ×S'H_k

${{\mathrm{<span><span>V</span>V</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>k</span>k</span>}}^{\mathrm{<span><span>T</span>T</span>}}<span><span>=</span>=</span>{\mathrm{<span><span>S</span>S</span>}}^{<span><span>\&prime;</span>\&prime;</span>}\left[\begin{array}{ccc}\mathrm{<span><span>1</span>1</span>}& \mathrm{<span><span>0</span>0</span>}& {{\mathrm{<span><span>T</span>T</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>x</span>x</span>}}\\ \mathrm{<span><span>0</span>0</span>}& \mathrm{<span><span>1</span>1</span>}& {{\mathrm{<span><span>T</span>T</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>y</span>the\; y</span>}}\\ \mathrm{<span><span>0</span>0</span>}& \mathrm{<span><span>0</span>0</span>}& \mathrm{<span><span>1</span>1</span>}\end{array}\right]{\mathrm{<span><span>V</span>V</span>}}_{\mathrm{<span><span>k</span>k</span>}}^{\mathrm{<span><span>T</span>T</span>}}$

其中参数：x_k′为样本图像的第k个顶点单元质心的x轴坐标；y_k′为样本图像的第k个顶点单元质心的y轴坐标；R_k′为样本图像的第k个顶点单元的边框；W_k′为其边框宽度；H_k′为其边框长度；T′_x为样本图像中x轴的平移因子；T′_y为样本图像中y轴的平移因子。Among them, the parameters: x_k ′ is the x-axis coordinate of the kth vertex unit mass center of the sample image; y_k ′ is the y-axis coordinate of the kth vertex unit mass center of the sample image; R_k ′ is the kth vertex of the sample image The border of the cell; W_k ′ is its border width; H_k ′ is its border length; T′_x is the translation factor of the x-axis in the sample image; T′_y is the translation factor of the y-axis in the sample image.

(3)对样本图像进行高斯平滑，然后进行1/2金字塔降采样，针对绣布的中心不妨取1/4绣布大小的矩形区域，通过彩色直方图统计出出现频率最高的颜色作为绣布背景色。(3) Gaussian smoothing is performed on the sample image, and then 1/2 pyramid downsampling is performed. For the center of the embroidered cloth, it is advisable to take a rectangular area of 1/4 the size of the embroidered cloth, and use the color histogram to count the color with the highest frequency as the embroidered cloth background color.

(4)在样本图像中，以坐标V′_k为中心，然后以1.5W_k′×1.5H_k′为长宽的矩形区域内进行11点漫水填充二值化处理。如图8(其中图8为9点漫水填充)所示，首先随机选择11个点作为备选种子点，对每个候选的种子点，在11×11邻域的区域内计算出色彩平均值，如果平均值接近绣布的背景色，而且邻域颜色方差较小，则将该点列为漫水填充种子点。(4) In the sample image, take the coordinate V′_k as the center, and then perform 11-point flood filling binarization in a rectangular area with a length and width of 1.5W_k ′×1.5H_k ′. As shown in Figure 8 (where Figure 8 is filled with 9 points of water), first, 11 points are randomly selected as candidate seed points, and for each candidate seed point, the color average is calculated in the area of 11×11 neighborhood value, if the average value is close to the background color of the embroidered cloth, and the neighborhood color variance is small, then this point will be listed as the seed point for flood filling.

(5)对每个种子点执行漫水填充算法，得到单元V′_k的二值化图像。然后，进行滑动窗口搜索，在窗口内，计算轮廓面积比和区域Hu矩，并与DST中对应顶点V_k的轮廓特征进行相似性比较，如下式所示：(5) Execute the flood filling algorithm for each seed point to obtain the binarized image of unit_V'k . Then, a sliding window search is carried out. In the window, the contour area ratio and the area Hu moment are calculated, and the similarity is compared with the contour features of the corresponding vertex V_k in DST, as shown in the following formula:

$\mathrm{<span><span>I</span>I</span>}<span><span>(</span>(</span>\mathrm{<span><span>A</span>A</span>}<span><span>,</span>,</span>\mathrm{<span><span>B</span>B</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\underset{\mathrm{<span><span>p</span>p</span>}<span><span>=</span>=</span>\mathrm{<span><span>1</span>1</span>}}{\overset{\mathrm{<span><span>7</span>7</span>}}{\mathrm{<span><span>\&Sigma;</span>\&Sigma;</span>}}}<span><span>|</span>|</span>\frac{{\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>B</span>B</span>}}}{{\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}}<span><span>|</span>|</span>$

${\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}<span><span>=</span>=</span>\mathrm{<span><span>sign</span>sign</span>}<span><span>(</span>(</span>{\mathrm{<span><span>h</span>h</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}<span><span>)</span>)</span>\mathrm{<span><span>log</span>log</span>}<span><span>|</span>|</span>{\mathrm{<span><span>h</span>h</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}<span><span>|</span>|</span>$

${\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>B</span>B</span>}}<span><span>=</span>=</span>\mathrm{<span><span>sign</span>sign</span>}<span><span>(</span>(</span>{\mathrm{<span><span>h</span>h</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>B</span>B</span>}}<span><span>)</span>)</span>\mathrm{<span><span>log</span>log</span>}<span><span>|</span>|</span>{\mathrm{<span><span>h</span>h</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>B</span>B</span>}}<span><span>|</span>|</span>$

I(A，B)＜T_huI(A,B)<T_hu

其中：A为选择的DST解析后图像的单元；B为选择的样本图像的单元；

和

分别是A和B的Hu矩。T_hu是Hu矩相似度阈值，一般取为1.5。Among them: A is the unit of the selected DST analyzed image; B is the unit of the selected sample image;

and

are the Hu moments of A and B, respectively. T_hu is the Hu moment similarity threshold, which is generally taken as 1.5.

(6)将满足面积比并且Hu矩相似度阈值的矩形区域作为单元V′_k的候选边界框，然后，将Hu矩相似度最小的候选边界作为单元V′_k的真实边界框。并得到单元的真实质心V″_k。(6) The rectangular area that satisfies the area ratio and the Hu moment similarity threshold is used as the candidate bounding box of unit_V'k , and then the candidate boundary with the smallest Hu moment similarity is taken as the real bounding box of unit_V'k . And get the true centroid V″_k of the unit.

(7)由图像中顶点向量和DST绣花文件的顶点向量

之间的夹角θ_k以及平移因子T′，可以得到旋转因子估计值R′＝θ。(7) By the vertex vector in the image and the vertex vectors of the DST embroidery file

Between the angle θ_k and the translation factor T', the estimated value of the rotation factor R'=θ can be obtained.

(8)根据DST文件中解析得到的各个DST单元边框左上角齐次坐标L_k＝(x_k，y_k，1)，和估计得到的变换因子，可以估计出图像中各个花型单元的边框左上角齐次坐标L′_k。对应的DST边框大小缩放S′倍。(8) According to the homogeneous coordinates L_k = (x_k , y_k , 1) of the upper left corner of each DST unit frame analyzed in the DST file, and the estimated transformation factor, the frame of each pattern unit in the image can be estimated The homogeneous coordinate L′_k of the upper left corner. The corresponding DST frame size is scaled by S' times.

${{\mathrm{<span><span>L</span>L</span>}}^{<span><span>\&prime;</span>\&prime;</span>}}_{\mathrm{<span><span>k</span>k</span>}}^{\mathrm{<span><span>T</span>T</span>}}<span><span>=</span>=</span>{\mathrm{<span><span>S</span>S</span>}}^{<span><span>\&prime;</span>\&prime;</span>}\left[\begin{array}{ccc}\mathrm{<span><span>cos</span>cos</span>}\mathrm{<span><span>\&theta;</span>\&theta;</span>}& <span><span>-</span>-</span>\mathrm{<span><span>sin</span>sin</span>}\mathrm{<span><span>\&theta;</span>\&theta;</span>}& {\mathrm{<span><span>T</span>T</span>}}_{\mathrm{<span><span>x</span>x</span>}}^{<span><span>\&prime;</span>\&prime;</span>}\\ \mathrm{<span><span>sin</span>sin</span>}\mathrm{<span><span>\&theta;</span>\&theta;</span>}& \mathrm{<span><span>cos</span>cos</span>}\mathrm{<span><span>\&theta;</span>\&theta;</span>}& {\mathrm{<span><span>T</span>T</span>}}_{\mathrm{<span><span>y</span>the\; y</span>}}^{<span><span>\&prime;</span>\&prime;</span>}\\ \mathrm{<span><span>0</span>0</span>}& \mathrm{<span><span>0</span>0</span>}& \mathrm{<span><span>1</span>1</span>}\end{array}\right]{\mathrm{<span><span>L</span>L</span>}}_{\mathrm{<span><span>k</span>k</span>}}^{\mathrm{<span><span>T</span>T</span>}}$

2.3.2刺绣样本的二值化2.3.2 Binarization of embroidery samples

进行单元定位后，可以把各个花型单元的基本位置确定下来，然后每个边框内部进行局部定位。不但减少了处理量，加快了处理速度，而且提高了定位精度。花型单元的二值化是准确提取每个单元内部元素轮廓特征的基础。通过对比各种二值化方法，对比二值化的结果，为了保证二值化的鲁棒性，改进了原始的漫水填充算法，设计了一种自适应的漫水填充算法：After unit positioning, the basic position of each pattern unit can be determined, and then local positioning is performed inside each frame. Not only the processing capacity is reduced, the processing speed is accelerated, but also the positioning accuracy is improved. The binarization of flower-shaped units is the basis for accurately extracting the contour features of the internal elements of each unit. By comparing various binarization methods and binarization results, in order to ensure the robustness of binarization, the original flood filling algorithm is improved, and an adaptive flood filling algorithm is designed:

1、获取绣布的主背景色。根据DST文件和相机参数，选择绣布中心区域，计算中心区域的颜色直方图，以频数最高的颜色作为绣布主颜色，即背景色；1. Get the main background color of the embroidery cloth. According to the DST file and camera parameters, select the central area of the embroidered cloth, calculate the color histogram of the central area, and use the color with the highest frequency as the main color of the embroidered cloth, that is, the background color;

2、采用9点法计算漫水填充的种子点。在图像中采样如图8所示的9个点，计算该点所在邻域窗口是否为背景色，如果是背景色，则将其纳入种子点；2. Use the 9-point method to calculate the seed point for flood filling. Sample 9 points as shown in Figure 8 in the image, calculate whether the neighborhood window where the point is located is the background color, and if it is the background color, incorporate it into the seed point;

3、自适应阈值选取。采用基本阈值分割法获得一个二值分割的阈值，并向两边计算出波峰，量化波峰间隔。通过量化间隔的离散化，实现自适应阈值；3. Adaptive threshold selection. The basic threshold segmentation method is used to obtain a binary segmentation threshold, and the peaks are calculated to both sides, and the peak interval is quantified. Through the discretization of the quantization interval, the adaptive threshold is realized;

4、以自适应阈值对保留下来的种子点进行漫水填充，并统计每个种子点的覆盖范围以及覆盖的颜色平均值；4. Fill the remaining seed points with adaptive threshold, and count the coverage of each seed point and the average color of the coverage;

5、通过HSV颜色过滤漫水填充图，实现绣样二值化。通过9点漫水填充过程中统计的种子点覆盖范围，可以进一步更新背景色。然后计算背景色的HSV颜色值，以该颜色值过滤掉二值图中的残留像素块；5. Use the HSV color filter to filter the flooded filling image to realize the binarization of the embroidery sample. The background color can be further updated through the coverage of the seed points counted during the 9-point flood filling process. Then calculate the HSV color value of the background color, and filter out the residual pixel blocks in the binary image with this color value;

6、统计二值图像中的联通域面积，联通域边界框长宽比，过滤掉部分错误区域；6. Count the area of the Unicom domain in the binary image, the aspect ratio of the bounding box of the Unicom domain, and filter out some error areas;

7、填充保留下来的轮廓，输出完整二值图。7. Fill the remaining contours and output a complete binary image.

2.3.3花型单元特征提取2.3.3 Pattern unit feature extraction

根据2.3.1中的单元定位，在矩形框内计算刺绣单元二值化后的区域Hu矩以及几何特征。According to the unit positioning in 2.3.1, calculate the area Hu moment and geometric features of the embroidery unit binarized in the rectangular frame.

3匹配3 matches

以上部分已经计算出DST单元的内部特征与待匹配的花型单元的内部特征，接下来需要进行匹配。匹配分为：The above part has calculated the internal features of the DST unit and the internal features of the pattern unit to be matched, and then needs to be matched. Matching is divided into:

1、几何特征匹配：a.包括面积过滤，将轮廓面积太大或者太小的轮廓剔除；b.将边界框长宽比不满足阈值的剔除。1. Geometric feature matching: a. Including area filtering, removing contours with too large or too small contour areas; b. Removing those whose bounding box aspect ratio does not meet the threshold.

2、轮廓匹配，对比每个花型单元与DST文件中对应单元的Hu矩，计算匹配结果。匹配公式：2. Contour matching, compare the Hu moment of each pattern unit with the corresponding unit in the DST file, and calculate the matching result. Matching formula:

$\mathrm{<span><span>I</span>I</span>}<span><span>(</span>(</span>\mathrm{<span><span>A</span>A</span>}<span><span>,</span>,</span>\mathrm{<span><span>B</span>B</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\underset{\mathrm{<span><span>p</span>p</span>}<span><span>=</span>=</span>\mathrm{<span><span>1</span>1</span>}}{\overset{\mathrm{<span><span>7</span>7</span>}}{\mathrm{<span><span>\&Sigma;</span>\&Sigma;</span>}}}<span><span>|</span>|</span>\frac{{\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>B</span>B</span>}}}{{\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}}<span><span>|</span>|</span>$

${\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}<span><span>=</span>=</span>\mathrm{<span><span>sign</span>sign</span>}<span><span>(</span>(</span>{\mathrm{<span><span>h</span>h</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}<span><span>)</span>)</span>\mathrm{<span><span>log</span>log</span>}<span><span>|</span>|</span>{\mathrm{<span><span>h</span>h</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>A</span>A</span>}}<span><span>|</span>|</span>$

${\mathrm{<span><span>m</span>m</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>B</span>B</span>}}<span><span>=</span>=</span>\mathrm{<span><span>sign</span>sign</span>}<span><span>(</span>(</span>{\mathrm{<span><span>h</span>h</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>B</span>B</span>}}<span><span>)</span>)</span>\mathrm{<span><span>log</span>log</span>}<span><span>|</span>|</span>{\mathrm{<span><span>h</span>h</span>}}_{\mathrm{<span><span>p</span>p</span>}}^{\mathrm{<span><span>B</span>B</span>}}<span><span>|</span>|</span>$

I(A，B)＜T_huI(A,B)<T_hu

并将成功匹配的单元质心保留下来，进一步做最小二乘变换；And keep the unit centroids that are successfully matched, and further perform least square transformation;

3、最小二乘变换，通过花型单元的质心和DST中各个对应单元质心坐标的最小二乘变换，可以获得一个透视变换矩阵C。3. Least square transformation, through the least square transformation of the centroid of the pattern unit and the coordinates of the centroid of each corresponding unit in DST, a perspective transformation matrix C can be obtained.

4DST映射4DST mapping

根据匹配得到的透视变换矩阵，将DST文件解析得到的切割线映射到刺绣样本图像上，并显示映射后的刺绣样本图像。本发明还配备了个人微调功能，如果切割线定位不准确，可以人工改变切割线位置。DST映射完成后，保存映射到刺绣样本图像的切割线信息，并将该信息转化到切割机坐标系上：根据坐标系标定获得的相机坐标系到切割机坐标系的变换矩阵M，X是DST中各个单元的切割线坐标，而Y是切割机坐标系的坐标According to the perspective transformation matrix obtained by matching, the cutting line obtained by parsing the DST file is mapped to the embroidery sample image, and the mapped embroidery sample image is displayed. The invention is also equipped with a personal fine-tuning function, if the cutting line is not positioned accurately, the position of the cutting line can be manually changed. After the DST mapping is completed, save the cutting line information mapped to the embroidery sample image, and convert this information to the cutting machine coordinate system: the transformation matrix M from the camera coordinate system to the cutting machine coordinate system obtained according to the coordinate system calibration, X is DST The cutting line coordinates of each unit in , and Y is the coordinate of the cutting machine coordinate system

Y＝CMXY=CMX

保存变换后的切割线图像，此时该切割线图像即为激光切割机到头的运动轨迹线。Save the transformed cutting line image, at this time, the cutting line image is the movement track line of the laser cutting machine to the head.

由于激光切割机控制卡只接受PLT格式的运动轨迹，因此需要将最后得到的切割线图像保存为PLT格式的文件。Since the laser cutting machine control card only accepts the motion trajectory in PLT format, it is necessary to save the final cutting line image as a file in PLT format.

操作实例Operation example

如图9所示为本发明所做的MFC界面，本界面主要分为五个部分：绣花样式、相机控制、图像检测、相机标定以及右上角显示框。其中，绣花样式主要用于DST文件解析，通过打开DST文件，在其下方的窗口中显示解析出来的DST图像；相机控制主要是控制样本图像输入，用于获得样本图像，通过在右上角显示框显示切割机上样本信息；图像检测是样本图像输入后的进一步处理，主要是图像匹配，切割线定位等工作；相机标定是在第一次使用该发明时用于确定相机基本参数、坐标系变换等原始信息。As shown in Fig. 9, the MFC interface made by the present invention is mainly divided into five parts: embroidery style, camera control, image detection, camera calibration and display frame in the upper right corner. Among them, the embroidery style is mainly used for DST file analysis. By opening the DST file, the parsed DST image will be displayed in the window below it; the camera control is mainly used to control the sample image input, used to obtain the sample image, by displaying the box in the upper right corner Display the sample information on the cutting machine; image detection is the further processing after the sample image is input, mainly image matching, cutting line positioning and other work; camera calibration is used to determine the basic parameters of the camera, coordinate system transformation, etc. when the invention is used for the first time original information.

图10为本发明的改进后的切割机示意图。如图中所示，将相机架设于切割面正上方1.4m处，这样可以较理想的获取绣布信息，由于光照不均匀对图像后期处理有很大的影响，因此在切割机两边架设了两盏日光灯，这样可以最大限度的降低光线不均匀对于后期处理的影响。相机的有效视野约为62.98cm*42cm，因此在放置绣布时需要考虑视野问题，在切割面上我们已经标记出理想的视野，只需要在放置绣布时大体放入标记区域即可。Fig. 10 is a schematic diagram of the improved cutting machine of the present invention. As shown in the figure, the camera is set up 1.4m directly above the cutting surface, so that the embroidery information can be obtained more ideally. Since the uneven illumination has a great impact on the post-processing of the image, two sets of cameras are set up on both sides of the cutting machine. A fluorescent lamp, which can minimize the impact of uneven light on post-processing. The effective field of view of the camera is about 62.98cm*42cm, so the field of view needs to be considered when placing the embroidery cloth. We have marked the ideal field of view on the cutting surface, and only need to roughly place it in the marked area when placing the embroidery cloth.

下面给出一个实例的操作步骤：The following is an example of the operation steps:

(1)在第一次使用本发明时需要进行标定。在MFC界面的右下角有两个按钮用于标定，通过点击这两个按钮完成标定，获得系统的原始信息。(1) Calibration is required when using the present invention for the first time. There are two buttons in the lower right corner of the MFC interface for calibration. Click these two buttons to complete the calibration and obtain the original information of the system.

(2)将刺绣样本放入切割机的工作面上，在工作面上印有工作区的边界线，每次放置绣布时只要左上角大致对齐边界线即可保证绣布位置在有效检测范围。(2) Put the embroidery sample on the working surface of the cutting machine, and the boundary line of the working area is printed on the working surface. When placing the embroidered cloth, as long as the upper left corner is roughly aligned with the boundary line, the position of the embroidered cloth is within the effective detection range. .

(3)点击“打开”按钮，输入与之对应的DST文件并解析，在该部分的显示窗口中显示解析后的DST图像。(3) Click the "Open" button, input the corresponding DST file and analyze it, and display the analyzed DST image in the display window of this part.

(4)在相机控制部分点击“取景”按钮，在右上方的显示框内显示相机的视野，通过观察判断刺绣样本是否放在有效视野中。(4) Click the "viewfinder" button in the camera control part, and the field of view of the camera will be displayed in the display frame on the upper right, and judge whether the embroidery sample is placed in the effective field of view through observation.

(5)点击“拍照”按钮用于将样本图像输入进系统。(5) Click the "photograph" button to input the sample image into the system.

(6)此时点击“位置检测”按钮，获得DST图像与刺绣样本的匹配结果，如图11所示。图中的红色方框，表示匹配的结果，即寻找到特征区域。(6) At this point, click the "Position Detection" button to obtain the matching result between the DST image and the embroidery sample, as shown in Figure 11. The red box in the figure indicates the matching result, that is, the feature area is found.

(7)点击“切割线检测”得到图12、图13中显示的图像，图像根据匹配结果得到的切割线图像，通过观测切割线定位的结果判断是否合格。(7) Click "cutting line detection" to get the images shown in Fig. 12 and Fig. 13, and judge whether the image is qualified or not by observing the result of cutting line positioning based on the cutting line image obtained from the matching result.

(8)判断切割线合格后，点击“导出”按钮导出切割机可以读取的PLT格式文件，开始切割，切割过程如图14所示。(8) After judging that the cutting line is qualified, click the "Export" button to export the PLT format file that the cutting machine can read, and start cutting. The cutting process is shown in Figure 14.

(9)图15为经过激光切割后的布料。(9) Figure 15 shows the fabric after laser cutting.

Claims (5)

Translated fromChinese

1.一种基于DST文件解析与机器视觉结合的绣布自动切割方法，其特征在于该方法用到一种激光切割机，该激光切割机切割面正上方架设有相机，所述绣布自动切割方法包括如下步骤：1. An embroidered cloth automatic cutting method based on DST file parsing and machine vision combination, it is characterized in that the method uses a kind of laser cutting machine, this laser cutting machine is set up with camera directly above the cutting surface, and described embroidered cloth is automatically cut The method includes the following steps:（1）DST文件解析，获得原始刺绣文件图像信息和内部特征信息，为图像匹配做准备工作；(1) DST file analysis, obtaining original embroidery file image information and internal feature information, and preparing for image matching;（2）样本提取与处理，通过相机拍照获得样本图像信息，并对样本图像进行畸变矫正、二值化处理以及提取样本内部特征操作，获得样本的特征信息；(2) Sample extraction and processing. The sample image information is obtained by taking photos with the camera, and the sample image is subjected to distortion correction, binarization processing, and internal feature extraction of the sample to obtain the feature information of the sample;（3）匹配，将得到的DST文件与样本的特征信息进行匹配，匹配单元质心并根据最小二乘法进行质心匹配与切割线校正；(3) Matching, matching the obtained DST file with the characteristic information of the sample, matching the centroid of the unit, and performing centroid matching and cutting line correction according to the least square method;（4）DST映射，将匹配后的切割线坐标通过坐标变换映射到切割面坐标系下用于切割，然后输出激光切割机能够执行的PLT文件，由激光切割机进行切割。(4) DST mapping, the coordinates of the matched cutting line are mapped to the coordinate system of the cutting surface for cutting through coordinate transformation, and then output the PLT file that the laser cutting machine can execute, and the laser cutting machine cuts.2.如权利要求1所述的基于DST文件解析与机器视觉结合的绣布自动切割方法，其特征在于DST文件解析的过程包括：2. the embroidered cloth automatic cutting method based on DST file parsing and machine vision as claimed in claim 1 is characterized in that the process of DST file parsing comprises:（1.1）根据DST文件的编码方式解析每一针的功能，记录每一针的偏移坐标与操作，并且记录解析过程中所有出现的闭合数据段；(1.1) Analyze the function of each stitch according to the encoding method of the DST file, record the offset coordinates and operations of each stitch, and record all closed data segments that appear during the analysis process;（1.2）根据（1.1）解析得到的结果，创建一幅空白图像，在图像上将每一针的具体动作勾画出来，得到DST解析图像；(1.2) Create a blank image based on the analysis result of (1.1), draw the specific action of each needle on the image, and obtain the DST analysis image;（1.3）根据（1.1）得到的闭合数据段，判断闭合数据段是否为DST单元的边框，如果该闭合数据段是边界框，则提取其内部的信息作为DST单元的信息，如果不是则继续判断，直到得到DST单元的边界框。(1.3) According to the closed data segment obtained in (1.1), judge whether the closed data segment is the border of the DST unit. If the closed data segment is a bounding box, extract its internal information as the information of the DST unit. If not, continue to judge , until the bounding box of the DST cell is obtained.3.如权利要求1所述的基于DST文件解析与机器视觉结合的绣布自动切割方法，其特征在于从相机坐标系映射到切割机坐标系的变换矩阵采用“基于最小二乘法圆检测的坐标系标定法”获得，步骤如下：3. The embroidered cloth automatic cutting method based on DST file parsing and machine vision as claimed in claim 1 is characterized in that the transformation matrix mapped to the cutting machine coordinate system from the camera coordinate system adopts "coordinates detected based on the least squares method circle". system calibration method", the steps are as follows:（1）首先设计好用于切割标定圆的PLT文件；(1) First design the PLT file for cutting the calibration circle;（2）从PLT文件中获取切割机坐标系下精确的圆心坐标                                               

，其中

、

分别表示第i行第j列的圆的圆心在切割机坐标系上的x轴坐标和y轴坐标； (2) Obtain the precise circle center coordinates in the cutting machine coordinate system from the PLT file

,in

,

Respectively represent thex -axis coordinates and y-axis coordinates of the circle center of thei-th row and j-th column on the cutting machine coordinate system;（3）利用相机拍摄一张清晰的标定圆图像，然后利用已有的畸变参数，进行图像矫正；(3) Use the camera to take a clear calibration circle image, and then use the existing distortion parameters to correct the image;（4）采用漫水填充法对矫正后的图像进行二值化处理，将图像中的圆变成白色，背景变为黑色；(4) Binarize the rectified image by flood filling method, turning the circle in the image into white and the background into black;（5）对二值化后图像进行轮廓提取，检测出每个标定圆的轮廓点序列；(5) Extract the contour of the binarized image, and detect the contour point sequence of each calibration circle;（6）对每个轮廓点序列进行RANSAC圆拟合，获得相机坐标系下标定圆圆心坐标

，其中

、

分别表示第i行第j列的圆的圆心在相机坐标系上的x轴坐标和y轴坐标； (6) Perform RANSAC circle fitting on each contour point sequence to obtain the coordinates of the center of the calibration circle in the camera coordinate system

,in

,

Respectively represent thex- axis coordinates and y-axis coordinates of the circle center of thei-th row and j-th column on the camera coordinate system;（7）将切割机坐标系下的圆心坐标按照从左到右，从上往下的形式排列为齐次坐标矩阵P，将相机坐标系下圆心坐标也按照对应的次序排列为齐次坐标矩阵Q；(7) Arrange the coordinates of the center of the circle under the cutting machine coordinate system into a homogeneous coordinate matrix P from left to right and from top to bottom, and arrange the coordinates of the center of the circle under the camera coordinate system into a homogeneous coordinate matrix according to the corresponding order Q;（8）采用最小二乘法，计算出从相机坐标系映射到切割机坐标系的变换矩阵M，使之满足：Q*M=P。(8) Use the least square method to calculate the transformation matrix M mapped from the camera coordinate system to the cutting machine coordinate system, so that it satisfies: Q*M=P.4.如权利要求1所述的基于DST文件解析与机器视觉结合的绣布自动切割方法，其特征在于样本提取与处理过程中，采用改进的自适应漫水填充法对样本图像进行二值化处理，即通过在刺绣样本中随机选择多个点作为样本点，判断这些随机选择的点是否为背景点，如果是则保留当作种子点，如果不是则忽略。4. The embroidered cloth automatic cutting method based on DST file parsing and machine vision as claimed in claim 1 is characterized in that in the sample extraction and processing process, the sample image is binarized by using the improved adaptive flood filling method Processing, that is, by randomly selecting multiple points in the embroidery sample as sample points, judging whether these randomly selected points are background points, if they are, they will be kept as seed points, and if not, they will be ignored.5.如权利要求1所述的基于DST文件解析与机器视觉结合的绣布自动切割方法，其特征在于为了减小光照不均匀对图像后期处理产生的影响，在切割机两侧分别加装一盏日光灯。5. The embroidered cloth automatic cutting method based on DST file parsing and machine vision as claimed in claim 1 is characterized in that in order to reduce the impact of uneven illumination on image post-processing, install a a fluorescent lamp.

Images