CN102663348B - Marine ship detection method in optical remote sensing image - Google Patents
Marine ship detection method in optical remote sensing imageDownload PDFInfo
- Publication number
- CN102663348B CN102663348BCN 201210077407CN201210077407ACN102663348BCN 102663348 BCN102663348 BCN 102663348BCN 201210077407CN201210077407CN 201210077407CN 201210077407 ACN201210077407 ACN 201210077407ACN 102663348 BCN102663348 BCN 102663348B
- Authority
- CN
- China
- Prior art keywords
- target
- image
- ship
- suspected
- ship target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000001514detection methodMethods0.000titleclaimsabstractdescription37
- 230000003287optical effectEffects0.000titleabstractdescription14
- 238000000034methodMethods0.000claimsabstractdescription22
- 239000000284extractSubstances0.000claimsabstractdescription4
- 238000012549trainingMethods0.000claimsdescription12
- 230000000007visual effectEffects0.000claimsdescription11
- 238000004422calculation algorithmMethods0.000claimsdescription9
- 238000012706support-vector machineMethods0.000claimsdescription9
- 230000011218segmentationEffects0.000claimsdescription7
- 238000000605extractionMethods0.000claimsdescription4
- 238000006243chemical reactionMethods0.000claimsdescription2
- 239000004744fabricSubstances0.000claims1
- 238000012795verificationMethods0.000claims1
- 238000012790confirmationMethods0.000description6
- 238000004088simulationMethods0.000description6
- 238000012360testing methodMethods0.000description5
- 238000010586diagramMethods0.000description4
- 238000003384imaging methodMethods0.000description3
- 238000004458analytical methodMethods0.000description2
- 230000009286beneficial effectEffects0.000description1
- 238000007635classification algorithmMethods0.000description1
- 238000013145classification modelMethods0.000description1
- 239000011248coating agentSubstances0.000description1
- 238000000576coating methodMethods0.000description1
- 238000002474experimental methodMethods0.000description1
- 238000010191image analysisMethods0.000description1
- 238000003909pattern recognitionMethods0.000description1
- 238000012545processingMethods0.000description1
Images
Landscapes
- Image Analysis (AREA)
Abstract
Translated fromChinese
Description
Translated fromChinese技术领域technical field
本发明涉及遥感图像分析领域的智能化舰船目标检测技术，更具体地说，涉及一种复杂海面条件下的光学遥感图像海上舰船检测方法。The invention relates to an intelligent ship target detection technology in the field of remote sensing image analysis, and more specifically, relates to a sea ship detection method of an optical remote sensing image under complex sea surface conditions.
背景技术Background technique
在光学遥感图像中，复杂海面背景情况下的海上舰船目标检测问题一直是难点。一方面，由于成像器件、大气、拍摄角度、时间、气象等诸多因素影响，以及不同的海面波浪状况对光照的反射能力不同，使得光学遥感图像的亮度、对比度等信息存在很大变化，海面背景具有不稳定性，平均亮度存在起伏，其高频信息在幅度上受海浪、航迹的影响变化很大。具体到舰船检测，由于受光照、舰船表面涂层的影响，舰船目标灰度表现不确定性，可见光舰船目标在亮度上可能高于或低于海面背景亮度，分别称为舰船的白极性表现和黑极性表现。此时，传统的基于阈值分割的检测方法无法选择一个合适的阈值将目标与背景分离，造成较高的虚警率。In optical remote sensing images, the problem of ship target detection in the complex sea background has always been a difficult point. On the one hand, due to the influence of many factors such as imaging devices, atmosphere, shooting angle, time, weather, etc., as well as the different reflection capabilities of different sea surface waves to light, the brightness, contrast and other information of optical remote sensing images vary greatly. It is unstable, the average brightness fluctuates, and the amplitude of its high-frequency information is greatly changed by the influence of waves and track. Specific to ship detection, due to the influence of light and ship surface coating, the gray scale performance of ship targets is uncertain, and the brightness of visible light ship targets may be higher or lower than the background brightness of the sea surface, which are called ship targets respectively. White polarity performance and black polarity performance. At this time, the traditional detection method based on threshold segmentation cannot select an appropriate threshold to separate the target from the background, resulting in a high false alarm rate.
另一方面，由于可见光成像容易受天气等因素的影响，所以可见光图像中往往存在大量的云等情况，舰船检测往往受到云、海浪等干扰，导致目前的舰船检测方法往往虚警较多，虽然在目前一些海上舰船检测方法在海面区域目标的粗检测得到疑似目标后，增加对舰船目标候选区域或舰船疑似目标的目标自身特征进行分析确认，去除部分虚警，获取舰船检测结果。这些方法中常用的舰船目标自身特征有：灰度、尺寸、形状、纹理特征，在提取目标自身多种特征描述后，采用分类器对疑似舰船目标进行分类确认识别。然而，对于云、海浪、海岛等疑似目标干扰，有时在疑似目标块上提取出的目标特征与实际舰船目标非常相似，并且不同分辨率、不同时相图像中舰船目标反映出的特征又有差异，为舰船的确认识别带来很多困难，所以目前基于目标自身特征分析方法仍然存在较多虚警，阻碍了可见光图像海上舰船检测方法的应用。On the other hand, since visible light imaging is easily affected by factors such as weather, there are often a large number of clouds in visible light images, and ship detection is often interfered by clouds, waves, etc., resulting in more false alarms in current ship detection methods. , although some current maritime ship detection methods have obtained suspected targets through rough detection of targets in the sea surface area, they will increase the analysis and confirmation of the target’s own characteristics in the target candidate area of the ship or the suspected target of the ship, remove some false alarms, and obtain ship Test results. The characteristics of ship targets commonly used in these methods include: grayscale, size, shape, and texture features. After extracting multiple feature descriptions of the target itself, a classifier is used to classify and identify suspected ship targets. However, for the interference of suspected targets such as clouds, waves, and islands, sometimes the target features extracted from the suspected target block are very similar to the actual ship target, and the features reflected by the ship target in images of different resolutions and different time phases are different. There are differences, which bring many difficulties to the confirmation and identification of ships. Therefore, there are still many false alarms in the current analysis method based on the characteristics of the target itself, which hinders the application of the detection method of ships at sea in visible light images.
发明内容Contents of the invention
本发明为了有效解决复杂海面光学遥感图像的海上舰船检测问题，提供了一种基于局部对比度信息和空间金字塔特征的光学遥感图像海上舰船检测方法。本方法有效抑制了舰船的白极性表现和黑极性表现问题，同时对于舰船目标与其他干扰的相似性问题以及舰船目标本身具有的差异性问题，本方法引入目标的局部邻域上下文信息用于舰船的特征描述与识别，区分目标与背景干扰，有效抑制舰船检测的虚警率。In order to effectively solve the problem of detecting ships at sea in optical remote sensing images of complex sea surfaces, the present invention provides a method for detecting ships at sea in optical remote sensing images based on local contrast information and spatial pyramid features. This method effectively suppresses the problem of white polarity and black polarity performance of the ship. At the same time, for the similarity between the ship target and other interference and the difference of the ship target itself, this method introduces the local neighborhood of the target Context information is used for feature description and recognition of ships, to distinguish targets from background interference, and to effectively suppress the false alarm rate of ship detection.
本发明的基本思路是：首先，针对舰船目标的黑白极性问题，基于局部对比度在海面区域滑动窗口进行舰船的疑似目标检测，减少舰船检测的漏警；然后，对分割得到的疑似目标区域按一定窗口大小取其邻域，利用空间金字塔匹配模型提取空间上下文信息来进行分类，删除背景干扰，获取舰船检测结果，减少舰船检测的虚警。The basic idea of the present invention is: firstly, aiming at the black-and-white polarity problem of the ship target, the suspected target detection of the ship is carried out in the sliding window of the sea surface area based on the local contrast, and the false alarm of the ship detection is reduced; Select the neighborhood of the target area according to a certain window size, use the spatial pyramid matching model to extract spatial context information for classification, delete background interference, obtain ship detection results, and reduce false alarms in ship detection.
本发明的技术方案是：一种光学遥感图像海上舰船目标检测方法，具体包括下述步骤：The technical solution of the present invention is: a method for detecting a ship target in an optical remote sensing image at sea, specifically comprising the following steps:
已知：一幅输入图像I1是光学遥感图像。It is known that an input image I1 is an optical remote sensing image.
第一步：海陆区域分割Step 1: Segmentation of sea and land areas
对输入图像I1进行海面区域与陆地区域的分割，得到对海面区域进行标记的海域图像I2。Segment the sea surface area and the land area on the input image I1 to obtain the sea area image I2 marked with the sea surface area.
第二步：疑似舰船目标的粗检测Step 2: Coarse detection of suspected ship targets
对海域图像I2，采用修订了判决准则的Contrast Box算法进行处理，检测得到一组包含疑似舰船目标的矩形区域，称之为疑似舰船目标区域。For the sea area image I2, the Contrast Box algorithm with revised judgment criteria is used to process, and a group of rectangular areas containing suspected ship targets are detected, which are called suspected ship target areas.
特别的，在Contrast Box算法中定义如下判决准则来进行判断：In particular, the following judgment criteria are defined in the Contrast Box algorithm for judgment:
|(μT-μB)|/δB＞K (1)|(μT -μB )|/δB ＞K (1)
其中μT表示Contrast Box算法中目标窗口的灰度均值，μB表示背景窗口的灰度均值，δB表示背景窗口的灰度标准差，K为检测阈值。满足准则(1)则认为是疑似目标点，对多个相邻的疑似目标点所组成的区域求得最小面积外接矩形即是疑似舰船目标区域，通常选择包含2个以上相邻疑似目标点的区域作为疑似舰船目标区域。Among them, μT represents the average gray value of the target window in the Contrast Box algorithm, μB represents the average gray value of the background window, δB represents the gray standard deviation of the background window, and K is the detection threshold. If the criterion (1) is met, it is considered as a suspected target point. The circumscribed rectangle with the minimum area is obtained for the area composed of multiple adjacent suspected target points, which is the suspected ship target area. Usually, it is selected to contain more than two adjacent suspected target points. The area is regarded as the suspected ship target area.
第三步：疑似舰船目标的上下文特征提取Step 3: Context feature extraction of suspected ship targets
对每一个疑似舰船目标区域使用如下方法得到疑似目标邻域图像块：以疑似舰船目标区域的中心为中心，长、宽分别为疑似舰船目标区域长、宽的2倍大小，得到包含疑似目标及其邻域的矩形区域，称之为疑似目标邻域图像块。对疑似目标邻域图像块利用空间金字塔匹配模型提取疑似舰船目标的空间上下文特征。For each suspected ship target area, the following method is used to obtain the suspected target neighborhood image block: take the center of the suspected ship target area as the center, and the length and width are twice the length and width of the suspected ship target area, respectively. The rectangular area of the suspected target and its neighborhood is called the suspected target neighborhood image block. The spatial context features of the suspected ship target are extracted by using the spatial pyramid matching model for the neighborhood image block of the suspected target.
第四步：疑似舰船目标识别确认Step 4: Identification and confirmation of suspected ship targets
对疑似舰船目标的空间上下文特征，利用基于直方图交叉核的SVM(supportvector machine)分类器进行分类，得到该疑似舰船目标是否是舰船的确认结果。For the spatial context features of the suspected ship target, the SVM (support vector machine) classifier based on the histogram intersection kernel is used to classify, and the confirmation result of whether the suspected ship target is a ship is obtained.
本发明的有益效果是：一方面，根据舰船成像特性重新定义了Contrast Box算法的判断准则，从而克服了舰船在海域中存在的黑白极性问题，减小了目标漏警率。另一方面，在获取到疑似舰船目标后，不同于传统的只对舰船目标本身提取特征进行分析的方法，本发明针对舰船目标及其邻域提取空间上下文特征进行分析，一定程度上解决了舰船目标与虚警之间特征可区分性不强的问题，可直接用于光学遥感图像的海上舰船目标检测。The beneficial effects of the present invention are: on the one hand, the judging criterion of the Contrast Box algorithm is redefined according to the imaging characteristics of the ship, thereby overcoming the problem of black and white polarity of the ship in the sea area, and reducing the target false alarm rate. On the other hand, after obtaining the suspected ship target, different from the traditional method of analyzing only the feature extraction of the ship target itself, the present invention analyzes the space context features of the ship target and its neighborhood, to a certain extent The problem of weak distinguishability between ship targets and false alarms is solved, and it can be directly used for maritime ship target detection in optical remote sensing images.
附图说明Description of drawings
图1为本发明所提供的光学遥感图像中的海上舰船检测方法流程图；Fig. 1 is the flow chart of the ship detection method at sea in the optical remote sensing image provided by the present invention;
图2为仿真实验中第二步进行疑似舰船目标粗检测示意图；Figure 2 is a schematic diagram of the rough detection of suspected ship targets in the second step of the simulation experiment;
图3为仿真实验中第三步疑似舰船目标的空间上下文特征描述方法示意图；Figure 3 is a schematic diagram of the spatial context feature description method of the third step of the suspected ship target in the simulation experiment;
图4为仿真实验中第四步对空间上下文特征利用基于直方图交叉核的SVM分类器进行分类的流程图。Fig. 4 is a flowchart of the fourth step in the simulation experiment to classify the spatial context features using the SVM classifier based on the histogram intersection kernel.
具体实施方式Detailed ways
下面结合附图对本发明提供的光学遥感图像舰船目标检测方法进行详细说明。The method for detecting ship targets in optical remote sensing images provided by the present invention will be described in detail below in conjunction with the accompanying drawings.
图1是本发明所提供的光学遥感图像中的海上舰船检测方法流程图。该流程图的第一步是海陆区域分割，通过海陆区域分割获取海面区域，首先基于海面与陆地的灰度差异根据OTSU方法确定二值化分割阈值，得到海域和陆地初始分割，再在海域中选择种子点采用区域生长法得到海面区域。第二步疑似舰船目标的粗检测，是利用修订了判决准则的Contrast Box算法在海面区域中逐像素滑动窗口检测得到一组疑似舰船目标区域，Contrast Box算法具体实现详见文章“CASASENT D.P.SU W.，TURAGA D.，et al，SAR ship detection using newconditional contrast box filter[C]，SPIE，1999，372l：274-284.”。第三步疑似舰船目标的上下文特征提取，对疑似舰船目标区域获取疑似目标邻域图像块，利用空间金字塔匹配模型提取空间上下文特征。第四步疑似舰船目标识别确认，是使用基于直方图交叉核的SVM分类器对疑似舰船目标的空间上下文特征进行分类得到确认结果。Fig. 1 is a flow chart of a method for detecting ships at sea in optical remote sensing images provided by the present invention. The first step of the flow chart is the segmentation of sea and land areas. The sea area is obtained through the segmentation of sea and land areas. First, based on the gray level difference between the sea surface and land, the binary segmentation threshold is determined according to the OTSU method, and the initial segmentation of sea area and land is obtained. Then, in the sea area Select the seed point and use the region growing method to obtain the sea surface area. The second step is the rough detection of suspected ship targets, which is to use the Contrast Box algorithm with revised judgment criteria to detect a group of suspected ship target areas pixel by pixel in the sea surface area. SU W., TURAGA D., et al, SAR ship detection using newconditional contrast box filter[C], SPIE, 1999, 372l: 274-284.". The third step is the context feature extraction of the suspected ship target. The suspected target neighborhood image block is obtained for the suspected ship target area, and the spatial context feature is extracted by using the spatial pyramid matching model. The fourth step is to identify and confirm the suspected ship target, which is to use the SVM classifier based on the histogram intersection kernel to classify the spatial context features of the suspected ship target to obtain the confirmation result.
图2是仿真实验中第二步进行疑似舰船目标粗检测示意图。图2(a)为输入的一幅光学遥感图像的海域图像，为了表示清楚，在该海域图像中指示出舰船、海岛和云，图2(b)所示为通过检测得到的4个疑似舰船目标区域，图2(c)所示为分别对4个疑似舰船目标区域获取到的疑似目标邻域图像块。在利用修订了判决准则的Contrast Box算法进行疑似舰船目标的粗检测中，目标窗口T的大小为遥感图像中最大舰船目标尺寸，为保证背景窗口覆盖足够充分的背景特征数据，取背景窗口B为目标窗口T面积的4倍。检测阈值K控制检测虚警率，通常取1.25。Figure 2 is a schematic diagram of the rough detection of suspected ship targets in the second step of the simulation experiment. Figure 2(a) is an input sea area image of an optical remote sensing image. In order to show clearly, ships, islands and clouds are indicated in the sea area image. Figure 2(b) shows four suspected Ship target area, Figure 2(c) shows the suspected target neighborhood image blocks obtained for four suspected ship target areas. In the coarse detection of suspected ship targets using the Contrast Box algorithm with revised judgment criteria, the size of the target window T is the largest ship target size in the remote sensing image. In order to ensure that the background window covers sufficient background feature data, the background window T is taken as B is 4 times the area of the target window T. The detection threshold K controls the detection false alarm rate, usually 1.25.
图3是仿真实验中第三步疑似舰船目标的空间上下文特征描述方法示意图。首先对疑似目标邻域图像块进行规则网格的划分，再利用空间金字塔匹配模型提取疑似舰船目标的空间上下文特征。具体描述如下：Fig. 3 is a schematic diagram of the spatial context feature description method of the suspected ship target in the third step of the simulation experiment. Firstly, the image blocks in the neighborhood of the suspected target are divided into regular grids, and then the spatial context features of the suspected ship target are extracted using the spatial pyramid matching model. The specific description is as follows:
图3左侧图像为检测到的一幅疑似目标邻域图像块，对其进行规则的均匀网格分割，如图3中间图像所示，称之为图像块节点，再提取每个图像块节点的SIFT(Scale Invariant Feature Transform)特征，SIFT具体实现详见文章“DavidG.Lowe，Distinctive image features from scale-invariant keypoints.InternationalJournal of Computer Vision，200460(2)：91-110”。图像块节点尺寸根据遥感图像分辨率的高低不同来确定，中高分辨率可取16×16像素、8×8像素，低分辨率可取4×4像素。同时随机选出一部分疑似目标邻域图像块作为训练图像，将这些训练图像的各个图像块节点所提取的SIFT特征进行K均值聚类，实验中K均值聚类数目，即视觉词汇数目设为100，得到图像的视觉词汇码本。然后对所有疑似目标邻域图像块的图像块节点提取的SIFT特征都按照此视觉词汇码本进行量化。这样每个图像块节点将对应一个视觉词汇。从而完成了图像的像素空间表示到图像的视觉词汇空间表示的转换，转换得到的图称之为词汇图，图3右侧图像为输入的疑似目标邻域图像块转换得到的词汇图。The image on the left side of Figure 3 is a detected image block of a suspected target neighborhood, which is divided into a regular uniform grid, as shown in the middle image of Figure 3, which is called an image block node, and then each image block node is extracted The SIFT (Scale Invariant Feature Transform) feature of SIFT, the specific implementation of SIFT can be found in the article "David G. Lowe, Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 200460(2): 91-110". The size of the image block node is determined according to the resolution of the remote sensing image. The medium and high resolution can be 16×16 pixels and 8×8 pixels, and the low resolution can be 4×4 pixels. At the same time, a part of the suspected target neighborhood image blocks are randomly selected as training images, and the SIFT features extracted from each image block node of these training images are clustered by K-means. In the experiment, the number of K-means clusters, that is, the number of visual words, is set to 100 , to get the visual vocabulary codebook of the image. Then, the SIFT features extracted from the image block nodes of all suspected target neighborhood image blocks are quantized according to this visual vocabulary codebook. In this way, each image block node will correspond to a visual vocabulary. Thus, the conversion from the pixel space representation of the image to the visual vocabulary space representation of the image is completed, and the converted graph is called a vocabulary graph. The image on the right side of Figure 3 is the vocabulary graph obtained by converting the input suspected target neighborhood image block.
对于得到的词汇图再利用空间金字塔匹配模型获取疑似舰船目标的空间上下文特征。其中空间金字塔匹配模型具体详见文献“Lazibnik etc.Beyond Bagsof Features：Spatial Pyramid Matching for Recognizing Natural Scene Categories.Proceedings of IEEE Computer Society Conference on Computer Vision and PatternRecognition.New York：2006：2169-2178”。记得到的空间上下文特征为Pw。For the obtained vocabulary graph, the spatial context features of suspected ship targets are obtained by using the spatial pyramid matching model. For details on the spatial pyramid matching model, see the literature "Lazibnik etc. Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene Categories. Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition. New York: 2006: 2169-2178". The spatial context feature recalled is Pw .
图4是本发明仿真实验中第四步对空间上下文特征利用基于直方图交叉核的SVM分类器进行分类的流程图。该分类算法流程分为训练和测试两个部分，图4垂直方向的虚线左侧为训练阶段，右侧为测试阶段。样本库共包含海岛、云、海浪和舰船这四类样本。在训练阶段，训练样本库由每一类样本中随机抽取一部分疑似目标邻域图像块构成，图4虚线左侧最上方为输入的训练样本示意，然后根据第三步方法获得视觉词汇码本，提取训练样本库中的所有疑似目标邻域图像块的空间上下文特征Pw，再基于直方图交叉核函数由训练样本对SVM分类器进行训练，获取SVM分类模型。图4虚线右侧最上方为输入的一幅疑似目标邻域图像块，在测试阶段，对疑似目标邻域图像块根据训练阶段得到的视觉词汇码本，得到其空间上下文特征，再利用SVM分类模型对测试样本进行分类，得到疑似目标是否是舰船目标的确认结果。其中，直方图交叉核函数详见文章“Barla A，Odone F，and Verri A.Histogram intersection kernel for imageclassification[C].Proceedings of the International Conference on Image Processing，Barcelona，Catalonia，Spain，Sept.14-17，2003，Vol.2：513-516.”。Fig. 4 is a flowchart of the fourth step in the simulation experiment of the present invention to classify the spatial context features using the SVM classifier based on the histogram intersection kernel. The classification algorithm process is divided into two parts: training and testing. The left side of the dotted line in the vertical direction in Figure 4 is the training phase, and the right side is the testing phase. The sample library contains four types of samples: islands, clouds, waves and ships. In the training phase, the training sample library is composed of a part of the suspected target neighborhood image blocks randomly selected from each type of sample. The top left of the dotted line in Figure 4 is the input training sample, and then the visual vocabulary codebook is obtained according to the third step method. Extract the spatial context features Pw of all suspected target neighborhood image blocks in the training sample library, and then train the SVM classifier from the training samples based on the histogram intersection kernel function to obtain the SVM classification model. The uppermost part on the right side of the dotted line in Figure 4 is an input image block of a suspected target neighborhood. In the test phase, the spatial context features of the image block of the suspected target neighborhood are obtained according to the visual vocabulary codebook obtained in the training phase, and then classified by SVM The model classifies the test samples and obtains the confirmation result of whether the suspected target is a ship target. Among them, the histogram intersection kernel function is detailed in the article "Barla A, Odone F, and Verri A.Histogram intersection kernel for imageclassification[C].Proceedings of the International Conference on Image Processing, Barcelona, Catalonia, Spain, Sept.14-17 , 2003, Vol.2:513-516.".
Claims (2)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201210077407CN102663348B (en) | 2012-03-21 | 2012-03-21 | Marine ship detection method in optical remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201210077407CN102663348B (en) | 2012-03-21 | 2012-03-21 | Marine ship detection method in optical remote sensing image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102663348A CN102663348A (en) | 2012-09-12 |
| CN102663348Btrue CN102663348B (en) | 2013-10-16 |
Family
ID=46772832
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201210077407Expired - Fee RelatedCN102663348B (en) | 2012-03-21 | 2012-03-21 | Marine ship detection method in optical remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102663348B (en) |
Families Citing this family (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102867196B (en)* | 2012-09-13 | 2015-10-21 | 武汉大学 | Based on the complicated sea remote sensing image Ship Detection of Gist feature learning |
| CN103020592B (en)* | 2012-11-22 | 2015-09-23 | 北京大学 | A kind of remote sensing images naval vessel based on local context detects false-alarm elimination method |
| CN103150567B (en)* | 2013-02-04 | 2016-01-20 | 北京理工大学 | The remote sensing image cloud method of discrimination of integration mode identification and Context Knowledge |
| CN103150582B (en)* | 2013-02-04 | 2015-11-25 | 北京理工大学 | Modified result method is sentenced based on contextual remote sensing image cloud |
| CN103544502A (en)* | 2013-10-29 | 2014-01-29 | 上海市城市建设设计研究总院 | High-resolution remote-sensing image ship extraction method based on SVM |
| CN103617628A (en)* | 2013-12-16 | 2014-03-05 | 北京航空航天大学 | Naval ship detection method based on polarized azimuth secant function characteristic |
| CN103996017B (en)* | 2014-02-24 | 2018-08-31 | 航天恒星科技有限公司 | One kind being based on HuThe not Ship Detection of bending moment and support vector machines |
| CN104166853B (en)* | 2014-04-24 | 2019-06-25 | 中国人民解放军海军航空大学 | The method that rapidly extracting regularization naval vessel is sliced from high-resolution remote sensing image |
| CN103984945A (en)* | 2014-05-14 | 2014-08-13 | 武汉大学 | Optical remote sensing image ship detection method |
| CN104182731B (en)* | 2014-08-12 | 2017-05-17 | 中国科学技术大学 | A grid-shaped radar detection method based on cross detection |
| CN105095846B (en)* | 2014-09-28 | 2019-07-30 | 航天恒星科技有限公司 | Region growing seed point extracting method and system towards the segmentation of remote sensing images sea land |
| CN104992140B (en)* | 2015-05-27 | 2019-07-26 | 上海海事大学 | A method for detecting abnormal floating objects on sea surface based on remote sensing images |
| CN105139034B (en)* | 2015-08-27 | 2019-06-28 | 北京市遥感信息研究所 | A kind of ship detecting method that combination spectrum filters out |
| CN105303526B (en)* | 2015-09-17 | 2017-11-07 | 哈尔滨工业大学 | A kind of Ship Target Detection method based on coastline data and spectrum analysis |
| CN105976345A (en)* | 2016-04-27 | 2016-09-28 | 北京航空航天大学 | Visible light remote sensing image synthesis method |
| CN106056084B (en)* | 2016-06-01 | 2019-05-24 | 北方工业大学 | A Multi-resolution Hierarchical Screening Method for Port Ship Detection in Remote Sensing Images |
| CN106295498B (en)* | 2016-07-20 | 2019-04-16 | 湖南大学 | Remote sensing image target area detection device and method |
| CN106709426B (en)* | 2016-11-29 | 2020-04-14 | 上海航天测控通信研究所 | Ship target detection method based on infrared remote sensing image |
| CN106815553B (en)* | 2016-12-13 | 2019-07-23 | 华中科技大学 | A kind of infrared forward sight image Ship Detection based on edge matching |
| CN106845372B (en)* | 2016-12-31 | 2018-01-26 | 华中科技大学 | A ship target detection and recognition method and system for spaceborne remote sensing optical images |
| CN107704865A (en)* | 2017-05-09 | 2018-02-16 | 北京航空航天大学 | Fleet Targets Detection based on the extraction of structure forest edge candidate region |
| CN108229342B (en)* | 2017-12-18 | 2021-10-26 | 西南技术物理研究所 | Automatic sea surface ship target detection method |
| CN108376247A (en)* | 2018-02-05 | 2018-08-07 | 北方工业大学 | A split-strategy rough-fine combined sea-land separation method applied to optical remote sensing ship detection |
| CN108681691A (en)* | 2018-04-09 | 2018-10-19 | 上海大学 | A kind of marine ships and light boats rapid detection method based on unmanned water surface ship |
| CN110120073B (en)* | 2019-03-28 | 2020-12-22 | 上海大学 | A method for guiding the recovery of unmanned boats based on visual signals of beacons |
| CN110443201B (en)* | 2019-08-06 | 2023-01-10 | 哈尔滨工业大学 | Target identification method based on multi-source image joint shape analysis and multi-attribute fusion |
| CN110910421B (en)* | 2019-11-11 | 2022-11-11 | 西北工业大学 | Weak and small moving object detection method based on block characterization and variable neighborhood clustering |
| CN111160354B (en)* | 2019-12-30 | 2022-06-17 | 哈尔滨工程大学 | A method for image segmentation of ships in sea and sky background based on joint image information |
| CN111582089B (en)* | 2020-04-27 | 2021-07-09 | 中国人民解放军军事科学院国防科技创新研究院 | Maritime target information fusion method based on satellite infrared and visible light images |
| CN111931688A (en)* | 2020-08-27 | 2020-11-13 | 珠海大横琴科技发展有限公司 | Ship recognition method and device, computer equipment and storage medium |
| CN112070083B (en)* | 2020-09-04 | 2024-12-24 | 北京灵汐科技有限公司 | Image content detection method, device, electronic device and storage medium |
| CN112101250B (en)* | 2020-09-18 | 2022-07-15 | 电子科技大学 | Method for detecting offshore ship target based on context semantic perception |
| CN113674308B (en)* | 2021-05-06 | 2024-02-13 | 西安电子科技大学 | SAR image ship target rapid detection method based on image enhancement and multiple detection |
| CN115331113A (en)* | 2022-10-12 | 2022-11-11 | 浙江华是科技股份有限公司 | Ship target detection model training method and system and computer storage medium |
| CN115661434B (en)* | 2022-10-17 | 2023-05-02 | 中国人民解放军61540部队 | Night marine ship automatic identification method, system, electronic equipment and medium |
| CN117761695B (en)* | 2024-02-22 | 2024-04-30 | 中国科学院空天信息创新研究院 | Multi-angle SAR 3D imaging method based on adaptive partitioned SIFT |
| CN118097473B (en)* | 2024-04-22 | 2024-08-13 | 中国科学院西安光学精密机械研究所 | Sea surface target sensing method for aircraft |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102024156A (en)* | 2010-11-16 | 2011-04-20 | 中国人民解放军国防科学技术大学 | Method for positioning lip region in color face image |
- 2012
- 2012-03-21CNCN 201210077407patent/CN102663348B/ennot_activeExpired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102024156A (en)* | 2010-11-16 | 2011-04-20 | 中国人民解放军国防科学技术大学 | Method for positioning lip region in color face image |
Non-Patent Citations (8)
| Title |
|---|
| Learning Mid-Level Features For Recognition;Y-Lan Boureau等;《Proc 23rd International conference on computer vision and pattern recognition》;20100618;第2559-2566页* |
| Y-Lan Boureau等.Learning Mid-Level Features For Recognition.《Proc 23rd International conference on computer vision and pattern recognition》.2010, |
| 利用贝叶斯网络融合空间上下文的高分辨遥感图像分类;程环环等;《计算机工程与科学》;20110131;第33卷(第1期);第70-76页* |
| 可见光遥感图像中舰船目标检测方法;赵英海等;《光电工程》;20080831;第35卷(第8期);第102-106页* |
| 周晖等.引入PLSA模型的光学遥感图像舰船检测.《遥感学报》.2010,第14卷(第4期), |
| 引入PLSA模型的光学遥感图像舰船检测;周晖等;《遥感学报》;20100430;第14卷(第4期);第672-680页* |
| 程环环等.利用贝叶斯网络融合空间上下文的高分辨遥感图像分类.《计算机工程与科学》.2011,第33卷(第1期), |
| 赵英海等.可见光遥感图像中舰船目标检测方法.《光电工程》.2008,第35卷(第8期), |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102663348A (en) | 2012-09-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102663348B (en) | Marine ship detection method in optical remote sensing image | |
| CN109427055B (en) | Remote sensing image sea surface ship detection method based on visual attention mechanism and information entropy | |
| CN108304873B (en) | Target detection method and system based on high-resolution optical satellite remote sensing image | |
| CN106845372B (en) | A ship target detection and recognition method and system for spaceborne remote sensing optical images | |
| Dowden et al. | Sea ice classification via deep neural network semantic segmentation | |
| CN109902618A (en) | A kind of sea ship recognition methods and device | |
| Alsmadi et al. | Fish recognition based on robust features extraction from color texture measurements using back-propagation classifier | |
| CN105894037B (en) | A fully supervised classification method of remote sensing images based on SIFT training sample extraction | |
| CN107704865A (en) | Fleet Targets Detection based on the extraction of structure forest edge candidate region | |
| CN111898627A (en) | A PCA-based SVM Cloud Particle Optimal Classification and Recognition Method | |
| CN115471755A (en) | Image target rapid detection method based on segmentation | |
| Feng et al. | Rapid ship detection method on movable platform based on discriminative multi-size gradient features and multi-branch support vector machine | |
| Marie et al. | Real-time maritime situation awareness based on deep learning with dynamic anchors | |
| Huang et al. | Ship detection using texture statistics from optical satellite images | |
| CN109325958B (en) | A nearshore vessel detection method based on contour refinement and improved generalized Hough transform | |
| US10115195B2 (en) | Method and apparatus for processing block to be processed of urine sediment image | |
| CN111476074A (en) | Human body foreign matter detection method based on millimeter wave image | |
| CN116563591A (en) | Optical smoke detection method based on feature extraction under sea-sky background | |
| Chowdhury et al. | Scene text detection using sparse stroke information and MLP | |
| Kukreja et al. | Image Processing | |
| Weinmann et al. | Detection, segmentation and localization of individual trees from MMS point cloud data | |
| CN116520320A (en) | A Sea Target Recognition Method | |
| CN111539242A (en) | Millimeter wave imaging image processing method | |
| van den Broek et al. | Discrimination of classes of ships for aided recognition in a coastal environment | |
| Sharma | Extraction of text regions in natural images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | Granted publication date:20131016 | |
| CF01 | Termination of patent right due to non-payment of annual fee |