CN102331883B - Identification method of three-dimensional control endpoint and computer readable medium using same - Google Patents

Abstract

A method for identifying a three-dimensional control endpoint and a computer readable medium using the same are provided. The identification method of the three-dimensional control endpoint comprises the following steps: receiving depth information, wherein the depth information is related to an image captured by an image capturing device; generating three-dimensional block information related to the three-dimensional block according to the depth information; generating a reference plane according to the depth information; generating a connection group according to the three-dimensional block information and the reference plane; and selecting the three-dimensional block closest to the image capturing device in the connection group as a control endpoint.

Description

Translated fromChinese

三维控制端点的辨识方法及应用其的计算机可读介质Identification method of three-dimensional control endpoint and computer readable medium using same

技术领域technical field

本发明涉及一种控制端点的辨识方法及应用该方法的装置，且特别涉及一种三维控制端点的辨识方法及使用其的计算机可读介质。The present invention relates to a control endpoint identification method and a device applying the method, and in particular to a three-dimensional control endpoint identification method and a computer-readable medium using the method.

背景技术Background technique

多点触控功能是触控屏幕接口中一项十分便利的功能，多点触控的精神在于可以利用更贴进人类的动作习惯来操作系统，更加拉进人与计算机的距离。第一种已知技术的做法先定义物件特征，如拳头的颜色特征、形状特征或纹理特征。之后再撷取图像。接着将物件特征与图像中的特征区块进行比对以找出控制端点。第二种已知技术的做法进一步利用深度特征来过滤背景以防止复杂背景所造成的误判。第三种已知技术的做法则是找出三维控制区域，并在三维控制区域中利用深度信息与手部特征找出最接近摄影机的控制端点。The multi-touch function is a very convenient function in the touch screen interface. The spirit of multi-touch is that it can use the action habits that are more suitable for human beings to operate the system, and further shorten the distance between people and computers. The method of the first known technology first defines the object features, such as the color feature, shape feature or texture feature of the fist. Capture the image later. Then the object features are compared with the feature blocks in the image to find the control endpoints. The method of the second known technology further utilizes depth features to filter the background to prevent misjudgment caused by complex backgrounds. The third known technique is to find out the 3D control area, and use the depth information and hand features to find the control endpoint closest to the camera in the 3D control area.

发明内容Contents of the invention

本发明涉及一种三维控制端点的辨识方法及使用其的计算机可读介质。The invention relates to a method for identifying three-dimensional control endpoints and a computer-readable medium using the same.

根据本发明的一方面，提出一种三维控制端点的辨识方法。三维控制端点的辨识方法包括：接收深度信息，深度信息相关于图像撷取装置所撷取的图像；根据深度信息产生相关于三维区块的三维区块信息；根据深度信息产生参考平面；根据三维区块信息及参考平面产生连接群组；以及选择连接群组中最接近图像撷取装置的三维区块做为控制端点。According to one aspect of the present invention, a method for identifying a three-dimensional control endpoint is proposed. The identification method of the 3D control endpoint includes: receiving depth information, the depth information is related to the image captured by the image capture device; generating 3D block information related to the 3D block according to the depth information; generating a reference plane according to the depth information; The block information and the reference plane generate a connection group; and the 3D block closest to the image capture device in the connection group is selected as the control endpoint.

根据本发明的另一方面，提出一种计算机可读介质。计算机可读介质具有数个程序指令以执行一三维控制端点的辨识方法，三维控制端点的辨识方法包括：接收深度信息，深度信息相关于图像撷取装置所撷取的图像；根据深度信息产生相关于三维区块的三维区块信息；根据深度信息产生参考平面；根据三维区块信息及参考平面产生连接群组；以及选择连接群组中最接近图像撷取装置的三维区块做为控制端点。According to another aspect of the present invention, a computer-readable medium is proposed. The computer readable medium has several program instructions to execute a method for identifying a three-dimensional control endpoint. The method for identifying a three-dimensional control endpoint includes: receiving depth information, the depth information is related to an image captured by an image capture device; generating a correlation according to the depth information 3D block information in the 3D block; generating a reference plane according to the depth information; generating a connection group according to the 3D block information and the reference plane; and selecting the 3D block closest to the image capture device in the connection group as a control endpoint .

为让本发明的上述内容能更明显易懂，下文特举一优选实施例，并配合附图，作详细说明如下：In order to make the above-mentioned content of the present invention more obvious and understandable, a preferred embodiment is specifically cited below, together with the accompanying drawings, and described in detail as follows:

附图说明Description of drawings

图1绘示为一种三维控制端点的辨识系统。FIG. 1 shows a three-dimensional control endpoint identification system.

图2绘示为一种三维控制端点的辨识方法的流程图。FIG. 2 is a flowchart of a method for identifying three-dimensional control endpoints.

图3绘示为产生三维区块信息的细部流程图。FIG. 3 is a detailed flowchart of generating 3D block information.

图4绘示为凸点的示意图。FIG. 4 is a schematic diagram of bumps.

图5绘示为滤除噪声区块前的示意图。FIG. 5 is a schematic diagram before filtering noise blocks.

图6绘示为滤除噪声区块后的示意图。FIG. 6 is a schematic diagram after filtering noise blocks.

图7绘示为图像中所有凸点的示意图。FIG. 7 is a schematic diagram of all the bumps in the image.

图8绘示为图像中所有三维区块的示意图。FIG. 8 is a schematic diagram of all three-dimensional blocks in an image.

图9绘示为产生参考平面的细部流程图。FIG. 9 is a detailed flowchart of generating a reference plane.

图10绘示为空间分布统计的示意图。FIG. 10 is a schematic diagram of spatial distribution statistics.

图11绘示为产生连接群组的细部流程图。FIG. 11 is a detailed flowchart of generating connection groups.

图12绘示为第一种三维区块连接结果示意图。FIG. 12 is a schematic diagram of the first three-dimensional block connection result.

图13绘示为第二种三维区块连接结果示意图。FIG. 13 is a schematic diagram of the second type of 3D block connection results.

图14绘示为第三种三维区块连接结果示意图。FIG. 14 is a schematic diagram of the third type of 3D block connection results.

图15绘示为于二个三维区块的连结上找出基准参考点的示意图。FIG. 15 is a schematic diagram of finding a reference point on the connection of two 3D blocks.

图16绘示为第一种三维区块类型示意图。FIG. 16 is a schematic diagram of the first three-dimensional block type.

图17绘示为第二种三维区块类型示意图。FIG. 17 is a schematic diagram of the second three-dimensional block type.

图18绘示为第三种三维区块类型示意图。FIG. 18 is a schematic diagram of the third three-dimensional block type.

图19绘示为连接群组的示意图。FIG. 19 is a schematic diagram of connection groups.

【主要元件符号说明】[Description of main component symbols]

10：三维控制端点的辨识系统10: Identification system of 3D control endpoints

21～25、221～225、231～233、241～243：步骤21～25, 221～225, 231～233, 241～243: steps

410、410(1)～410(4)：凸点410, 410(1)～410(4): bump

50：参考平面50: Reference plane

60(1)、60(2)：连接群组60(1), 60(2): Connection groups

110：计算机110: Computer

120：计算机可读介质120: Computer-readable media

420、420(a)、420(b)、420(1)～420(9)：三维区块420, 420(a), 420(b), 420(1)～420(9): three-dimensional blocks

430：基准参考点430: datum reference point

具体实施方式Detailed ways

为了正确地辨识出控制端点，下述实施例提供一种三维控制端点的辨识方法及使用其的计算机可读介质。计算机可读介质具有数个程序指令以执行一三维控制端点的辨识方法，三维控制端点的辨识方法包括：接收深度信息，深度信息相关于图像撷取装置所撷取的图像；根据深度信息产生相关于三维区块的三维区块信息；根据深度信息产生参考平面；根据三维区块信息及参考平面产生连接群组；以及选择连接群组中最接近图像撷取装置的三维区块做为控制端点。In order to correctly identify control endpoints, the following embodiments provide a method for identifying three-dimensional control endpoints and a computer-readable medium using the same. The computer readable medium has several program instructions to execute a method for identifying a three-dimensional control endpoint. The method for identifying a three-dimensional control endpoint includes: receiving depth information, the depth information is related to an image captured by an image capture device; generating a correlation according to the depth information 3D block information in the 3D block; generating a reference plane according to the depth information; generating a connection group according to the 3D block information and the reference plane; and selecting the 3D block closest to the image capture device in the connection group as a control endpoint .

三维控制端点的辨识方法及计算机可读介质Three-dimensional control endpoint identification method and computer readable medium

请同时参照图1及图2，图1绘示为一种三维控制端点的辨识系统，图2绘示为一种三维控制端点的辨识方法的流程图。三维控制端点的辨识系统10包括计算机110及计算机可读介质(Computer-Readable Medium)120。计算机可读介质120具有数个程序指令，供计算机110载入以执行三维控制端点的辨识方法。计算机可读介质120例如为磁盘、光盘、磁带或是硬盘机。三维控制端点的辨识方法包括如下步骤：首先如步骤21所示，接收深度信息。深度信息相关于一图像撷取装置所撷取的图像，而图像撷取装置例如为红外线摄影机或双摄影机。Please refer to FIG. 1 and FIG. 2 at the same time. FIG. 1 shows a three-dimensional control endpoint identification system, and FIG. 2 shows a flow chart of a three-dimensional control endpoint identification method. Thesystem 10 for identifying three-dimensional control endpoints includes a computer 110 and a computer-readable medium (Computer-Readable Medium) 120 . The computer-readable medium 120 has several program instructions for the computer 110 to load to execute the method for identifying three-dimensional control endpoints. The computer readable medium 120 is, for example, a magnetic disk, an optical disk, a magnetic tape, or a hard disk drive. The identification method of the three-dimensional control endpoint includes the following steps: First, as shown instep 21, depth information is received. The depth information is related to an image captured by an image capture device, such as an infrared camera or a dual camera.

接着如步骤22所示，根据深度信息产生相关于三维区块的三维区块信息。跟着如步骤23所示，根据深度信息产生参考平面。然后如步骤24所示，根据三维区块信息及参考平面产生连接群组。最后如步骤25所示，从连接群组中选择最接近图像撷取装置的三维区块做为控制端点。后续将进一步分别说明产生三维区块信息、参考平面及连接群组的步骤。Next, as shown instep 22, 3D block information related to the 3D block is generated according to the depth information. Then, as shown instep 23, a reference plane is generated according to the depth information. Then, as shown instep 24, a connection group is generated according to the 3D block information and the reference plane. Finally, as shown instep 25, the 3D block closest to the image capture device is selected from the connection group as the control endpoint. The steps of generating the 3D block information, the reference plane and the connection group will be further described later.

产生三维区块信息Generate 3D block information

请同时参照图3至图8，图3绘示为产生三维区块信息的细部流程图，图4绘示为凸点的示意图，图5绘示为滤除噪声区块前的示意图，图6绘示为滤除噪声区块后的示意图，图7绘示为图像中所有凸点的示意图，图8绘示为图像中所有三维区块的示意图。前述步骤22进一步包括步骤221至225。Please refer to FIG. 3 to FIG. 8 at the same time. FIG. 3 is a detailed flow chart for generating three-dimensional block information. FIG. 4 is a schematic diagram of bumps. FIG. 5 is a schematic diagram before filtering out noise blocks. FIG. 6 It is a schematic diagram after filtering noise blocks, FIG. 7 is a schematic diagram of all bumps in the image, and FIG. 8 is a schematic diagram of all three-dimensional blocks in the image. The foregoingstep 22 further includes steps 221 to 225 .

首先如步骤221所示，根据深度信息沿水平方向及垂直方向检测出图像撷取装置所撷取的图像中所有凸点410。所谓的凸点410是指在一个图像中存在一个相对于其他周边更为隆起的一定高度的特征像素点(如图4绘示)。First, as shown in step 221 , allbumps 410 in the image captured by the image capture device are detected along the horizontal and vertical directions according to the depth information. The so-calledconvex point 410 refers to a feature pixel point of a certain height that is more elevated than other surrounding areas in an image (as shown in FIG. 4 ).

接着如步骤222所示，根据凸点410与凸点410的周边像素点的深度差异将图7绘示的凸点410扩展为图8绘示的三维区块420。需说明的是，步骤222系检查凸点410周边所有像素点的深度信息。当凸点410周边所有像素点的深度差异在一预设范围内时，即将凸点410周边深度差异在一预设范围内的像素点纳入扩展范围内，使得凸点410扩展为三维区块420。Next, as shown in step 222 , thebump 410 shown in FIG. 7 is expanded into the three-dimensional block 420 shown in FIG. 8 according to the depth difference between thebump 410 and the surrounding pixels of thebump 410 . It should be noted that, step 222 is to check the depth information of all pixels around thebump 410 . When the depth difference of all the pixels around thebump 410 is within a preset range, the pixels whose depth difference around thebump 410 is within a preset range are included in the expansion range, so that thebump 410 is expanded into a three-dimensional block 420 .

跟着如步骤223所示，根据深度变化过滤三维区块420中的噪声区块。需说明的是，步骤223也可与步骤222同步执行。举例来说，在图5绘示的凸点410(1)扩展时，扩展的动作接触到另一个凸点410(2)。此时步骤223即将凸点410(2)与凸点410(1)的深度信息进行比对。当凸点410(2)的深度差异在一预设范围内时，我们便会将凸点410(2)融入凸点410(1)的扩展范围内，并且去除掉凸点410(2)扩展的权力(如图6绘示)，而凸点410(1)扩展为三维区块420(a)。相似地，当凸点410(4)的深度差异在一预设范围内时，我们便会将凸点410(4)融入凸点410(3)的扩展范围内，并且去除掉凸点410(4)扩展的权力，而凸点410(3)扩展为三维区块420(b)。藉此，我们可以简化前述步骤222的运算次数，并在准确度不变的前提下达到加速的效果。Then, as shown in step 223 , the noise blocks in the3D block 420 are filtered according to the depth variation. It should be noted that step 223 may also be executed synchronously with step 222 . For example, when the bump 410(1) shown in FIG. 5 expands, the expanding action contacts another bump 410(2). At this time, step 223 is to compare the depth information of the bump 410(2) with that of the bump 410(1). When the depth difference of the bump 410(2) is within a preset range, we will integrate the bump 410(2) into the extension range of the bump 410(1), and remove the extension of the bump 410(2) power (as shown in FIG. 6 ), and the bump 410(1) expands into a three-dimensional block 420(a). Similarly, when the depth difference of the bump 410(4) is within a preset range, we will integrate the bump 410(4) into the extended range of the bump 410(3), and remove the bump 410( 4) The power of expansion, while the bump 410(3) is expanded into a three-dimensional block 420(b). In this way, we can simplify the number of calculations in the aforementioned step 222, and achieve the effect of speeding up while keeping the accuracy unchanged.

然后如步骤224所示，判断图像中所有凸点410是否检查完毕，如果否则重复前述步骤222及步骤223。不过在步骤223中被去除扩展的权力的凸点将不会再纳入运算，如图5及图6绘示的凸点410(2)及凸点410(4)。步骤224会重复检查直到前述步骤221中所检测的所有凸点410都被检查完毕为止。Then, as shown in step 224, it is judged whether allbumps 410 in the image have been inspected, and if not, the aforementioned steps 222 and 223 are repeated. However, the bumps whose extended power is removed in step 223 will not be included in the calculation, such as bumps 410( 2 ) and 410 ( 4 ) shown in FIG. 5 and FIG. 6 . Step 224 repeats the inspection until allbumps 410 detected in the aforementioned step 221 are inspected.

接着如步骤225所示，产生相关于三维区块420的三维区块信息。进一步来说，当前述步骤224找出所有三维区块420后，步骤225系先于所有三维区块420分别找出代表点，代表点例如为三维区块420的重心。当三维区块420的代表点决定后，再产生相关于代表点的三维区块信息。Next, as shown in step 225 , the 3D block information related to the3D block 420 is generated. Further speaking, after all the 3D blocks 420 are found in the aforementioned step 224 , the step 225 is to find representative points respectively before all the 3D blocks 420 , such as the center of gravity of the 3D blocks 420 . After the representative point of the3D block 420 is determined, the 3D block information related to the representative point is generated.

产生参考平面generate reference plane

请同时参照图9至图10，图9绘示为产生参考平面的细部流程图，图10绘示为空间分布统计的示意图。前述步骤23进一步包括步骤231至233。首先如步骤231所示，根据深度信息计算空间分布统计。进一步来说，步骤231通过深度信息评估图像中每一个像素在三维空间的对应位置，进而重建一个三维场景。藉此通过三维场景可得到如图10绘示的空间分布统计。在图10中，原点表示图像撷取装置所在位置，而x轴表示距离图像撷取装置的远近，y轴则表示像素个数的多寡。由于x轴表示距离图像撷取装置的远近，因此图10绘示的空间分布统计即表示不同深度下的像素个数。Please refer to FIG. 9 to FIG. 10 at the same time. FIG. 9 is a detailed flowchart of generating a reference plane, and FIG. 10 is a schematic diagram of spatial distribution statistics. The foregoingstep 23 further includessteps 231 to 233 . First, as shown instep 231, the spatial distribution statistics are calculated according to the depth information. Further,step 231 evaluates the corresponding position of each pixel in the image in the 3D space through the depth information, and then reconstructs a 3D scene. In this way, the spatial distribution statistics shown in FIG. 10 can be obtained through the three-dimensional scene. In FIG. 10 , the origin represents the location of the image capture device, the x-axis represents the distance from the image capture device, and the y-axis represents the number of pixels. Since the x-axis represents the distance from the image capture device, the spatial distribution statistics shown in FIG. 10 represent the number of pixels at different depths.

接着如步骤232所示，区间平滑化及整体加权调整图10绘示的空间分布统计。进一步来说，为了去除空间分布统计中小细节噪声的影响，步骤232利用模糊化的作法，平滑了原始的空间分布统计，减轻小细节噪声的影响。举例来说，步骤232可选择取一定范围的深度区段(y轴坐标中一定大小的区域)，将深度区段所对应的像素加总后平均取得此深度区段的代表值。接着，再逐步移动此深度区段并且重复进行前述运算，让这些噪声不会影响到其后的运算，接下来通过与摄影机距离的关系，加权评估整体的空间分布统计，进而达到整体加权调整的目标。Next, as shown instep 232 , interval smoothing and overall weighting are performed to adjust the spatial distribution statistics shown in FIG. 10 . Further, in order to remove the influence of small detail noise in the spatial distribution statistics,step 232 smoothes the original spatial distribution statistics by using a fuzzy method to reduce the influence of small detail noise. For example, step 232 may select a certain range of depth segments (an area of a certain size in the y-axis coordinates), sum the pixels corresponding to the depth segments and average them to obtain a representative value of the depth segment. Then, move this depth segment step by step and repeat the above calculations, so that the noise will not affect the subsequent calculations. Next, through the relationship with the camera distance, the overall spatial distribution statistics are weighted, and then the overall weighted adjustment is achieved. Target.

跟着如步骤233所示，过滤噪声平面以产生参考平面。进一步来说，为了在图10绘示的空间分布统计中找出适格的参考平面，步骤233会先找出空间分布统计中的最高峰值y_max，并以最高峰值y_max为基准产生评估标准。举例来说，以最高峰值y_max的30％为判定依据，如果有一个峰值的像素值没有超过这个基准值则不会被选取为适合的参考平面，而被判定成噪声平面后过滤。经由步骤233的过滤运算后，图10绘示的空间分布统计中可以发现y₁合乎要求。步骤233再以最高峰值y_max及峰值y₁为基准，分析出峰值区间内的x轴的空间像素分布图，分割出近似深度的不同使用者的参考平面。Next, as shown instep 233, the noise plane is filtered to generate a reference plane. Further, in order to find a suitable reference plane in the spatial distribution statistics shown in FIG. 10 ,step 233 will first find out the highest peak y_max in the spatial distribution statistics, and generate an evaluation standard based on the highest peak y_max . For example, based on 30% of the highest peak y_max , if the pixel value of a peak does not exceed this reference value, it will not be selected as a suitable reference plane, but will be judged as a noise plane and then filtered. After the filtering operation instep 233 , it can be found that y₁ meets the requirements in the spatial distribution statistics shown in FIG. 10 . Instep 233 , based on the highest peak value y_max and peak value y₁ , analyze the spatial distribution of pixels on the x-axis in the peak range, and segment the reference planes of different users with approximate depths.

产生连接群组generate connection group

请同时图11至图19，图11绘示为产生连接群组的细部流程图，图12绘示为第一种三维区块连接结果示意图，图13绘示为第二种三维区块连接结果示意图，图14绘示为第三种三维区块连接结果示意图，图15绘示为于二个三维区块的连结上找出基准参考点的示意图，图16绘示为第一种三维区块类型示意图，图17绘示为第二种三维区块类型示意图，图18绘示为第三种三维区块类型示意图，图19绘示为连接群组的示意图。当前述三维区块信息及三维区块信息产生完毕后，前述步骤24即可根据三维区块信息及参考平面分析出三维区块间的连结性，并且依照三维区块的连结性去产生连接群组。前述步骤24进一步包括步骤241至243。Please see Figure 11 to Figure 19 at the same time, Figure 11 is a detailed flowchart for generating a connection group, Figure 12 is a schematic diagram of the first type of 3D block connection results, and Figure 13 is a diagram of the second type of 3D block connection results Schematic diagram, Figure 14 is a schematic diagram of the connection result of the third three-dimensional block, Figure 15 is a schematic diagram of finding a reference point on the connection of two three-dimensional blocks, and Figure 16 is a schematic diagram of the first three-dimensional block Type schematic diagram, FIG. 17 is a schematic diagram of the second three-dimensional block type, FIG. 18 is a schematic diagram of the third three-dimensional block type, and FIG. 19 is a schematic diagram of a connection group. After the aforementioned 3D block information and 3D block information are generated, theaforementioned step 24 can analyze the connectivity between the 3D blocks according to the 3D block information and the reference plane, and generate connection groups according to the connectivity of the 3D blocks Group. The foregoingstep 24 further includessteps 241 to 243 .

首先如步骤241所示，根据三维区块之间的距离将距离相近的三维区块相连接。进一步来说，步骤241会依照三维区块之间的距离将相近的三维区块连接在一起，而三维区块之间的距离可通过欧几里德距离来计算区块间的距离。一般来说，三维区块420(1)至420(3)正确的连接方向应是如图12绘示朝向参考平面50进行连接，且三维区块420(4)至420(6)正确的连接方向也应是如图12绘示朝向参考平面50进行连接。然而，如果单纯地仅以欧几里德距离来评估二个三维区块是否相连时，则有可能会发生如图13或图14绘示的连接错误。为了避免错误连接的情事发生，步骤241利用欧几里德距离找出最接近的三维区块后，尚须通过后续步骤242检查三维区块间的连接是否正确。如果三维区块间的连接无法通过步骤242的检查时，便需重回到步骤241。当回到步骤241后，则需找出次接近的三维区块，并再次通过后续步骤242检查。First, as shown instep 241, the 3D blocks with similar distances are connected according to the distance between the 3D blocks. Further,step 241 connects the adjacent 3D blocks together according to the distance between the 3D blocks, and the distance between the 3D blocks can be calculated by the Euclidean distance. Generally speaking, the correct connection direction of the three-dimensional blocks 420(1) to 420(3) should be connected towards thereference plane 50 as shown in FIG. 12, and the correct connection of the three-dimensional blocks 420(4) to 420(6) The direction should also be to connect towards thereference plane 50 as shown in FIG. 12 . However, if only the Euclidean distance is used to evaluate whether two 3D blocks are connected, connection errors as shown in FIG. 13 or FIG. 14 may occur. In order to avoid wrong connections, afterstep 241 uses the Euclidean distance to find the closest 3D block, the followingstep 242 is required to check whether the connection between the 3D blocks is correct. If the connection between the 3D blocks fails to pass the check instep 242 , it is necessary to go back tostep 241 . After returning to step 241 , it is necessary to find the next closest 3D block and check it again throughsubsequent step 242 .

接着如步骤242所示，在相连接的二个三维区块的连结上找出深度最低的一点做为基准参考点430，并根据基准参考点430判断连结是否符合预设连接状态，如果连结不符合预设连接状态，则重复执行步骤241。进一步来说，如图15绘示，步骤241先将三维区块420(7)及420(8)连线上所有的点的深度计算出来，再从这些点中找出深度最底的点，并且定义深度最底的点为基准参考点430。当深度最底的点和基准参考点430的关系如图16绘示时，表示三维区块420(7)及420(8)连结上没有相对低的基准参考点430。这样的连接代表沿着手部往参考平面连接，所以图16绘示的三维区块420(7)及420(8)的连结符合预设连接状态。Then, as shown instep 242, find the point with the lowest depth on the connection of the two connected three-dimensional blocks as thereference point 430, and judge whether the connection meets the default connection status according to thereference point 430, if the connection is not If the preset connection status is met,step 241 is repeated. Further, as shown in FIG. 15,step 241 first calculates the depths of all points on the line connecting the three-dimensional blocks 420(7) and 420(8), and then finds the point with the lowest depth from these points, And the point at the bottom of the depth is defined as thedatum reference point 430 . When the relationship between the lowest depth point and thereference point 430 is shown in FIG. 16 , it means that there is no relativelylow reference point 430 on the connection of the three-dimensional blocks 420 ( 7 ) and 420 ( 8 ). Such a connection represents a connection along the hand to the reference plane, so the connection of the three-dimensional blocks 420(7) and 420(8) shown in FIG. 16 conforms to the default connection state.

此外，当深度最底的点和基准参考点430的关系如图17绘示时，表示基准参考点430虽低于三维区块420(7)及420(8)，但基准参考点430与三维区块420(7)及420(8)的距离在一定的限度内。就连结的整体性看来，图17绘示仍旧是向着参考平面进行连接，所以图17绘示的三维区块420(7)及420(8)的连结符合预设连接状态。In addition, when the relationship between the point at the bottom of the depth and thereference point 430 is shown in FIG. The distance between blocks 420(7) and 420(8) is within certain limits. From the point of view of the integrity of the connection, the connection shown in FIG. 17 is still towards the reference plane, so the connection of the three-dimensional blocks 420( 7 ) and 420( 8 ) shown in FIG. 17 conforms to the default connection status.

另外，当深度最底的点和基准参考点430的关系如图18绘示时，表示基准参考点430远低于三维区块420(7)及420(8)。如果三维区块420(7)及420(8)分别为双手之前端，而基准参考点430在参考平面上。此时就算三维区块420(7)及420(8)靠的再近，三维区块420(7)及420(8)间的连接关也不合乎预设连接状态。因此三维区块420(7)及420(8)不会被连接。需重回到步骤241，并找出次接近的三维区块。In addition, when the relationship between the bottommost point and thereference point 430 is shown in FIG. 18 , it means that thereference point 430 is much lower than the three-dimensional blocks 420 ( 7 ) and 420 ( 8 ). If the three-dimensional blocks 420 ( 7 ) and 420 ( 8 ) are the front ends of the hands respectively, and thereference point 430 is on the reference plane. At this time, no matter how close the three-dimensional blocks 420(7) and 420(8) are, the connection between the three-dimensional blocks 420(7) and 420(8) does not conform to the default connection state. Therefore, the 3D blocks 420(7) and 420(8) will not be connected. It is necessary to go back to step 241 and find the next closest 3D block.

同理，不论二个人靠的再近，只要其手部端点的三维区块连线上存在一个陷落的基准参考点430，我们就不会将其连接起来。如此一来可以克服多点对应在多人意欲进行操作时彼此互相干扰的问题，以确保三维区块间的连接符合预设的连接状态。In the same way, no matter how close two people are, as long as there is a submergedreference point 430 on the three-dimensional block connection line of their hand endpoints, we will not connect them. In this way, the problem of multi-point correspondence interference with each other when multiple people intend to operate can be overcome, so as to ensure that the connection between the three-dimensional blocks conforms to the preset connection status.

跟着如步骤243所示，集合彼此相连且连回参考平面的三维区块以产生连接群组。进一步来说，如图19绘示，如果三维区块420(1)及420(4)可以正确连回参考平面50时，则三维区块420(1)及420(4)为合理的控制端点。相反地，如果三维区块420(9)无法正确连回参考平面50时，则三维区块420(9)无法成为合理的控制端点，而为空间中的噪声。藉此，步骤243集合彼此相连且连回参考平面的三维区块420(1)至420(3)以产生连接群组60(1)，并集合彼此相连且连回参考平面的三维区块420(4)至420(6)以产生连接群组60(2)。之后前述步骤25可在连接群组60(1)及连接群组60(2)中选择最接近图像撷取装置的三维区块420(1)及420(4)做为控制端点。Next, as shown instep 243 , the 3D blocks that are connected to each other and back to the reference plane are assembled to generate a connection group. Further, as shown in FIG. 19, if the 3D blocks 420(1) and 420(4) can be correctly connected back to thereference plane 50, then the 3D blocks 420(1) and 420(4) are reasonable control endpoints . On the contrary, if the 3D block 420 ( 9 ) cannot be connected back to thereference plane 50 correctly, then the 3D block 420 ( 9 ) cannot be a reasonable control endpoint, but is noise in the space. Thereby, step 243 aggregates 3D blocks 420(1) to 420(3) connected to each other and back to the reference plane to generate connection group 60(1), and assembles 3D blocks 420 connected to each other and back to the reference plane (4) to 420(6) to generate connection group 60(2). Afterwards, theaforementioned step 25 may select the three-dimensional blocks 420(1) and 420(4) closest to the image capture device in the connection group 60(1) and the connection group 60(2) as control endpoints.

本发明上述实施例所公开的三维控制端点的辨识方法及计算机可读介质，具有多项优点，以下仅列举部分优点说明如下：The identification method of the three-dimensional control endpoint and the computer-readable medium disclosed in the above-mentioned embodiments of the present invention have many advantages, and only some of the advantages are listed below:

一、在复杂背景中亦能正确地检测出控制端点。1. Control endpoints can be detected correctly even in complex backgrounds.

二、不需限定使用者的操作位置及距离。2. There is no need to limit the user's operating position and distance.

三、能以人体类型趋势找出最正确的控制端点，并将多人操作时彼此造成的影响降到最低。3. The most correct control endpoint can be found based on the trend of human body type, and the mutual influence caused by multiple people during operation can be minimized.

综上所述，虽然本发明已以一优选实施例公开如上，然其并非用以限定本发明。本发明所属领域技术人员，在不脱离本发明的精神和范围内，当可作各种的更动与润饰。因此，本发明的保护范围当视所附权利要求书所界定者为准。In summary, although the present invention has been disclosed as above with a preferred embodiment, it is not intended to limit the present invention. Those skilled in the art to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be defined by the appended claims.

Claims (10)

1. three-dimensional discrimination method of controlling end points comprises:

Receive a depth information, this depth information is relevant to the image that an image capturing device captures;

Produce a plurality of spaces block message that is relevant to a plurality of three-dimensional blocks according to this depth information;

Produce at least one reference planes according to this depth information;

Produce at least one connection group according to these space block messages and this reference planes; And

Select to control end points near the three-dimensional block of this image capturing device as one in this connection group.

2. three-dimensional as claimed in claim 1 is controlled the discrimination method of end points, and this step that wherein produces a plurality of spaces block message comprises:

Detect a plurality of salient points according to this depth information;

According to the depth difference of the neighboring pixel point of these salient points and these salient points, these salient points are expanded to these three-dimensional blocks; And

Generation is relevant to these space block messages of these three-dimensional blocks.

3. three-dimensional as claimed in claim 2 is controlled the discrimination method of end points, and this step that wherein produces a plurality of spaces block message also comprises:

Filter noise block in these three-dimensional blocks according to change in depth.

4. three-dimensional as claimed in claim 2 is controlled the discrimination method of end points, and this step that wherein produces a plurality of spaces block message also comprises:

Judge whether these salient points check complete, if otherwise repeat these salient points are expanded to this steps of these three-dimensional blocks.

5. three-dimensional as claimed in claim 2 is controlled the discrimination method of end points, and this step that wherein produces a plurality of spaces block message also comprises:

Find out respectively a plurality of representative points in these three-dimensional blocks; And

Generation is relevant to these space block messages of these representative points.

6. three-dimensional as claimed in claim 5 is controlled the discrimination method of end points, and wherein these representative points are respectively the center of gravity of these three-dimensional blocks.

7. three-dimensional as claimed in claim 1 is controlled the discrimination method of end points, and this step that wherein produces at least one reference planes comprises:

Calculate a space distribution statistics according to this depth information;

Interval smoothing and overall weight are adjusted this space distribution statistics; And

Filter a noise plane to produce this reference planes.

8. three-dimensional as claimed in claim 7 is controlled the discrimination method of end points, and wherein this space distribution statistics is the corresponding number of pixels of different depth.

9. the three-dimensional discrimination method of controlling end points as claimed in claim 7 wherein filters a noise plane and produces this reference planes with the peak-peak according to this space distribution statistics in this step that produces these reference planes.

10. three-dimensional as claimed in claim 1 is controlled the discrimination method of end points, wherein produces at least one this step that connects group and comprises:

The three-dimensional block that distance between block three-dimensional according to these is close with the distance in these three-dimensional blocks is connected;

In one of wherein two the three-dimensional blocks of these distances in close three-dimensional block that are connected connect find out the degree of depth minimum a bit as a reference point, and judge according to this reference point whether this connection meets a default connection status, if this connection does not meet this default connection status, repeat this Connection Step; And

Set is connected with each other and is linked back the three-dimensional block of these reference planes producing this connection group,

Wherein first will be connected apart from immediate two three-dimensional blocks in this Connection Step system, if connections of immediate two the three-dimensional blocks of these distances do not meet this default connection status, two three-dimensional blocks approaching of chosen distance.

Images