Target identification method for specific luminescence characteristics in robot shooting confrontation environment

Abstract

The invention belongs to the technical field of robots, and particularly relates to a target identification method for specific light-emitting features in a robot shooting confrontation environment. The method comprises the following steps of: acquiring data; a step of data conversion; a step of data calculation; a step of data judgment; a data screening step; and a data output step. Compared with the prior art, the method has the following beneficial effects: compared with other existing methods, the method has a higher recognition rate in the scene. Compared with color-based recognition, the recognition accuracy is improved by 55%, and the robustness is improved by 30%; compared with identification based on depth, the hardware cost is reduced by 50%, the frame number is increased by 533%, and the processing comprehensive delay is reduced by 81%. The method provided by the invention provides excellent data for subsequent strike.

Description

Technical Field

[0001] The invention belongs to the technical field of robots, and in particular relates to a target recognition method with specific luminous features in a robot shooting confrontation environment. Background Art

[0002] Artificial intelligence is a discipline that studies how robots can simulate human intelligent behavior. It gives robots the ability to learn, reason, perceive and understand the environment. Artificial intelligence technology has a wide range of applications in many fields, such as machine learning, natural language processing, computer vision, etc.

[0003] Robot shooting confrontation is a robot competition project that simulates war scenes. It requires robots to be able to autonomously identify and attack enemy robots in complex environments, demonstrating the robot's intelligence and combat capabilities. One of the core technologies of robot shooting confrontation is target recognition, that is, the robot can accurately identify the position and state of the enemy robot so as to carry out effective attacks. The difficulty of target recognition lies in the fact that enemy robots may have different appearances, colors, movement methods and other characteristics, and there may be a variety of interference factors in the environment, such as lighting, occlusion, noise, etc., which will affect the robot's visual perception and judgment.

[0004] In order to realize target recognition, the existing methods mainly include the following: one method is color-based target recognition, that is, the color characteristics of the target robot are used for recognition. The advantage of this method is that it is simple and easy to implement, but the disadvantage is that it has high requirements for the selection of color and the lighting conditions of the environment. If the color of the target robot is similar to the background color or the lighting changes drastically, it will cause recognition errors or failures. Compared with the present invention, this method cannot identify the target robot's area to be hit, nor can it obtain the motion state information of the target robot. Another method is shape-based target recognition, that is, it uses the shape characteristics of the target robot for recognition. The advantage of this method is that it has little influence on color and lighting, but the disadvantage is that it has high requirements for shape extraction and matching. If the shape of the target robot is complex or there is occlusion, it will cause recognition difficulties or inaccuracies. Compared with the present invention, this method cannot identify the target robot's area to be hit, nor can it obtain the motion state information of the target robot. Another method is depth-based target recognition, that is, it uses the depth information of the target robot for recognition. The advantage of this method is that it can obtain the distance and position information of the target robot, but the disadvantage is that it needs to use equipment such as depth cameras or binocular cameras, which increases the cost and complexity of the robot, and the acquisition and processing of depth information is also time-consuming and resource-consuming. Compared with the present invention, this method greatly increases hardware costs and computing resource consumption.

[0005] In summary, the existing target recognition methods all have certain limitations and shortcomings, and cannot meet the requirements of efficiency and accuracy of robot shooting confrontation. Therefore, the technical purpose of the present invention is to provide a method for automatically identifying characteristic targets in a robot shooting confrontation environment, which utilizes the specific luminous characteristics of the target robot for identification, and can identify the target robot's area to be hit, while obtaining the distance, position and motion state information of the target robot, thereby improving the strike efficiency and accuracy of one's own robot. The beneficial effect of the present invention is that, compared with the existing target recognition method, the method of the present invention has higher recognition rate and robustness, lower cost and complexity, faster processing speed and response time. It is possible to achieve target recognition with extremely high frame number and response speed, providing excellent data for subsequent strikes. Summary of the invention

[0006] The purpose of the present invention is to provide a target recognition method with specific luminous features in a robot shooting confrontation environment in view of the defects of the prior art.

[0007] The specific technical solution adopted by the present invention is as follows: a method for identifying a target with specific luminous features in a robot shooting confrontation environment, which comprises the following steps:

[0008] Step 1: To ensure the quality and consistency of the image, the resolution, exposure time, color temperature, and focal length parameters of the monocular color camera are fixed to facilitate subsequent image processing and analysis.

[0009] Step 2: Capture the current image through a monocular color camera and transmit the image information to the computing module through a data line for image processing and analysis.

[0010] Step 3: Convert the image from BGR color space to HSV color space to obtain the image information of HSV color space, where each element represents the hue, saturation and brightness value of a pixel.

[0011] Step 4: Extract the V channel from the image in the HSV color space to obtain the image information of the brightness channel, where each element represents the brightness value of a pixel.

[0012] Step 5: Binarize the image of the brightness channel to obtain the binarized image information, where each element represents the binarization result of a pixel.

[0013] Step 6: Dilate the binarized image to enhance the connectivity and visibility of the target robot’s luminous features.

[0014] Step 7: Perform contour detection on the expanded image to obtain the position and shape information of the target robot's luminous features.

[0015] Step 8: Solve the minimum bounding rectangle for each contour in the image and obtain the information of the minimum bounding rectangle, including the center coordinates, length, width, area, perimeter and angle of the rectangle.

[0016] Step 9: According to the characteristics of the target robot's luminous features, set the screening conditions, including the aspect ratio threshold, area threshold, perimeter threshold and angle threshold of the minimum bounding rectangle, and screen all the minimum bounding rectangles in the image, and only keep the minimum bounding rectangle that meets the conditions.

[0017] Step 10: Sort the filtered minimum enclosing rectangles in descending order according to their area, and store the sorted results in the form of a list. Each element in the list represents the information of a minimum enclosing rectangle.

[0018] Step 11: Traverse each element in the list, perform affine transformation on it respectively, and then input it into an SVM classifier to determine whether the minimum bounding rectangle is the target robot's attack area. If the classification result of the SVM classifier is no, continue to traverse the next minimum bounding rectangle; if it is yes, stop traversing and record the information of the minimum bounding rectangle.

[0019] Step 12: Calculate the distance between the target robot and the monocular color camera, that is, the depth information of the target robot, based on the area of ​​the recorded minimum circumscribed rectangle and the internal parameters of the monocular color camera.

[0020] Step 13: Based on the area and distance of the recorded minimum circumscribed rectangle, calculate the coordinates of the target robot in the three-dimensional coordinate system with the monocular color camera as the origin, that is, the spatial position information of the target robot.

[0021] Step 14: By acquiring two frames of images from a monocular color camera, the motion characteristics of the target robot are calculated, including speed, direction, acceleration, that is, the motion state information of the target robot.

[0022] As described above, a method for target recognition of specific luminous features in a robot shooting confrontation environment, wherein the step one includes setting the resolution of the monocular color camera to 320*240-1920*1080, the exposure time to 0.2s-2.0s, preferably 0.5s, the color temperature to 5500k-9000k, and the focal length to 2-8m.

[0023] As described above, a method for target recognition of specific luminous features in a robot shooting confrontation environment, wherein the formula for converting the image from the BGR color space to the HSV color space in step three is:

[0024] V=max{B,G,R}

[0025]

[0026] Where {B,G,R} is the value of BGR color space.

[0027] The method for identifying a target with specific luminous features in a robot shooting confrontation environment as described above, wherein the step five comprises:

[0028] The image of the brightness channel is binarized, that is, the brightness value of each pixel in the image is compared with a fixed threshold value, which is set to 200-254, preferably 220. If it is greater than the threshold value, it is set to 1, otherwise it is set to 0 to obtain the binarized image information, in which each element represents the binarization result of a pixel.

[0029] The method for identifying a target with specific luminous features in a robot shooting confrontation environment as described above, wherein the step six comprises:

[0030] The binary image is expanded, that is, the pixel values ​​around the pixel with a value of 1 are set to 1, and the expanded image information is obtained, in which each element represents the expansion result of a pixel.

[0031] The target recognition method of a specific luminous feature in a robot shooting confrontation environment as described above, wherein the contour detection in step seven includes:

[0032] Use the findcontour algorithm to surround each connected area with a brightness of 1 in the image with one or more closed curves to obtain the image information after contour detection, where each element represents the contour detection result of a pixel.

[0033] The method for identifying a target with specific luminous features in a robot shooting confrontation environment as described above, wherein the step nine comprises:

[0034] According to the characteristics of the target robot's luminous features, the screening conditions are set, including the aspect ratio threshold, area threshold, perimeter threshold and angle threshold of the minimum enclosing rectangle, where the aspect ratio threshold is set to 2-4.5, the area threshold is set to 20-5000, the perimeter threshold is set to 15-350, and the angle threshold is set to 0-15. All minimum enclosing rectangles in the image are screened, and only the minimum enclosing rectangles that meet the conditions are retained.

[0035] The method for identifying a target with specific luminous features in a robot shooting confrontation environment as described above, wherein the step ten comprises:

[0036] The filtered minimum enclosing rectangles are sorted in descending order according to the size of the area, and the Timsort sorting algorithm is used to establish a list List whose storage data type is an integer array to save the sorted results. That is, each element in the list represents the information of a minimum enclosing rectangle, and the minimum enclosing rectangles are stored in this list in descending order according to the size of the area.

[0037] The method for identifying a target with specific luminous features in a robot shooting confrontation environment as described above, wherein the step twelve comprises:

[0038] According to the recorded area of ​​the minimum circumscribed rectangle and the internal parameters of the monocular color camera, the distance between the target robot and the monocular color camera, that is, the depth information of the target robot, is calculated. The specific calculation formula is as follows: D represents the distance between the object and the camera, which is also a parameter to be determined. P is the width of the target, that is, the width coordinate of the minimum circumscribed rectangle. P represents the width of the object in reality. Since the target feature is fixed, this parameter is known. f represents the focal length of the camera, which varies depending on the camera hardware itself and is a known parameter.

[0039]

[0040] Compared with the prior art, the present invention has the following beneficial effects: Compared with other existing methods, the present method has a higher recognition rate in this scenario. Compared with color-based recognition, the recognition accuracy is increased by 55%, and the robustness is increased by 30%; compared with depth-based recognition, the hardware cost is reduced by 50%, the number of frames is increased by 533%, and the overall processing delay is reduced by 81%. The method provided by the present invention provides excellent data for subsequent strikes. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is the first specific luminous feature of the target robot.

[0042] Figure 2 It is the second specific luminous feature of the target robot.

[0043] Figure 3 A flow chart of a target recognition method provided by an embodiment of the present invention.

[0044] In the picture: 1. Fixing bracket, 2. Fixing screws, 3. Light bar that can emit red or blue characteristic light, 4. Area to be struck DETAILED DESCRIPTION

[0045] The present invention will be further described and illustrated below in conjunction with the accompanying drawings and specific implementation methods.

[0046] The present invention provides a method for automatically identifying characteristic targets in a robot shooting confrontation environment, and the method is applicable to the following system shooting confrontation robot scenario. The system consists of the following parts: a self-robot, a target robot, a pan-tilt head, and a monocular color camera. The self-robot is a robot that can identify and attack the target robot in a complex environment. The target robot is the robot to be attacked, and it has a specific luminous feature that can be used as a basis for identification. The pan-tilt head is a rotatable platform installed on the self-robot, which is used to adjust the direction of the monocular color camera. The monocular color camera is a camera that can capture color images. It serves as a visual sensor of the self-robot and is used to obtain information about the target robot.

[0047] This application includes the following steps:

[0048] Step 1: To ensure the quality and consistency of the image, the resolution, exposure time, color temperature, and focal length parameters of the monocular color camera are fixed to facilitate subsequent image processing and analysis.

[0049] Step 2: Capture the current image through a monocular color camera and transmit the image information to the computing module through a data line for image processing and analysis.

[0050] Step 3: Convert the image from BGR color space to HSV color space to obtain image information in HSV color space, where each element represents the hue, saturation, and brightness value of a pixel.

[0051] Step 4: Extract the V channel from the image in the HSV color space to obtain the image information of the brightness channel, where each element represents the brightness value of a pixel.

[0052] Step 5: Binarize the image of the brightness channel to obtain binarized image information, where each element represents the binarization result of a pixel.

[0053] Step 6: Dilate the binarized image to enhance the connectivity and visibility of the target robot’s luminous features.

[0054] Step 7: Perform contour detection on the expanded image to obtain the position information and shape information of the target robot's luminous features.

[0055] Step 8: Solve the minimum enclosing rectangle for each contour in the image and obtain the information of the minimum enclosing rectangle, including the center coordinates, length, width, area, perimeter and angle of the rectangle.

[0056] Step 9: According to the characteristics of the target robot's luminous features, set the screening conditions, including the aspect ratio threshold, area threshold, perimeter threshold and angle threshold of the minimum bounding rectangle, filter all the minimum bounding rectangles in the image, and only retain the minimum bounding rectangles that meet the conditions.

[0057] Step 10: Sort the filtered minimum enclosing rectangles in descending order according to their areas, and store the sorted results in the form of a list, where each element in the list represents the information of a minimum enclosing rectangle.

[0058] Step 11: Traverse each element in the list, perform affine transformation on it, and then input it into an SVM classifier to determine whether the minimum bounding rectangle is the target robot's attack area. If the classification result of the SVM classifier is no, continue to traverse the next minimum bounding rectangle; if it is yes, stop traversing and record the information of the minimum bounding rectangle.

[0059] Step 12: According to the area of ​​the recorded minimum circumscribed rectangle and the internal parameters of the monocular color camera, calculate the distance between the target robot and the monocular color camera, that is, the depth information of the target robot.

[0060] Step 13: Based on the area and distance of the recorded minimum circumscribed rectangle, calculate the coordinates of the target robot in the three-dimensional coordinate system with the monocular color camera as the origin, that is, the spatial position information of the target robot.

[0061] Step 14: By acquiring two frames of images from a monocular color camera, the motion characteristics of the target robot are calculated, including speed, direction, acceleration, that is, the motion state information of the target robot.

[0062] The above steps achieve complete identification of the target robot.

[0063] Through this method, the friendly robot can accurately identify the position and status of the target robot, thereby improving the efficiency and accuracy of the attack.

[0064] The present invention uses a monocular color camera as a visual sensor, simplifies the structure and cost of the system, and improves the speed and response time of image acquisition and processing; uses a brightness channel to perform binarization processing, separates the luminous characteristics of the target robot from the background, reduces the influence on color and illumination, and enhances the robustness of recognition; uses a minimum bounding rectangle to perform screening and sorting, reduces irrelevant contours, and improves the accuracy of recognition; uses a SVM classifier to perform matching, judges whether the minimum bounding rectangle is the target robot's area to be struck, and improves the reliability of recognition; uses the area of ​​the minimum bounding rectangle and the internal parameters of the camera to calculate the distance and position information of the target robot, and improves the accuracy of positioning; uses two frames of images to calculate the motion characteristics of the target robot, and improves the accuracy of prediction.

[0065] In order to further illustrate the present application, a specific example is given below.

[0066] The shooting confrontation robot system applicable to the present application includes: a friendly robot, a target robot, a pan / tilt, and a monocular color camera. The friendly robot is a robot that can identify and attack the target robot in a complex environment. The target robot is the robot to be attacked, and it has a specific luminous feature, such as the attached Figure 1 , Attachment Figure 2 As shown, it can be used as the basis for identification. The pan-tilt is a rotatable platform installed on the robot to adjust the direction of the monocular color camera. The monocular color camera is a camera that can capture color images. It is used as a visual sensor of the robot to obtain information about the target robot.

[0067] Attached Figure 3 The flow chart of the target recognition method provided by the embodiment of the present invention is shown. The specific flow of the target recognition method is as follows:

[0068] Step 1: To ensure the quality and consistency of the image, the resolution of the monocular color camera can be set to 320*240-1920*1080, preferably 640*480, the exposure time can be set to 0.2s-2.0s, preferably 0.5s, the color temperature can be set to 5500k-9000k, preferably 7000k, and the focal length can be set to 2-8m, preferably 5m, to facilitate subsequent image processing and analysis.

[0069] Step 2: Capture the current image through a monocular color camera and save the image as image matrix data. At this time, the image data is a 3*640*480 matrix, where 3 refers to three color channels, namely B, G, and R channels, 640 is the image width, and 480 is the image height in pixels. The image data information is transmitted to the computing module through the data line for image processing and analysis.

[0070] Step 3: Convert the image from BGR color space to HSV color space using the following formula, where B, G, R are the values ​​of the B, G, R channels of a pixel in the original image, and H, S, V are the values ​​of the H, S, V channels of the pixel in the converted image. H, S, V represent the hue, saturation, and brightness of each pixel.

[0071] V=max{B,G,R}

[0072]

[0073] Step 4: Since the image in the HSV color space is an image with three color channels, the V channel, that is, the brightness channel, can be directly extracted. At this time, the image data is a 1*640*480 matrix. Changing from the original 3 to 1 means extracting the V channel from HSV. 640 and 480 are the same as above, and the image information of the brightness channel is obtained, where each element represents the brightness value of a pixel.

[0074] Step 5: Binarize the image of the brightness channel, that is, compare the brightness value of each pixel in the image with a fixed threshold value, which can be set to 200-254, preferably 220. If it is greater than the threshold, it is set to 1, otherwise it is set to 0, and the binarized image information is obtained, in which each element represents the binarization result of a pixel. In this way, the area with higher brightness in the image, that is, the luminous features of the target robot, can be separated from the background.

[0075] Step 6: Dilate the binary image, that is, set the pixel values ​​around the pixel with a value of 1 to 1, and obtain the dilated image information, where each element represents the dilation result of a pixel. In this way, the noise in the image, that is, the area with high brightness but not the luminous feature of the target robot, can be eliminated, and the connectivity and visibility of the luminous feature of the target robot can also be enhanced.

[0076] Step 7: Perform contour detection on the expanded image using the findcontour algorithm, that is, surround each connected area with a brightness of 1 in the image with one or more closed curves to obtain the image information after contour detection, where each element represents the contour detection result of a pixel. In this way, the position information and shape information of all areas with a brightness of 1 in the image can be obtained, that is, the position information and shape information of the target robot's luminous features.

[0077] Step 8: Solve the minimum bounding rectangle for each contour in the image. Since the contours in step 7 are screened based on pixel values, there cannot be intersecting or tangent contours. Therefore, each contour can be surrounded by a minimum rotated rectangle to obtain the minimum bounding rectangle information, including the center coordinates, length, width, area, perimeter and angle of the rectangle, which is stored in the form of an integer array.

[0078] Step 9: According to the characteristics of the target robot's luminous features, set the screening conditions, including the aspect ratio threshold, area threshold, perimeter threshold and angle threshold of the minimum bounding rectangle, where the aspect ratio threshold is set to 2-4.5, the area threshold is set to 20-5000, the perimeter threshold is set to 15-350, and the angle threshold is set to 0-15. All minimum bounding rectangles in the image are screened, and only the minimum bounding rectangles that meet the conditions are retained.

[0079] Step 10: Sort the filtered minimum enclosing rectangles in descending order according to their area, use the Timsort sorting algorithm, create a list List whose storage data type is an integer array, and save the sorted results, that is, each element in the list represents the information of a minimum enclosing rectangle, and the minimum enclosing rectangles are stored in this list in descending order according to their area size.

[0080] Step 11: Traverse the list and perform an affine transformation on each minimum bounding rectangle, that is, transform it into a horizontal rectangle, and then input it into an SVM classifier. SVM is a support vector machine, which can classify the input data to distinguish whether the minimum bounding rectangle is the target robot's area to be hit. If the classification result of the SVM classifier is no, that is, the minimum bounding rectangle is not the target robot's area to be hit, then continue to traverse the next minimum bounding rectangle; if the classification result of the SVM classifier is yes, that is, the minimum bounding rectangle is the target robot's area to be hit, then stop traversing and record the information of the minimum bounding rectangle.

[0081] The SVM classifier here is a classifier specially trained for this classification scenario and has extremely high accuracy for this classification task.

[0082] Step 12: According to the recorded area of ​​the minimum circumscribed rectangle and the internal parameters of the monocular color camera, calculate the distance between the target robot and the monocular color camera, that is, the depth information of the target robot. The specific calculation formula is as follows, where D represents the distance between the object and the camera, which is also a parameter to be determined. P is the width of the target, that is, the width coordinate of the minimum circumscribed rectangle. P represents the width of the object in reality. Since the target features are fixed, this parameter is known. f represents the focal length of the camera, which varies depending on the camera hardware itself and is a known parameter.

[0083]

[0084] Step 13: Based on the recorded area and distance of the minimum circumscribed rectangle and the relative position of the camera relative to the transmitting structure, calculate the coordinates of the target robot in the three-dimensional coordinate system with the monocular color camera as the origin, that is, the spatial position information of the target robot.

[0085] Step 14: By acquiring two frames of images from the monocular color camera and the previously acquired one frame of image, the Kalman filter algorithm is used to calculate the motion characteristics of the target robot, including speed, direction, acceleration, that is, the motion state information of the target robot.

[0086] The present invention provides a method for identifying targets with specific luminous features in a robot shooting confrontation environment. Compared with other existing methods, this method has a higher recognition rate in this scenario. Compared with color-based recognition, the recognition accuracy is improved by 55%, and the robustness is improved by 30%; compared with depth-based recognition, the hardware cost is reduced by 50%, the number of frames is increased by 533%, and the overall processing delay is reduced by 81%. The method provided by the present invention provides excellent data for subsequent strikes.

Claims

1. A method for identifying targets with specific luminous features in a robot shooting confrontation environment, characterized in that: The steps include: Step 1: To ensure the quality and consistency of the image, the resolution, exposure time, color temperature, and focal length parameters of the monocular color camera are fixed to facilitate subsequent image processing and analysis. Step 2: Capture the current image through a monocular color camera and transmit the image information to the computing module through a data line for image processing and analysis. Step 3: Convert the image from BGR color space to HSV color space to obtain the image information of HSV color space, where each element represents the hue, saturation and brightness value of a pixel. Step 4: Extract the V channel from the image in the HSV color space to obtain the image information of the brightness channel, where each element represents the brightness value of a pixel. Step 5: Binarize the image of the brightness channel to obtain the binarized image information, where each element represents the binarization result of a pixel. Step 6: Dilate the binarized image to enhance the connectivity and visibility of the target robot’s luminous features. Step 7: Perform contour detection on the expanded image to obtain the position and shape information of the target robot's luminous features. Step 8: Solve the minimum bounding rectangle for each contour in the image and obtain the information of the minimum bounding rectangle, including the center coordinates, length, width, area, perimeter and angle of the rectangle. Step 9: According to the characteristics of the target robot's luminous features, set the screening conditions, including the aspect ratio threshold, area threshold, perimeter threshold and angle threshold of the minimum bounding rectangle, and screen all the minimum bounding rectangles in the image, and only keep the minimum bounding rectangle that meets the conditions. Step 10: Sort the filtered minimum enclosing rectangles in descending order according to their area, and store the sorted results in the form of a list. Each element in the list represents the information of a minimum enclosing rectangle. Step 11: Traverse each element in the list, perform affine transformation on it respectively, and then input it into an SVM classifier to determine whether the minimum bounding rectangle is the target robot's attack area. If the classification result of the SVM classifier is no, continue to traverse the next minimum bounding rectangle; if it is yes, stop traversing and record the information of the minimum bounding rectangle. Step 12: Calculate the distance between the target robot and the monocular color camera, that is, the depth information of the target robot, based on the area of ​​the recorded minimum circumscribed rectangle and the internal parameters of the monocular color camera. Step 13: Based on the area and distance of the recorded minimum circumscribed rectangle, calculate the coordinates of the target robot in the three-dimensional coordinate system with the monocular color camera as the origin, that is, the spatial position information of the target robot. Step 14: By acquiring two frames of images from a monocular color camera, the motion characteristics of the target robot are calculated, including speed, direction, acceleration, that is, the motion state information of the target robot.

2. A method for identifying a target with specific luminous features in a robot shooting confrontation environment as claimed in claim 1, characterized in that: The step 1 includes setting the resolution of the monocular color camera to 320*240-1920*1080, the exposure time to 0.2s-2.0s, preferably 0.5s, the color temperature to 5500k-9000k, and the focal length to 2-8m.

3. A method for identifying a target with specific luminous features in a robot shooting confrontation environment as claimed in claim 2, characterized in that: The formula for converting the image from the BGR color space to the HSV color space in step 3 is: V=max{B,G,R} Where {B,G,R} is the value of BGR color space.

4. A method for identifying a target with specific luminous features in a robot shooting confrontation environment as claimed in claim 3, characterized in that: The step five comprises: The image of the brightness channel is binarized, that is, the brightness value of each pixel in the image is compared with a fixed threshold value, which is set to 200-254, preferably 220. If it is greater than the threshold value, it is set to 1, otherwise it is set to 0 to obtain the binarized image information, in which each element represents the binarization result of a pixel.

5. A method for identifying a target with specific luminous features in a robot shooting confrontation environment as claimed in claim 4, characterized in that: The step six comprises: The binary image is expanded, that is, the pixel values ​​around the pixel with a value of 1 are set to 1, and the expanded image information is obtained, in which each element represents the expansion result of a pixel.

6. A method for identifying a target with specific luminous features in a robot shooting confrontation environment as claimed in claim 5, characterized in that: The contour detection in step seven includes: Use the findcontour algorithm to surround each connected area with a brightness of 1 in the image with one or more closed curves to obtain the image information after contour detection, where each element represents the contour detection result of a pixel.

7. A method for identifying a target with specific luminous features in a robot shooting confrontation environment as claimed in claim 6, characterized in that: The step nine comprises: According to the characteristics of the target robot's luminous features, the screening conditions are set, including the aspect ratio threshold, area threshold, perimeter threshold and angle threshold of the minimum enclosing rectangle, where the aspect ratio threshold is set to 2-4.5, the area threshold is set to 20-5000, the perimeter threshold is set to 15-350, and the angle threshold is set to 0-15. All minimum enclosing rectangles in the image are screened, and only the minimum enclosing rectangles that meet the conditions are retained.

8. A method for identifying a target with specific luminous features in a robot shooting confrontation environment as claimed in claim 7, characterized in that: The step ten comprises: The filtered minimum enclosing rectangles are sorted in descending order according to the size of the area, and the Timsort sorting algorithm is used to establish a list List whose storage data type is an integer array to save the sorted results. That is, each element in the list represents the information of a minimum enclosing rectangle, and the minimum enclosing rectangles are stored in this list in descending order according to the size of the area.

9. A method for identifying a target with specific luminous features in a robot shooting confrontation environment as claimed in claim 8, characterized in that: The step twelve comprises: According to the recorded area of ​​the minimum circumscribed rectangle and the internal parameters of the monocular color camera, the distance between the target robot and the monocular color camera, that is, the depth information of the target robot, is calculated. The specific calculation formula is as follows: D represents the distance between the object and the camera, which is also a parameter to be determined. P is the width of the target, that is, the width coordinate of the minimum circumscribed rectangle. P represents the width of the object in reality. Since the target feature is fixed, this parameter is known. f represents the focal length of the camera, which varies depending on the camera hardware itself and is a known parameter.

Images