Method and system for moving target recognition, real-time positioning and attitude solution

Abstract

The present invention provides a method and system for motion target recognition, real-time positioning, and attitude calculation, including: Step S1: Real-time obtain an image video containing a dynamic target, and preprocess the obtained image video to obtain a preprocessed image; Step S2: Extract the image contour of the preprocessed image and the position coordinates of the feature points corresponding to the contour based on the contour recognition method based on the inclusion relationship; Step S3: Complete the pose calculation and ID recognition of the dynamic target based on the position coordinates of the feature points corresponding to the contour; The contour recognition method based on the inclusion relationship is to recognize the pattern according to the inclusion relationship between the contours with different features contained in the pattern and obtain the position coordinates of the feature points corresponding to the contour.

Description

Technical Field

[0001] The present invention relates to the technical field of dynamic target pose estimation. Specifically, it relates to methods and systems for moving target recognition, real-time positioning, and attitude solution. More specifically, it relates to a method for designing a feature marker board based on contour inclusion relationships. Background Art

[0002] Placing various markers on an unmanned aerial vehicle or other moving targets and using a camera to identify and detect the markers to achieve pose solution is a very common and effective method.

[0003] The design of the marker generally requires it to be light in weight, have little interference with the unmanned aerial vehicle, and be easily distinguishable from the background environment. For example, in Rudol P, Wzorek M, Conte G, et al. Micro unmanned aerial vehicle visual servoing for cooperative indoor exploration[C]. in: 2008 IEEE Aerospace Conference. 2008: 1-10., a lightweight cube frame was installed around a helicopter, and high-brightness LED marker points of different colors were installed at each vertex of the cube. By adjusting the exposure parameters of the camera, the LED marker points could be well separated in the image. However, such a marking method also has deficiencies. It can be seen from the experiments in the paper that when using this method for observation, there is no obvious light source indoors, especially when the camera cannot be directed towards the direction where the light source emits light. Otherwise, the bright spots formed by the light source will significantly interfere with the recognition of the LED marker points, resulting in the failure of the observation.

[0004] One way to avoid the interference of indoor light sources in observations is to use infrared light sources and add an infrared filter in front of the camera lens to filter out visible light. Generally, the wavelength of the light emitted by indoor lighting LEDs is below 850 nm and can be filtered out by the filter. For example: Yan X, Deng H, Quan Q. Active Infrared Coded Target Design and Pose Estimation for Multiple Objects[C / OL]. in: 2019 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Macao, China: IEEE, 2019: 6885 - 6890[2021 - 02 - 14]. Yan X, Deng H, Quan Q. Active infrared coded target design and pose estimation for multiple objects[C]. in: 2019 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). 2019: 6885 - 6890. also used a similar idea. This paper designed an infrared LED light board fixed on the drone. There are multiple isolated LED lights on the marker board. Multiple LED lights are connected into a long strip-shaped contour to distinguish the background noise. The limitations generated thereby are also very obvious. The contours generated by the background noise information are often random. If the noise contour happens to be long strip-shaped, it will be recognized as part of the marker board, creating difficulties and unstable factors for later recognition.

[0005] After ordinary pattern recognition, it can only determine the position of the target in the image. Using a single camera lacks data in the depth direction and cannot achieve the positioning of the target object in space. Since the pattern of the marker board on the drone can be accurately measured and known, the depth direction information can be determined according to the size of the pattern. This problem is the PnP (Perspective-n-Point) problem. The PnP problem has a stable and unique solution with at least 4 points. Therefore, the reflective marker board of the drone must meet the requirement that 4 points are easy to identify in the image.

[0006] Patent document CN107622499B (application number: 201710734207.5) discloses a recognition and spatial positioning method based on a target two-dimensional contour model, which relates to the technical field of image processing. The object is an object with a two-dimensional planar contour, and the recognition target is the contour feature on a certain plane of interest of the object. The recognition target is a closed contour formed by connecting a certain number of straight line segments and / or arc segments and / or circles. The image can be taken of the target and its background from any angle and height, but it is necessary to ensure that all contours of the target are visible and clear. Then, according to the two-dimensional model contour information of the target, the target is recognized from the image, and the spatial position of the target relative to the camera is calculated.

[0007] After comprehensively absorbing the advantages and avoiding the disadvantages of different markers, the present invention designs a reflective marker and implements an efficient and robust recognition algorithm, and the effectiveness of the algorithm can be proved in various indoor and outdoor lighting environments. Summary of the Invention

[0008] Aiming at the defects in the prior art, the object of the present invention is to provide a method and system for recognizing moving targets, real-time positioning and attitude solution.

[0009] A method for recognizing moving targets, real-time positioning and attitude solution according to the present invention includes:

[0010] Step S1: Real-time obtain an image video containing a dynamic target, and preprocess the obtained image video to obtain a preprocessed image;

[0011] Step S2: Extract the image contour of the preprocessed image and the position coordinates of the feature points corresponding to the contour based on the contour recognition method based on the inclusion relationship;

[0012] Step S3: Complete the pose solution and ID recognition of the dynamic target based on the position coordinates of the feature points corresponding to the contour;

[0013] The contour recognition method based on the inclusion relationship is to recognize the pattern according to the inclusion relationship between the contours with different features contained in the pattern and obtain the position coordinates of the feature points corresponding to the contour.

[0014] Preferably, in step S1, a feature marker board is used in cooperation with the infrared light source and infrared filter of the camera to real-time obtain an image video containing a dynamic target;

[0015] The feature marker board is a reflective marker board with a whole-board design, and at the same time, the outermost contour of the reflective marker board is a whole.

[0016] Preferably, step S1 adopts:

[0017] Step S1.1: Use the Laplacian of Gaussian operator to sharpen the image and obtain the sharpened image.

[0018] Step S1.2: Use the Otsu adaptive binarization algorithm to binarize the sharpened image and obtain the binarized image.

[0019] Preferably, in step S2, the preprocessed image uses the findContours function of the OpenCV open-source library to find the contours of the marker board and obtain the inclusion relationship of the found contours.

[0020] Preferably, in step S3, the real-time pose calculation of the moving target is completed based on the contour centroid coordinates combined with the PnP algorithm.

[0021] Preferably, the current marker board is identified by the preset ID pattern.

[0022] According to a moving target recognition, real-time positioning and pose calculation system provided by the present invention, it includes:

[0023] Module M1: Real-time obtain an image video containing a dynamic target, and preprocess the obtained image video to obtain a preprocessed image.

[0024] Module M2: Extract the image contours of the preprocessed image and the position coordinates of the feature points corresponding to the contours based on the contour recognition method of the inclusion relationship.

[0025] Module M3: Complete the pose calculation and ID recognition of the dynamic target based on the position coordinates of the feature points corresponding to the contours.

[0026] The contour recognition method of the inclusion relationship is to identify the pattern based on the inclusion relationship between the contours with different features contained in the pattern and obtain the position coordinates of the feature points corresponding to the contours.

[0027] Preferably, module M1 uses: a feature marker board, an infrared light source of the camera, and an infrared filter to real-time obtain an image video containing a dynamic target.

[0028] The feature marker board is a reflective marker board with a whole-board design, and at the same time, the outermost contour of the reflective marker board is an entire whole.

[0029] Module M1 uses:

[0030] Module M1.1: Use the Laplacian of Gaussian operator to sharpen the image and obtain the sharpened image.

[0031] Module M1.2: Use the Otsu adaptive binarization algorithm to binarize the sharpened image, obtaining the binarized image.

[0032] Preferably, the module M2 adopts: using the findContours function of the OpenCV open-source library for the preprocessed image to find the contours of the marker board and obtain the inclusion relationship of the found contours.

[0033] Preferably, the module M3 adopts: combining the contour centroid coordinates with the PnP algorithm to complete the real-time pose solution of the moving target;

[0034] Perform ID recognition on the current marker board according to the preset ID pattern.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. Through the design of the reflective marker, the present invention realizes an efficient and robust recognition algorithm, and the effectiveness of the algorithm can be proven in various indoor and outdoor lighting environments;

[0037] 2. Through the design of using a whole board for the reflective marker, and the outermost contour of the reflective marker board is an entire entity, the problem that individual bright spots are easily interfered by the background is avoided, and the pose solution and ID recognition of the target can be efficiently and robustly realized;

[0038] 3. The designed feature marker board and its recognition algorithm of the present invention have small estimation errors and are less affected by environmental factors, have high stability and accuracy, and can be widely applied to other existing target recognition scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objects and advantages of the present invention will become more obvious:

[0040] Figure 1 It is a flowchart of the method for moving target recognition and real-time positioning and pose solution.

[0041] Figure 2 It is a schematic diagram of the marker board; wherein: (a) is the design drawing of the marker board; (b) is the physical drawing of the marker board;

[0042] Figure 3 They are feature boards with different IDs;

[0043] Figure 4 It is the contour signal of the marker board and the hierarchical relationship obtained through the contour inclusion relationship;

[0044] Figure 5Schematic diagram of a drone; among which: (a) is a diagram of an unmanned vehicle and the infrared camera carried thereon, and (b) is a schematic diagram of a drone carrying a marker board;

[0045] Figure 6 Schematic diagram of contour detection; among which: (a) is the original image captured by the infrared camera, (b) is the diagram of directly performing contour detection on the basis of (a); (c) is the effect diagram of sharpening the original image, and (d) is the diagram of performing contour detection on the basis of (c);

[0046] Figure 7 Schematic diagram of the recognition effect of the marker board;

[0047] Figure 8 Schematic diagram of an outdoor environment with complete darkness and light interference;

[0048] Figure 9 Schematic diagram of an outdoor environment with sunlight interference;

[0049] Figure 10 Schematic diagram of the recognition effect of the marker board in the outdoor environment;

[0050] Figure 11 Schematic diagram of the recognition effect of the marker board in the outdoor environment; Specific implementation manner

[0051] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several changes and improvements can still be made. These all belong to the protection scope of the present invention.

[0052] Example 1

[0053] According to a method for motion target recognition, real-time positioning and attitude solution provided by the present invention, as Figure 1 shown, it includes:

[0054] Step S1: Real-time obtain an image video containing a dynamic target, and preprocess the obtained image to obtain a preprocessed image; among which, use a feature marker board to cooperate with the infrared light source and infrared filter of the camera to real-time obtain an image video containing a dynamic target, and preprocess the obtained image video;

[0055] Step S2: Extract the image contour of the preprocessed image and the position coordinates of the feature points corresponding to the contour based on the contour recognition method based on the inclusion relationship;

[0056] Step S3: Complete the pose solution and ID recognition of the dynamic target based on the position coordinates of the feature points corresponding to the contour;

[0057] The contour recognition method for the inclusion relationship identifies a pattern based on the inclusion relationship between the contours of different features contained in the pattern, and obtains the position coordinates of the feature points corresponding to the contours. The recognition effect obtained in this way is more robust and efficient.

[0058] Specifically, step S1 adopts:

[0059] Step S1.1: Use the Laplacian of Gaussian operator to sharpen the image. For different positions of the marking plate, convolution kernels of different sizes are adopted to obtain the sharpened image;

[0060] Step S1.2: Use the Otsu adaptive binarization algorithm to binarize the sharpened image to obtain the binarized image.

[0061] Specifically, step S2 adopts: The preprocessed image uses the findContours function of the OpenCV open source library to find the contours of the marking plate and obtain the inclusion relationship of the found contours. Among them, the inclusion relationship and the obtained contour hierarchy relationship are expressed as Figure 4 shown. According to Figure 4 the hierarchy relationship of the contours in, the contours can be defined as the parent contour a and the child contour b according to the relationship between the contours, that is, contour a contains contour b, and the same-level contours c and d, that is, contour e contains contour c and contour e contains contour d. For Figure 4 the contour hierarchy relationship shown, the specific recognition process is as follows:

[0062] Step S2.1: Find contours 5, 6, and 7 by the number of same-level contours and the existence of their child contours;

[0063] Step S2.2: Find contour 3 through the parent-child contour relationship;

[0064] Step S2.3: Find contour 2 through the same-level contour of contour 3;

[0065] Step S2.4: Find contour 1 through the parent contour of contours 2 and 3, and find contour 4 through the child contour of contour 2;

[0066] Step S2.5: Distinguish contours 5, 6, and 7 by calculating the relative relationship between the centroids of contours 4, 5, 6, and 7 and the center of the marking plate. The specific method is: The atan2(y, x) function can be used to obtain the direction angles of the above four contours and the center of the marking plate. Sort them according to the size of the direction angle, and then cycle and roll to arrange the 4th contour in the first place, and the sequential order of the 4 key grid contours arranged clockwise starting from the 4th contour can be obtained. Then, according to the coordinates of the centroids of contours 4, 5, 6, and 7 obtained by the fixed installation function in the opencv library.

[0067] Specifically, the feature marker board adopts: as Figure 2 shown, the reflective marker board adopts the design of a whole board. At the same time, the outermost contour of the reflective marker board is an integral whole, avoiding the problem that individual bright spots are easily interfered by the background. There are 4 square grids on the reflective marker board, and their centers of gravity are used for PnP pose calculation.

[0068] Specifically, the step S3 adopts: based on the center-of-gravity coordinates of the contour and combined with the PnP algorithm to complete the real-time pose calculation of the moving target.

[0069] Specifically, the current marker board is identified by the preset ID pattern. Among them, the ID recognition algorithm of the dynamic target is used for ID recognition. The specific method is: according to different target objects, the variable marker board is marked with different IDs. Figure 2 Whether there is reflective material pasted at the ID marking position in can be marked as different IDs in binary coding. According to the size design, there are four marking bits, which can be marked (0000 2 ~1111 2 ) that is, a total of 2 4 = 16 different IDs, and some ID patterns are as Figure 3 shown.

[0070] According to a moving target recognition, real-time positioning and attitude calculation system provided by the present invention, including:

[0071] Module M1: Real-time obtain an image video containing a dynamic target, and preprocess the obtained image to obtain a preprocessed image; among them, use the feature marker board to cooperate with the infrared light source and infrared filter of the camera to real-time obtain an image video containing a dynamic target, and preprocess the obtained image video;

[0072] Module M2: Extract the image contour of the preprocessed image and the position coordinates of the feature points corresponding to the contour based on the contour recognition method based on the inclusion relationship;

[0073] Module M3: Complete the pose calculation and ID recognition of the dynamic target based on the position coordinates of the feature points corresponding to the contour;

[0074] The contour recognition method based on the inclusion relationship is to recognize the pattern according to the inclusion relationship between the contours with different features contained in the pattern, and obtain the position coordinates of the feature points corresponding to the contour. The recognition effect obtained in this way has the characteristics of being more robust and efficient.

[0075] Specifically, the module M1 adopts:

[0076] Module M1.1: Sharpen the image using the Laplacian of Gaussian operator. For different positions of the marker board, use convolution kernels of different sizes to obtain the sharpened image;

[0077] Module M1.2: Binarize the sharpened image using the Otsu adaptive binarization algorithm to obtain the binarized image.

[0078] Specifically, the module M2 uses: The preprocessed image uses the findContours function of the OpenCV open-source library to find the contours of the marker board and obtain the inclusion relationship of the found contours. Among them, the inclusion relationship and the obtained contour level relationship are expressed as Figure 4 shown. According to Figure 4 the level relationship of the contours in, the contours can be defined as the parent contour a and the child contour b according to the relationship between the contours, that is, the contour a contains the contour b, and the same-level contours c and d, that is, the contour e contains the contour c, and the contour e contains the contour d. For Figure 4 the contour level relationship shown, the specific recognition process is as follows:

[0079] Module M2.1: Find contours 5, 6, and 7 by the number of same-level contours and the existence of their child contours;

[0080] Module M2.2: Find contour 3 through the parent-child contour relationship;

[0081] Module M2.3: Find contour 2 through the same-level contour of contour 3;

[0082] Module M2.4: Find contour 1 through the parent contour of contours 2 and 3, and find contour 4 through the child contour of contour 2;

[0083] Module M2.5: Distinguish contours 5, 6, and 7 by calculating the relative relationship between the centroids of contours 4, 5, 6, and 7 and the center of the marker board. The specific method is: Use the atan2(y, x) function to obtain the direction angles of the above four contours and the center of the marker board. Sort them according to the size of the direction angles, and then cycle and roll to place contour 4 in the first position, and the sequential order of the 4 key grid contours arranged clockwise starting from contour 4 can be obtained. Then, according to the coordinates of the centroids of contours 4, 5, 6, and 7 obtained by the fixed installation function in the opencv library.

[0084] Specifically, the feature marker board uses: As Figure 2 shown, the reflective marker board uses a whole-board design. At the same time, the outermost contour of the reflective marker board is an entire whole, avoiding the problem that single bright spots are easily interfered by the background. There are 4 square grids on the reflective marker board, and its centroid is used for PnP pose solution.

[0085] Specifically, the module M3 adopts: combining the contour centroid coordinates with the PnP algorithm to complete the real-time pose solution of the moving target.

[0086] Specifically, the current marker board is identified by its ID according to a preset ID pattern. Among them, the ID recognition algorithm of the dynamic target is used for ID recognition. The specific method is: according to different target objects, the variable marker board is marked with different IDs. Figure 2 Whether there is a reflective material pasted at the ID marking position can be marked with different IDs in binary coding. According to the size design, there are four marking bits, which can mark (0000 2 ~1111 2 ) that is, a total of 2 4 = 16 different IDs. Some ID patterns are as Figure 3 shown.

[0087] Example 2

[0088] Embodiment 2 is a preferred example of Embodiment 1

[0089] A method for moving target recognition, real-time positioning and pose solution provided by the present invention. This specific example is established under the background of ID recognition and pose calculation of the ground-air cooperation system unmanned vehicle and unmanned aerial vehicle. Figure 5 (a) shows a schematic diagram of the unmanned vehicle and its equipped infrared camera, and (b) shows the unmanned aerial vehicle equipped with a marker board used in the example, including propeller 1, motor 2, marker point 3, receiver 4 and feature board 5. The system processes the images obtained by the infrared camera carried by the unmanned vehicle to real-time solve the pose and its ID number of the unmanned aerial vehicle, and communicates between the ground vehicle and the unmanned aerial vehicle through the UWB (Ultra Wide Band) module. The pose of the unmanned aerial vehicle real-time solved by the ground vehicle is sent to the unmanned aerial vehicle through the UWB module in real-time, so as to complete the pose feedback of the unmanned aerial vehicle to itself and realize the control of the unmanned aerial vehicle.

[0090] The marker board recognition and contour acquisition mainly include three parts: image preprocessing, image contour extraction and unmanned aerial vehicle ID recognition. Image preprocessing mainly includes two parts: image sharpening and image binarization. First, use the image sharpening algorithm to reduce the halo around the bright spot and make the bright and dark contours clearer. Here, the Laplacian of Gaussian kernel is convolved with the original image to complete the sharpening process of the image.

[0091] 0 mean, σ 2 The Gaussian kernel of is:

[0092]

[0093] Among them, x and y represent the coordinates of the x and y axes of the image, and σ 2Represents the variance of the Laplacian of Gaussian kernel

[0094] The finally obtained Laplacian of Gaussian kernel is:

[0095]

[0096] Among them, Represents the second-order partial derivative, Represents "denoted as", LoG represents the Laplacian of Gaussian, G σ (x, y) represents the Gaussian kernel with variance σ

[0097] Figure 6 Shows the comparison of contour extraction between the sharpened image and the original image. After the sharpening process, in order to facilitate contour finding, the sharpened image is binarized. Since the brightness levels are different at different positions of the image, the Otsu adaptive binarization algorithm is used here for binarization. Let the statistical histogram of the read image be represented as P, that is, P(i) represents the probability of pixel points with brightness value i.

[0098] Obviously, in an 8-bit grayscale image

[0099]

[0100] This algorithm finds the threshold t to minimize the weighted within-class variance after classification To the minimum.

[0101]

[0102] Among them, q is the sum of the respective probabilities of the two classes after being segmented by the threshold t, μ is the weighted average, and σ 2 Is the within-class variance.

[0103]

[0104]

[0105]

[0106] Among them, μ 1 (t) represents the weighted mean of class 1, μ 2 (t) represents the weighted mean of class 2, I = 256 which is the upper limit of the grayscale value for pixel binarization,

[0107] As long as the pixels from 0 to 255 are traversed once to calculate the weighted within-class variance corresponding to each threshold You can find The best binarization threshold corresponding to the minimum. Then, contour recognition is performed, and using the contour recognition method based on the inclusion relationship proposed in the present invention, the completion of Figure 4Recognition of the middle contours 4, 5, 6, and 7. After obtaining the centroid coordinates of the contours 4, 5, 6, and 7, the pose of the marker board can be solved according to the PnP algorithm. Finally, it is judged whether the marker board is successfully recognized. When the marker board is in the camera's field of view and the inner contour of the marker board is clear, it can be successfully recognized; if the marker board is successfully recognized, the positions q of the four ID marker points can be obtained. 1 , q 2 , q 3 , q 4 (arranged counterclockwise), find these four pixels q on the binarized image. 1 , q 2 , q 3 , q 4 's brightness I 1 , I 2 , I 3 , I 4 , assuming that the brightness of the binarized image is 0 or 1, the recognized ID is:

[0108] ID = 2 3 I 1 + 22I 2 + 2I 3 + I 4

[0109] The recognition effect is as Figure 6 shown.

[0110] The recognition algorithm is tested in various indoor and outdoor environments and can realize the recognition of the UAV marker board in various environments. In indoor environments such as Figure 8 shown, whether it is the light on the ceiling or complete darkness, it does not affect the reception of infrared light, and the marker board can be stable within a range of 9 m from the camera. In outdoor environments such as Figure 9 shown, if the camera is not directly facing the sun, it can be stably recognized within a range of 6 m, and the recognition effect is as Figure 10 and 11 shown.

[0111] Those skilled in the art know that in addition to implementing the systems, devices, and their various modules provided by the present invention in the form of pure computer-readable program codes, the method steps can be logically programmed to enable the systems, devices, and their various modules provided by the present invention to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers, etc. to achieve the same program. Therefore, the systems, devices, and their various modules provided by the present invention can be regarded as a kind of hardware component, and the modules included therein for implementing various programs can also be regarded as the structures within the hardware component; the modules for implementing various functions can also be regarded as both software programs for implementing the method and the structures within the hardware component.

[0112] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for identifying moving targets, real-time positioning, and attitude calculation, characterized in that, it includes: Step S1: Real-time obtain an image video containing dynamic targets, and preprocess the obtained image video to obtain a preprocessed image; Step S2: Extract the image contour of the preprocessed image and the position coordinates of the feature points corresponding to the contour based on the contour recognition method based on the inclusion relationship; Step S3: Complete the pose calculation and ID recognition of the dynamic target based on the position coordinates of the feature points corresponding to the contour; The contour recognition method based on the inclusion relationship is to identify the pattern according to the inclusion relationship between the contours with different features contained in the pattern and obtain the position coordinates of the feature points corresponding to the contour; In step S1, a feature marker board is used in cooperation with the infrared light source and infrared filter of the camera to real-time obtain an image video containing dynamic targets; The feature marker board is a reflective marker board with a whole-board design, and at the same time, the outermost contour of the reflective marker board is an integral whole; Perform ID recognition on the current marker board according to the preset ID pattern.

2. The method for identifying moving targets, real-time positioning, and attitude calculation according to claim 1, characterized in that, step S1 adopts: Step S1.1: Use the Laplacian of Gaussian operator to sharpen the image to obtain a sharpened image; Step S1.2: Use the Otsu adaptive binarization algorithm to binarize the sharpened image to obtain a binarized image.

3. The method for identifying moving targets, real-time positioning, and attitude calculation according to claim 1, characterized in that, in step S2, the findContours function of the OpenCV open source library is used for the preprocessed image to find the contour of the marker board and obtain the inclusion relationship of the found contour.

4. The method for identifying moving targets, real-time positioning, and attitude calculation according to claim 1, characterized in that, in step S3, based on the contour centroid coordinates and the PnP algorithm, the real-time pose calculation of the moving target is completed.

5. A system for identifying moving targets, real-time positioning, and attitude calculation, characterized in that, it includes: Module M1: Real-time obtain an image video containing dynamic targets, and preprocess the obtained image video to obtain a preprocessed image; Module M2: Extract the image contour of the preprocessed image and the position coordinates of the feature points corresponding to the contour based on the contour recognition method based on the inclusion relationship; Module M3: Complete the pose calculation and ID recognition of the dynamic target based on the position coordinates of the feature points corresponding to the contour; The contour recognition method based on the inclusion relationship is to identify the pattern according to the inclusion relationship between the contours with different features contained in the pattern and obtain the position coordinates of the feature points corresponding to the contour; Module M1 adopts: A feature marker board is used in cooperation with the infrared light source and infrared filter of the camera to real-time obtain an image video containing dynamic targets; The feature marker board is a reflective marker board with a whole-board design, and at the same time, the outermost contour of the reflective marker board is an integral whole; Module M1 adopts: Module M1.1: Sharpen the image using the Laplacian of Gaussian operator to obtain the sharpened image; Module M1.2: Binarize the sharpened image using the Otsu adaptive binarization algorithm to obtain the binarized image; Perform ID recognition on the current marker board according to the preset ID pattern.

6. The moving target recognition, real-time positioning and attitude solution system according to claim 5, wherein, the module M2 adopts: the preprocessed image uses the findContours function of the OpenCV open source library to find the contours of the marker board and obtain the inclusion relationship of the found contours.

7. The moving target recognition, real-time positioning and attitude solution system according to claim 5, wherein, the module M3 adopts: based on the contour centroid coordinates and combined with the PnP algorithm to complete the real-time pose solution of the moving target.

Images