Method and device for automatically identifying and tracking targets

Abstract

The present application discloses a method and apparatus for automatically identifying and tracking a target, the method comprising: creating a data set comprising multiple images, wherein the images contain the target to be identified; training a target detection model using the data set of labeled targets; identifying the target using the trained target detection model; determining an aiming point on the target and calculating the pixel coordinates of the aiming point; searching a point cloud and determining the point cloud information of the aiming point; and controlling the movement of a radar based on the point cloud information of the aiming point so that the center of the radar's field of view is aligned with the aiming point. The method provided in the embodiments of the present application avoids the influence of humans on target identification. As long as the target is in the radar's field of view, it can be continuously tracked. Even if the target may be obscured or temporarily disappear during movement, it will be re-locked as soon as the target reappears.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to a method and device for automatically identifying and tracking a target. Background Art

[0002] At present, intelligent weapon systems based on information technology and unmanned mobile platform technology are receiving increasing attention from countries around the world as an important component of future wars. Countries around the world are conducting in-depth research on them as the technological commanding heights of new combat forces.

[0003] Despite the rapid development of intelligent unmanned systems, their aiming systems still need improvement. Currently, popular aiming systems include manual aiming via video and target tracking and aiming based on correlation filtering. However, manual aiming via video is subject to significant human influence. Firstly, it requires high operator quality, which undoubtedly increases hidden costs. Secondly, operator stability is questionable and can result in significant errors. Furthermore, target tracking and aiming based on correlation filtering has its limitations. The target may become occluded or temporarily disappear during motion. When this occurs, the tracking frame easily includes the occluder and background information, causing the tracked target in subsequent frames to drift over the occluder. Summary of the Invention

[0004] The embodiments of the present application provide a method and apparatus for automatically identifying and tracking targets, thereby solving the technical problems in the prior art that manual aiming in videos is significantly affected by humans, and that target tracking and aiming based on correlation filtering also has its limitations.

[0005] In a first aspect, the present application provides a method for automatically identifying and tracking targets, the method comprising: preparing a data set comprising multiple images; wherein the images contain targets to be identified; training a target detection model using the data set in which the targets have been labeled; identifying the target using the trained target detection model; determining a targeting point on the target and calculating the pixel coordinates of the targeting point; searching a point cloud and determining point cloud information of the targeting point; and controlling radar movement based on the point cloud information of the targeting point so that the center of the radar's field of view is aligned with the targeting point.

[0006] In combination with the first aspect, in a possible implementation, determining the aiming point on the target includes: cutting out a target image based on the identified bounding box of the target, and converting the target image into an edge detection binary image; calculating the domain gradient value of each pixel in the edge detection binary image; drawing a line along the gradient direction of each non-zero pixel in the edge retrieval binary image; counting the number of straight lines passed by each pixel in the edge detection binary image, and determining the pixel with the largest number of line segments passed as the aiming point.

[0007] In combination with the first aspect, in a possible implementation method, converting the target image into an edge detection binary image includes: performing grayscale processing on the target image; calculating the gradient amplitude and direction of each pixel in the target image; performing non-maximum suppression and solving the two grayscale values ​​of each pixel along the gradient direction within the neighborhood; judging whether each pixel is an edge pixel based on the two grayscale values ​​of each pixel, and setting the grayscale value of the non-edge pixel to 0; using a dual threshold algorithm to detect and connect the edge pixels to obtain the edge detection binary image.

[0008] In combination with the first aspect, in a possible implementation method, converting the target image into an edge detection binary image also includes: comparing the numerical relationship between the gradient value of the edge pixel point, a preset high threshold and a preset low threshold; if the gradient value of the edge pixel point is higher than the preset high threshold, marking the edge pixel point as a strong edge pixel point; if the gradient value of the edge pixel point is higher than the preset low threshold and lower than the preset high threshold, marking the edge pixel point as a weak edge pixel point; if the gradient value of the edge pixel point is lower than the preset low threshold, suppressing the edge pixel point; connecting the edge pixel points to obtain the edge detection binary image includes: connecting all the strong edge pixels and the weak edge pixels connected to the strong edge pixels.

[0009] In combination with the first aspect, in a possible implementation manner, converting the target image into an edge-detected binary image further includes: performing Gaussian filtering on the target image after performing grayscale processing on the target image.

[0010] In combination with the first aspect, in a possible implementation, after executing the search point cloud, the searched point cloud is cleaned to remove isolated points, and a "circular" search algorithm is used to obtain all point cloud data of the target.

[0011] In combination with the first aspect, in a possible implementation, when performing the identifying of the target using the target detection model, an accelerator is used for acceleration.

[0012] In a second aspect, an embodiment of the present application provides a device for automatically identifying and tracking a target, the device comprising: a data set production module for producing a data set comprising multiple images; wherein the images contain the target to be identified; a training module for training a target detection model using the data set in which the target has been labeled; an identification module for identifying the target using the trained target detection model; an aiming point determination module for determining the aiming point on the target; a pixel coordinate calculation module for calculating the pixel coordinates of the aiming point; a point cloud module for searching the point cloud and determining the point cloud information of the aiming point; and an aiming module for controlling the movement of the radar based on the point cloud information of the aiming point so that the center of the radar's field of view is aligned with the aiming point.

[0013] In combination with the second aspect, in a possible implementation, the aiming point determination module is specifically used to: cut out a target image based on the identified bounding box of the target, and convert the target image into an edge detection binary image; calculate the domain gradient value of each pixel point in the edge detection binary image; draw a line along the gradient direction of each non-zero pixel point in the edge retrieval binary image; count the number of straight lines passed by each pixel point in the edge detection binary image, and determine the pixel point with the largest number of line segments passed as the aiming point.

[0014] In combination with the second aspect, in a possible implementation method, the aiming point determination module is used to convert the target image into an edge detection binary image, specifically for: performing grayscale processing on the target image; calculating the gradient amplitude and direction of each pixel in the target image; performing non-maximum suppression and solving the two grayscale values ​​of each pixel along the gradient direction within the neighborhood; judging whether each pixel is an edge pixel based on the two grayscale values ​​of each pixel, and setting the grayscale value of the non-edge pixel to 0; using a dual threshold algorithm to detect and connect the edge pixels to obtain the edge detection binary image.

[0015] In combination with the second aspect, in a possible implementation, the aiming point determination module is used to convert the target image into an edge detection binary image and is also used to: compare the numerical relationship between the gradient value of the edge pixel point, a preset high threshold and a preset low threshold; if the gradient value of the edge pixel point is higher than the preset high threshold, the edge pixel point is marked as a strong edge pixel point; if the gradient value of the edge pixel point is higher than the preset low threshold and lower than the preset high threshold, the edge pixel point is marked as a weak edge pixel point; if the gradient value of the edge pixel point is lower than the preset low threshold, the edge pixel point is suppressed; and connecting the edge pixel points to obtain the edge detection binary image includes: connecting all the strong edge pixels and the weak edge pixels connected to the strong edge pixels.

[0016] In combination with the second aspect, in a possible implementation, when the aiming point determination module is used to convert the target image into an edge-detected binary image, it is also used to: perform Gaussian filtering on the target image after performing grayscale processing on the target image.

[0017] In combination with the second aspect, in a possible implementation, the point cloud module is further used to clean the searched point cloud to remove isolated points after executing the search point cloud, and adopt a "U-shaped" search algorithm to obtain all point cloud data of the target.

[0018] In combination with the second aspect, in a possible implementation, the apparatus further includes an acceleration module configured to use an accelerator to perform acceleration when the recognition module performs the recognition of the target using the target detection model.

[0019] In combination with the second aspect, in one possible implementation, the device also includes a fitting module; the aiming module is further used to control the center of the field of view of the radar to align with the aiming point at multiple positions at different distances from the target; the fitting module is used to fit a straight line using multiple aiming points, and use the straight line as the weapon firing line.

[0020] In a third aspect, an embodiment of the present application provides an automatic identification and tracking device, which includes a memory and a processor; the memory is used to store computer-executable instructions; the processor is used to execute the computer-executable instructions, and can implement the method described in the first aspect and various possible implementation methods of the first aspect.

[0021] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores executable instructions, and a computer executing the executable instructions can implement the method described in the first aspect and various possible implementation methods of the first aspect.

[0022] The technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0023] The method for automatically identifying and tracking targets provided in the embodiment of the present application uses a data set of labeled targets to train a target detection model, so that the trained detection model has the ability to identify targets. The trained target detection model is then used to identify targets without the need for manual aiming, thereby avoiding the influence of humans on target recognition. The method for automatically identifying and tracking targets provided in the embodiment of the present application finally controls the movement of the radar based on the point cloud information of the aiming point on the target so that the center of the radar's field of view is aligned with the aiming point, so that the trained target detection model is integrated with the radar data. As long as the target is in the radar's field of view, it can be continuously tracked. Even if the target may be obscured or disappear briefly during movement, it will be re-locked as soon as the target reappears. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments of the present application or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 A flowchart of a method for automatically identifying and tracking a target provided in an embodiment of the present application;

[0026] Figure 2 A flowchart for determining an aiming point on a target provided in an embodiment of the present application;

[0027] Figure 3 A flowchart of converting a target image into an edge detection binary image provided in an embodiment of the present application;

[0028] Figure 4 A flowchart showing a comparison and determination of the numerical relationship between the gradient value of an edge pixel point, a preset high threshold, and a preset low threshold provided in an embodiment of the present application;

[0029] Figure 5 A schematic diagram of the structure of an apparatus for automatically identifying and tracking a target provided in an embodiment of the present application;

[0030] Figure 6 This is a schematic diagram of the structure of the automatic identification and tracking device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0032] To facilitate understanding, the terms involved in the embodiments of this application are first explained.

[0033] A bounding box (BBOX) is a rectangular box that can be determined by the pixel coordinates of the upper left corner and the lower right corner of the rectangle, and is used to frame the location of an object.

[0034] Non-maximum suppression (NMS) refers to suppressing elements that are not maximum values, which can be understood as a local maximum search. This local represents a neighborhood, and the neighborhood has two variable parameters, one is the dimension of the neighborhood, and the other is the size of the neighborhood. It is mainly used to "thin" edges. After applying the gradient calculation, the edges extracted from the gradient values ​​are still very blurred. Non-maximum suppression can help suppress all gradient values ​​except the local maximum (by setting them to 0), which indicates the location with the strongest intensity value change.

[0035] An operator is a mapping from a function space to another function space.

[0036] Grayscale refers to a logarithmic range between white and black. It typically ranges from 0 to 255, with white at 255 and black at 0. In binary images, there are only two grayscale levels. That is, any pixel in the image has a grayscale value of either 0 or 255, representing black and white, respectively.

[0037] Non-zero pixels refer to pixels whose gradient values ​​are not zero.

[0038] Point cloud refers to a massive collection of points that represent the surface characteristics of an object.

[0039] The present application provides a method for automatically identifying and tracking a target, including: Figure 1 Steps S101 to S107 are shown.

[0040] S101: Create a dataset of multiple images. The images in the dataset contain objects to be identified. The objects to be identified can be buildings, vehicles, targets, etc.

[0041] S102. Train the target detection model using a dataset of labeled targets. It should be noted that the target annotation in the dataset is performed manually; for example, when the target to be identified is a vehicle, the annotator marks the vehicle in each image with a box. After the target detection model is trained using a dataset of labeled targets, the target detection model has the ability to automatically identify targets. Specifically, when the labeled target is a vehicle, the target detection model can automatically identify the vehicle; when the labeled target is a building, the target detection model can automatically identify the building; when the labeled target is a target, the target detection model can automatically identify the target.

[0042] S103: Use the trained target detection model to identify the target. Specifically, the target detection model identifies the target from the image and draws a bounding box around the target.

[0043] S104: Determine the aiming point on the target. The aiming point is the point you want to aim at. The specific determination process can be a point manually given on the identified target, a point randomly determined by a computer on the target, or a specific point determined by the computer according to a specific algorithm.

[0044] The embodiment of the present application provides a specific process for determining the aiming point on the target when the target is a target and the aiming point is the bull's eye of the target, including the following steps: Figure 2 Steps S201 to S205 are shown.

[0045] S201: Extract the target image based on the identified target's bounding box. The background outside the bounding box has no practical effect on determining the aiming point. Step S201 discards the background image outside the bounding box, retaining only the image inside the bounding box. This reduces the computational effort in subsequent steps, improves computational efficiency, and reduces interference from the background image in subsequent calculations.

[0046] S202: Convert the target image into an edge-detected binary image. The edge-detected binary image here means that the edges of the target are white and the non-edge parts of the image are black.

[0047] The embodiment of the present application provides a specific implementation method for converting a target image into an edge detection binary image, including: Figure 3 Steps S301 to S305 are shown.

[0048] S301: Perform grayscale processing on the target image.

[0049] S302. Calculate the gradient magnitude and direction of each pixel in the target image. Specifically, the Sobel operator can be used to calculate the gradient magnitude and direction. The Sobel operator's x-direction convolution template, y-direction convolution template, and neighborhood point marker matrix of the point to be processed are shown in the following formula:

[0050]

[0051]

[0052]

[0053] Based on the above x-direction convolution template, y-direction convolution template and the neighborhood point label matrix of the point to be processed, the gradient amplitude of each pixel can be calculated by the following formula:

[0054]

[0055] S x =(a2+2a3+a4)-(a0+2a7+a6);

[0056] S y =(a0+2a1+a2)-(a6+2a5+a4).

[0057] S303: Perform non-maximum suppression and solve the two grayscale values ​​of each pixel point along the gradient direction in the neighborhood.

[0058] S304 , judging whether each pixel is an edge pixel according to the two grayscale values ​​of each pixel, and setting the grayscale value of the pixel determined to be a non-edge pixel to 0.

[0059] S305: Use a dual threshold algorithm to detect and connect edge pixels to obtain an edge detection binary image. Specifically, a Canny detector may be used, but other types of detectors may also be used, which is not limited in this embodiment of the present application.

[0060] The process of converting the target image into an edge detection binary image includes steps S301 to S305, as follows: Figure 4 Steps S401 to S404 are shown.

[0061] S401 : Compare the numerical relationships among the gradient value of the edge pixel point, the preset high threshold value, and the preset low threshold value.

[0062] If the gradient value of the edge pixel is higher than the preset high threshold, S402 is executed to mark the edge pixel as a strong edge pixel.

[0063] If the gradient value of the edge pixel is higher than the preset lower threshold and lower than the preset upper threshold, S403 is executed to mark the edge pixel as a weak edge pixel.

[0064] If the gradient value of the edge pixel point is lower than the preset low threshold, S404 is executed to suppress the edge pixel point.

[0065] After performing any one of steps S402 to S404 , connecting edge pixels to obtain an edge detection binary image specifically includes: connecting all strong edge pixels and weak edge pixels connected to the strong edge pixels.

[0066] Converting the target image into an edge detection binary image also includes: performing a Gaussian filter on the target image after grayscale processing. The Gaussian function is specifically as follows:

[0067]

[0068] Where σ is the variance. Determining these parameters yields a one-dimensional kernel vector. The choice of Gaussian convolution kernel size affects the performance of the Canny detector. Larger Gaussian convolution kernels reduce the detector's sensitivity to noise, but also slightly increase the edge detector's localization error. Gaussian convolution typically uses a 5×5 kernel size.

[0069] After executing the search point cloud, the searched point cloud is cleaned to remove isolated points. This removes the noise from isolated points in the direct search point cloud and avoids random errors. A zigzag search algorithm is then used to obtain the complete point cloud data for the target. Of course, the point cloud data obtained by the zigzag search algorithm may contain point cloud data from a small area around the target, in addition to the complete point cloud data for the target.

[0070] The above is only a specific example of converting a target image into an edge detection binary image provided in an embodiment of the present application. It does not mean that the target image can only be converted into an edge detection binary image through the above implementation method. It can also be achieved through many other methods, and the embodiment of the present application does not limit this.

[0071] After the target image is converted into an edge detection binary image, S203 is executed to calculate the neighborhood gradient value of each pixel in the edge detection binary image. The specific calculation method in S203 can still be performed using the Sobel operator, which will not be further described.

[0072] S204: Draw a line along the gradient direction of each non-zero pixel in the edge-detected binary image. The non-zero pixel is the edge of the white target in the edge-detected binary image. Since the target is a circular edge, the gradient direction of the non-zero pixel is theoretically the radial direction of the circle.

[0073] S205. Count the number of straight lines passed by each pixel in the edge detection binary image, and determine the pixel with the largest number of line segments as the target point. Theoretically, straight lines drawn along the gradient direction of each non-zero pixel in the edge detection binary image should pass through the target point, and the intersection of two straight lines can be used to determine the target point. However, due to errors, white pixels may exist in the non-edge portion of the actual edge detection binary image, and the intersection of two straight lines may not be the desired target point. By counting the number of straight lines passed by each pixel in the edge detection binary image and determining the pixel with the largest number of line segments as the target point, the influence of errors on the determination of the target point can be minimized.

[0074] It should be noted that steps S201 to S205 are only a specific example of the embodiment of the present application for determining the aiming point on the target, and do not mean that the aiming point on the target can only be determined through S201 to S205. It can also be achieved through many other methods, and the embodiment of the present application does not limit this.

[0075] S105. Calculate the pixel coordinates of the aiming point. It should be noted that images are composed of pixels, and pixel coordinates are the positions of pixels in the image. To determine the coordinates of a pixel, the image coordinate system must first be determined. Common coordinate systems include the image coordinate system, the camera coordinate system, and the world coordinate system. Taking the image coordinate system as an example, a direct coordinate system uv is established with the upper left corner of the image as the origin, using pixels as the unit. The horizontal coordinate u and vertical coordinate v of a pixel are the column and row numbers, respectively, of the pixel in the image array.

[0076] S106: Search the point cloud and determine the point cloud information of the aiming point. The point cloud information includes at least the three-dimensional coordinates of the point, and may also include color information and / or reflection intensity information.

[0077] S107. Control the radar's movement based on the point cloud information of the aiming point so that the center of the radar's field of view is aligned with the aiming point. Specifically, the determined point cloud information of the aiming point is transmitted to the radar's pan / tilt (PTZ), which drives the radar to rotate and align the center of the radar's field of view with the aiming point. The pan / tilt's horizontal rotation angle w and vertical rotation angle h can be calculated using the following formulas.

[0078]

[0079]

[0080] Among them, the radar coordinates are (x1, y1, z1) and the aiming point coordinates are (x2, y2, z2).

[0081] The method for automatically identifying and tracking a target provided in an embodiment of the present application uses an accelerator to accelerate the target detection model when identifying the target, so that the target detection model can quickly identify the target, thereby improving the recognition efficiency of the method. Specifically, the accelerator can be a TensorRT accelerator.

[0082] The target detection model used in the embodiments of the present application may be a YOLO model.

[0083] The method for automatically identifying and tracking targets provided in an embodiment of the present application also includes: controlling the center of the field of view of the radar to align with the aiming point at multiple positions at different distances from the target; using multiple aiming points to fit a straight line, and using the straight line as the weapon firing line.

[0084] For example, control the radar at 10m, 20m, 30m, 40m, and 50m away from the target, so that the center of the radar's field of view is aligned with the aiming point, and take the pixel coordinates of the five aiming points and the corresponding point cloud information. The pixel coordinates are (x i ,y i ), the distance information in the point cloud information is z i , and then fit a straight line in the field of view. This straight line is the weapon's firing line. The specific fitting process can be done using the least squares method to find the minimum R in the following formula:

[0085]

[0086] r is the distance from the point to the fitted line in the y direction.

[0087] y=ax+b;

[0088] R=(x1a+b-y1) 2 +(x2a+b-y2) 2 +…+(x5a+b-y5) 2 .

[0089] Find a and b to get the fitting line, and calculate z i When aiming is needed, you only need to identify the aiming point, obtain the pixel coordinates and distance information from the point cloud, calculate the pixel coordinates of the fitted line based on the distance information, and then rotate the gimbal so that the pixel coordinates of the bull's eye coincide with the pixel position of the fitted line at the same distance to complete the aiming.

[0090] The method for automatically identifying and tracking targets provided in the embodiment of the present application uses a data set of labeled targets to train a target detection model, so that the trained detection model has the ability to identify targets. The trained target detection model is then used to identify targets without the need for manual aiming, thereby avoiding the influence of humans on target recognition. The method for automatically identifying and tracking targets provided in the embodiment of the present application finally controls the movement of the radar based on the point cloud information of the aiming point on the target so that the center of the radar's field of view is aligned with the aiming point, so that the trained target detection model is integrated with the radar data. As long as the target is in the radar's field of view, it can be continuously tracked. Even if the target may be obscured or disappear briefly during movement, it will be re-locked as soon as the target reappears.

[0091] Although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative work. The order of steps listed in this embodiment is only one way of executing the steps among many steps and does not represent the only execution order. When an actual device or client product is executed, the method shown in this embodiment or the accompanying drawings may be executed sequentially or in parallel (for example, in a parallel processor or multi-threaded processing environment).

[0092] The embodiment of the present application provides a device 500 for automatically identifying and tracking a target, such as Figure 5 As shown, the apparatus includes: a data set preparation module 501, a training module 502, an identification module 503, an aiming point determination module 504, a pixel coordinate calculation module 505, a point cloud module 506, and an aiming module 507. The data set preparation module 501 is used to prepare a data set comprising multiple images, wherein the images contain targets to be identified; the training module 502 is used to train a target detection model using a data set of labeled targets; the identification module 503 is used to identify targets using the trained target detection model; the aiming point determination module 504 is used to determine the aiming point on the target; the pixel coordinate calculation module 505 is used to calculate the pixel coordinates of the aiming point; the point cloud module 506 is used to search the point cloud and determine the point cloud information of the aiming point; and the aiming module 507 is used to control the movement of the radar based on the point cloud information of the aiming point so that the center of the radar field of view is aligned with the aiming point.

[0093] The aiming point determination module 504 is specifically used to: extract the target image based on the bounding box of the identified target and convert the target image into an edge detection binary image; calculate the neighborhood gradient value of each pixel in the edge detection binary image; draw a line along the gradient direction of each non-zero pixel in the edge detection binary image; count the number of straight lines passed by each pixel in the edge detection binary image, and determine the pixel with the largest number of line segments as the aiming point.

[0094] The aiming point determination module 504 is used to convert the target image into an edge detection binary image, specifically for: performing grayscale processing on the target image; calculating the gradient amplitude and direction of each pixel in the target image; performing non-maximum suppression and solving the two grayscale values ​​of each pixel along the gradient direction within the neighborhood; judging whether each pixel is an edge pixel based on the two grayscale values ​​of each pixel, and setting the grayscale value of the pixel determined to be a non-edge pixel to 0; and using a dual threshold algorithm to detect and connect the edge pixels to obtain an edge detection binary image.

[0095] When the aiming point determination module 504 is used to convert the target image into an edge detection binary image, it is also used to: compare the numerical relationship between the gradient value of the edge pixel point, the preset high threshold value and the preset low threshold value; if the gradient value of the edge pixel point is higher than the preset high threshold value, the edge pixel point is marked as a strong edge pixel point; if the gradient value of the edge pixel point is higher than the preset low threshold value and lower than the preset high threshold value, the edge pixel point is marked as a weak edge pixel point; if the gradient value of the edge pixel point is lower than the preset low threshold value, the edge pixel point is suppressed; connecting the edge pixels to obtain the edge detection binary image includes: connecting all strong edge pixels and weak edge pixels connected to the strong edge pixels.

[0096] When the aiming point determination module 504 is used to convert the target image into an edge detection binary image, it is also used to: perform Gaussian filtering on the target image after performing grayscale processing on the target image.

[0097] The point cloud module 506 is further configured to clean the searched point cloud to remove isolated points after executing the search point cloud, and to obtain all point cloud data of the target using a “circular” search algorithm.

[0098] The apparatus 500 for automatically identifying and tracking a target provided in an embodiment of the present application further includes an acceleration module for using an accelerator to accelerate when the identification module performs target identification using a target detection model.

[0099] The device for automatically identifying and tracking targets provided in an embodiment of the present application also includes a fitting module; the aiming module is also used to control the center of the radar's field of view to align with the aiming point at multiple positions at different distances from the target; the fitting module is used to use multiple aiming points to fit a straight line, and use the straight line as the weapon firing line.

[0100] Some modules in the apparatus described herein may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform specific tasks or implement specific abstract data types. Embodiments of the present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0101] An embodiment of the present application also provides an automatic identification and tracking device 600, which includes a memory 601 and a processor 602; the memory 601 is used to store computer-executable instructions; the processor 602 is used to execute computer-executable instructions, which can implement the method of automatic target identification and tracking provided above in the embodiment of the present application.

[0102] The controller can take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the memory control logic. Those skilled in the art will also appreciate that, in addition to implementing the controller in pure computer-readable program code, it is entirely possible to implement the same functionality in the form of logic gates, switches, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the means for implementing various functions included therein can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered both a software module implementing the method and a structure within the hardware component.

[0103] The embodiment of the present application also provides a computer-readable storage medium, which stores executable instructions. The computer executes the executable instructions to implement the method of automatically identifying and tracking the target provided in the embodiment of the present application. The above-mentioned computer-readable storage medium includes but is not limited to a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk (HDD), or a memory card. The memory can be used to store computer program instructions.

[0104] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other. Each embodiment focuses on the differences from other embodiments.

[0105] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit the present application. Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some or all of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present application.

Claims

1. A method for automatically identifying and tracking a target, characterized in that: include: Creating a data set comprising a plurality of images; wherein the images contain objects to be identified; Training a target detection model using the dataset in which the targets have been labeled; Identify the target using the trained target detection model; determining an aiming point on the target and calculating pixel coordinates of the aiming point; Searching the point cloud and determining the point cloud information of the aiming point; Controlling the movement of the radar according to the point cloud information of the aiming point so that the center of the field of view of the radar is aligned with the aiming point; At a plurality of positions at different distances from the target, controlling the center of the field of view of the radar to align with the aiming point; A straight line is fitted using the plurality of aiming points, and the straight line is used as the weapon firing line.

2. The method for automatically identifying and tracking a target according to claim 1, wherein: Determining the aiming point on the target comprises: Cutting out a target image according to the identified bounding box of the target, and converting the target image into an edge detection binary image; Calculating the neighborhood gradient value of each pixel in the edge detection binary image; Draw a line along the edge retrieval gradient direction of each non-zero pixel in the binary image; The number of straight lines passed by each pixel in the edge detection binary image is counted, and the pixel passing the largest number of line segments is determined as the aiming point.

3. The method for automatically identifying and tracking a target according to claim 2, wherein: The converting of the target image into an edge detection binary image comprises: Performing grayscale processing on the target image; Calculating the gradient magnitude and direction of each pixel in the target image; Perform non-maximum suppression and solve the two grayscale values ​​of each pixel along the gradient direction in the neighborhood; Determine whether each pixel is an edge pixel based on its two grayscale values, and set the grayscale value of the non-edge pixel to 0; A dual threshold algorithm is used to detect and connect the edge pixels to obtain the edge detection binary image.

4. The method for automatically identifying and tracking a target according to claim 3, wherein: The step of converting the target image into an edge detection binary image further comprises: Comparing the numerical relationship between the gradient value of the edge pixel point, the preset high threshold value and the preset low threshold value; If the gradient value of the edge pixel point is higher than the preset high threshold, the edge pixel point is marked as a strong edge pixel point; if the gradient value of the edge pixel point is higher than the preset low threshold and lower than the preset high threshold, the edge pixel point is marked as a weak edge pixel point; if the gradient value of the edge pixel point is lower than the preset low threshold, the edge pixel point is suppressed; Connecting the edge pixel points to obtain the edge detection binary image includes: connecting all the strong edge pixel points and the weak edge pixel points connected to the strong edge pixel points.

5. The method for automatically identifying and tracking a target according to claim 3 or 4, characterized in that: The step of converting the target image into an edge detection binary image further comprises: After grayscale processing is performed on the target image, Gaussian filtering is performed on the target image.

6. The method for automatically identifying and tracking a target according to claim 1, wherein: After executing the search point cloud, the searched point cloud is cleaned to remove isolated points, and a "circular" search algorithm is used to obtain all point cloud data of the target.

7. The method for automatically identifying and tracking a target according to claim 1, wherein: When performing the identifying of the target using the target detection model, an accelerator is used for acceleration.

8. A device for automatically identifying and tracking a target, characterized in that: include: A data set preparation module, configured to prepare a data set comprising a plurality of images, wherein the images include objects to be identified; A training module, configured to train a target detection model using the dataset in which the targets have been labeled; An identification module, configured to identify the target using the trained target detection model; an aiming point determination module, configured to determine an aiming point on the target; A pixel coordinate calculation module, used to calculate the pixel coordinates of the aiming point; A point cloud module, configured to search the point cloud and determine the point cloud information of the aiming point; A targeting module, configured to control the movement of the radar according to the point cloud information of the targeting point so that the center of the field of view of the radar is aligned with the targeting point; The aiming module is further configured to control the center of the field of view of the radar to align with the aiming point at multiple positions at different distances from the target; The aiming module is further configured to fit a straight line using the plurality of aiming points, and use the straight line as a weapon firing line.

9. An automatic identification and tracking device, characterized in that: including memory and processor; The memory is used to store computer-executable instructions; The processor is configured to execute the computer-executable instructions and can implement the method according to any one of claims 1 to 7.

Images