A Bird Detection and Tracking Method and System Based on Infrared Target Recognition

Abstract

This invention belongs to the field of automation technology, specifically a bird detection and tracking method and system based on infrared target recognition. It includes an infrared imaging device, a two-dimensional gimbal, a bird deterrent device, and a main controller. The system identifies and tracks birds in complex environments based on infrared images. It determines the bird's flight trajectory based on the detected bird's location information, thereby adjusting the attitude of the two-dimensional gimbal to track and directionally disperse the bird target. This system is mainly used for bird deterrence in specific areas such as substations, fruit orchards, and airports. The system has a high bird recognition rate, stable deterrence effect, and the directional bird deterrence operation effectively reduces bird deterrence failure caused by bird adaptation.

Description

Technical Field

[0001] This invention belongs to the field of automation engineering technology, specifically an automated system for bird directional tracking and dispersal based on infrared target recognition. Background Technology

[0002] In recent years, bird infestations at substations and converter stations have become increasingly serious. This is because some equipment in substations generates heat and warmth, creating a suitable environment for birds to live and move around, easily attracting them and causing bird damage. Previously, bird damage prevention could only be achieved by strengthening the structure of the main transformer and the portal frame of the transmission lines. In the past, bird damage prevention mainly relied on treatment and removal, making the entire process very passive and inefficient.

[0003] Meanwhile, bird control measures are being promoted in substations and other similar locations. Existing bird control measures are mainly divided into three categories: prevention, evacuation, and repulsion. Prevention refers to insulating equipment, such as installing bird spikes, bird barriers, bird-proof boxes, and bird-proof needle plates. These methods are effective initially, but over time, birds adapt to the environment, and equipment cleaning and maintenance become extremely limited. Bird evacuation involves reducing bird density in substations and similar locations by installing artificial nests and perches. Bird repulsion uses physical methods such as sound, light, and electricity, or specialized biological bird repellents. Currently, to achieve optimal control results, prevention, evacuation, and repulsion methods need to be used in combination, especially bird repulsion, which serves as the last line of defense to protect equipment from bird infestations; its effectiveness and reliability are particularly important. Summary of the Invention

[0004] To address the aforementioned issues and considering the lack of specificity in current bird dispersal methods and the ease with which birds can adapt, a bird detection and tracking system based on infrared target recognition was designed. This system includes device design and detection algorithm research.

[0005] The technical solution adopted by this invention to achieve the above objectives is: a bird detection and tracking method based on infrared target recognition, comprising the following steps:

[0006] Moving target detection based on background subtraction is performed on video image sequences to obtain the foreground and background of each frame in the video image sequence, with the foreground being the target;

[0007] For the target, the difference between the target and the background is obtained by measuring the local brightness difference of the window, thus obtaining the bird target.

[0008] The step of performing motion target detection based on background subtraction on the acquired images to obtain the foreground and background of each frame in the video image sequence includes the following steps:

[0009] Each frame of the input video image sequence is compared with the background model: if the pixel information of the image matches the pixel information of the background model, it is marked as a background point; otherwise, it is classified as a foreground point.

[0010] The background model is represented as follows:

[0011] B(x i )={B1(x i ), ..., B j (x i ), ..., B N (x i )}

[0012] Where N represents the number of samples, B j (x i ) represents x i The j-th sample pixel information is used as the sample value; for the first and second sample values, they are directly used as the first frame image x. i Pixel value V(x) at i The following N-2 samples are represented by pixel values ​​V(x). i Add random noise within a set threshold range t to fill the noise.

[0013] For each frame starting from the second frame, calculate the current frame image x. i Pixel value V(x) at i x in the background model i Sample value B at point k (x i The distance difference;

[0014] When the distance difference is within the threshold range R, and the number of samples in the background model that meets the threshold range is not less than # min If the pixel is not a background pixel, it is classified as a background pixel; otherwise, it is classified as a foreground pixel. In the following formula, the background pixel is marked as 0, and the foreground pixel is marked as 1:

[0015]

[0016] Where F(x) i ) represents the distance difference between a pixel and the sample.

[0017] The pixel information is grayscale value or infrared image grayscale value.

[0018] For a video image sequence, count the number of pixels x. i The number of times a pixel is identified as a foreground point within consecutive frames is counted. When the count reaches a fixed value, the pixel is marked as a background point.

[0019] The method of obtaining the difference between the target and the background by measuring the local brightness difference of the window includes the following steps:

[0020] For the target, a set of multi-scale windows is introduced, including multiple inner windows with the same center and corresponding outer windows;

[0021] By fixing the outer window and taking the neighborhood of the center pixel of different sizes as the inner window, the local brightness difference between each inner window and its adjacent outer window is calculated.

[0022] For a set of multi-scale windows, the maximum brightness difference is taken as the multi-scale local brightness difference at the window center point:

[0023]

[0024] Where E(x,y) represents the multi-scale local brightness difference, x,y represent pixels, k=1,2,...,K, and K is the number of windows; set Θ k The set Ω represents the pixels contained in the k-th inner window. k max I(s,t) and I(p,q) represent the pixels contained between the k-th inner window and the outer window, respectively. k and Ω k max The grayscale value of the pixel in NΘ k and NΩ k max They respectively represent those contained in set Θ k and Ω k max The number of pixels in the image;

[0025] When E(x,y) is greater than the threshold, the target within the inner window corresponding to the multi-scale local brightness difference is the bird target.

[0026] For an outer window with a size of (2k+1)×(2k+1), the number of inner windows is k-1.

[0027] When a bird target is detected, its representation in the image coordinate system is obtained (u i ,v i By controlling the camera's pitch and yaw angles, (u) i ,v i The target desired point, i.e. (u0, v0), is reached at the center of the image.

[0028] The control camera's pitch and yaw angles are such that (u i ,v i Reaching the target desired point (u0, v0) at the center of the image involves the following steps:

[0029] By using a control model based on pitch-yaw angles and bird target tracking, θ1 and θ2 are obtained to control the yaw and pitch angles of the camera; the control model is as follows:

[0030]

[0031]

[0032] Where: f is the camera focal length; L represents the distance between the bird target point and the camera on a plane perpendicular to the camera's imaging plane; Xi and Yi are the horizontal and vertical distances from the bird target point to the camera mounting point, respectively; θ1 and θ2 are the camera's rotation angles in the yaw and pitch directions, respectively; u l v l These are the pixel width and height of the image, respectively; u i v represents the x-coordinate of the bird target pixel in the image plane. i The vertical coordinate of the bird target pixel in the image plane.

[0033] A bird detection and tracking system based on infrared target recognition includes: a mechanical subsystem and a control subsystem and a sensing subsystem disposed on the mechanical subsystem;

[0034] The mechanical subsystem includes: a base mounting plate, a bird deterrent cannon base, and a bird deterrent cannon barrel; the bird deterrent cannon base is mounted on the base mounting plate.

[0035] The control subsystem includes: a main control box, a two-dimensional electric turntable, and a control igniter; the main control box is located on the base of the bird deterrent cannon, and the control igniter is located on the outer wall of the cannon barrel.

[0036] The sensing subsystem includes: an infrared camera; the infrared camera is mounted on the barrel of the bird deterrent cannon;

[0037] The control subsystem has a two-dimensional electric turntable mounted on the bird deterrent cannon base via a flange. The two-dimensional electric turntable is equipped with a bird deterrent cannon base mounting plate. The bird deterrent cannon barrel is connected to the bird deterrent cannon base mounting plate via a connector, so that the two-dimensional electric turntable can drive the bird deterrent cannon barrel to rotate in the circumferential and pitch directions, thereby controlling the infrared camera to capture bird images.

[0038] The main control box is used to detect and track birds based on bird images captured by infrared cameras. When a bird target is detected, the camera's pitch and yaw angles are controlled by a two-dimensional electric turntable to track the bird target; and the igniter is controlled to fire at the birds.

[0039] The normal to the imaging plane of the infrared camera is parallel to the central axis of the gun barrel.

[0040] The present invention has the following beneficial effects and advantages:

[0041] 1. This invention uses infrared technology for bird target identification, which has fast processing speed and high recognition accuracy. The algorithm uses infrared information to confirm animal targets, and can especially identify a variety of large and medium-sized birds such as magpies, falcons, and crows. The algorithm is also robust.

[0042] 2. This invention acquires images using an infrared camera, identifies bird targets using the proposed local brightness difference method, and verifies the continuity of bird flight and size stability based on multiple frames of images. After a series of processing steps, the expected effect of accurately identifying birds is achieved, facilitating subsequent directional tracking of birds.

[0043] 3. The infrared camera and the bird deterrent cannon are mounted coaxially, with a two-dimensional electric turntable below. Based on the detected bird target position, the control algorithm adjusts the pitch and yaw angles of the turntable, which can bring the bird target closer to the center of the image, thereby enabling the bird deterrent cannon to disperse the birds with maximum intensity.

[0044] 4. This invention has a simple structure, requiring only an infrared camera and a device with two-dimensional steering functionality, the specific requirements of which vary depending on the application. The device's position can be adjusted according to actual needs, making installation flexible and convenient. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the bird-repelling device in this invention;

[0046] The components include: 1. Base mounting plate; 2. Main control box; 3. Bird deterrent cannon base; 4. Flange; 5. Two-dimensional electric turntable; 6. Bird deterrent cannon barrel; 7. Control igniter; and 8. Infrared camera.

[0047] Figure 2 This is a flowchart of the bird deterrence device control process in this invention;

[0048] Figure 3 This is a schematic diagram of the target position in the pixel coordinate system during the bird-driving process in this invention;

[0049] Figure 4 This is a schematic diagram of directional tracking in the bird-repelling process of the present invention;

[0050] Figure 5 Schematic diagram of a target recognition method based on multi-scale local brightness difference;

[0051] Figure 6 Flowchart of a bird target localization algorithm based on multi-frame images;

[0052] Figure 7a This is a schematic diagram of single target recognition in this invention;

[0053] Figure 7b for Figure 7a A magnified view of a portion of the image;

[0054] Figure 8a This is a schematic diagram of multiple target recognition in this invention;

[0055] Figure 8b for Figure 8a A magnified view of a portion of the image. Detailed Implementation

[0056] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0057] This invention uses infrared images to identify and track birds in complex environments. Based on the detected bird location information and flight trajectory, it adjusts the attitude of a two-dimensional gimbal to track and directionally disperse the birds. This system is primarily used for bird dispersal in specific areas such as substations, fruit orchards, and airports. The system boasts a high bird recognition rate and stable dispersal effect; its targeted bird dispersal capabilities effectively reduce bird dispersal failures caused by bird adaptation.

[0058] (1) Device Design

[0059] Bird deterrent devices (such as Figure 1 It consists of a mechanical subsystem, a control subsystem, and a sensing subsystem.

[0060] The mechanical subsystem includes: a base mounting plate, a bird deterrent cannon base, and a bird deterrent cannon barrel;

[0061] The control subsystem includes: a main control box, a two-dimensional electric turntable, and a control igniter;

[0062] The sensing subsystem includes: an infrared camera;

[0063] The bird detection and tracking system based on infrared target recognition is characterized in that the base of the two-dimensional electric turntable in the control subsystem is mounted on the base of the bird deterrent cannon via a support column, and the two-dimensional electric turntable is provided with a bird deterrent cannon base mounting plate; the bird deterrent cannon base mounting plate is hinged to the armature on the outer wall of the cannon barrel via a connector, so that the two-dimensional electric turntable can drive the cannon barrel to rotate in the circumferential and pitch directions.

[0064] The bird detection and tracking system based on infrared target recognition acquires image information through an infrared camera and identifies and locates bird targets using a visual algorithm. The optical axis of the infrared camera is coaxial with the pointing direction of the bird deterrent cannon, and both are mounted on a two-dimensional motorized turntable. When a bird target appears in the infrared image, the system calculates the distance of the bird target from the image center and adjusts the two-dimensional motorized turntable using a control algorithm to bring the bird target closer to the image center origin, thus achieving the purpose of pointing the bird deterrent cannon at the bird target.

[0065] The two-dimensional electric turntable is hinged to the armature on the outer wall of the gun barrel, controlling the circumferential and pitch rotation of the gun barrel, thereby controlling the infrared camera to capture bird images.

[0066] The infrared camera is capable of identifying bird targets. The normal to the imaging plane of the infrared camera is parallel to the central axis of the gun barrel. By adjusting the elevation and yaw joint angles, bird tracking and targeted dispersion with maximum intensity can be achieved.

[0067] (2) Detection Algorithm

[0068] Bird flock capture falls under the category of small target capture under infrared imaging conditions on a moving platform. This presents challenges: firstly, it requires handling the false motion of the background image caused by platform movement and the true motion of the target; secondly, it necessitates addressing noise interference and inter-frame pixel brightness differences caused by non-uniform infrared imaging. To address these issues, the capture and recognition method employed in this project is divided into two stages.

[0069] The first stage is moving target detection with background subtraction. The platform needs to be stationary. The background of the project is usually a clean sky, so the first moving target entering the field of view can be regarded as a bird.

[0070] Infrared small target detection based on background subtraction treats distinguishing foreground targets from background as a binary classification problem. Each frame of the input video image sequence is compared with a background model; if pixel information matches the model, it is marked as a background point; otherwise, it is classified as a foreground point. Specifically, it consists of three steps: background modeling, foreground point segmentation, and background model updating.

[0071] Building the background model is the process of filling in the model sample values, and it is the first step of this algorithm. Since the background is not static, a single model sample cannot adapt to changes in the background; therefore, multiple samples are needed to describe the background. For any pixel x... i There is a background model:

[0072] B(x i )={B1(x i ), ..., B j (x i ), ..., B N (x i )}

[0073] There are N samples in total. (B) j (x i ) represents x i The j-th sample value is determined by the data in the first frame image. For the first two sample values, they are directly taken from the first frame image x. i Pixel value V(x) at i The following N-2 samples are represented by pixel values ​​V(x). i Add random noise within ±10.

[0074] Starting from the second frame, calculate x. iPixel value V(x) at i ) and x i Sample value B at point j (x i The distance difference is the absolute value between a pixel value and a sample value in a single-channel image (grayscale image). When the distance difference between a pixel and a sample is within a threshold range R, and the number of pixels meeting the threshold range standard is no less than # min If the pixel is not specified, it is classified as a background pixel; otherwise, it is classified as a foreground pixel. In the following formula, background is marked as 0, and foreground is marked as 1:

[0075] Since the background subtraction method performs image processing based on pixels, its results contain a large number of discrete erroneous foreground points. This method incorporates a foreground point counting module to correct the detection and segmentation results: counting the number of pixels x... i This method involves counting the number of times a point is identified as a foreground point within consecutive frames. When the count reaches a fixed value, the point is marked as a background point. This method is effective for detecting small moving targets in ground-to-air scenes. In ground-to-air scenes, most pixels in a frame image are background, and the probability of being repeatedly marked as foreground within a short period is very low. Therefore, pixels with counts exceeding a fixed value are mostly noise points or false detections caused by background clutter, and should be classified as background.

[0076] To ensure accurate foreground segmentation of the target in each frame, the background model must adapt to background changes and accurately describe the current background conditions. Therefore, model updates are crucial. This patent employs ViBe's random update strategy: for points identified as background, a model sample value is randomly replaced with the current pixel value, making the model effective over a longer time span without increasing the sample size; simultaneously, a background model for the neighboring pixels of the background point is randomly updated with a certain probability. This utilizes spatial transmission characteristics to diffuse the background, gradually eliminating erroneous foreground points.

[0077] The second stage involves the platform tracking the moving target after it is detected, which involves single-frame small target detection. To improve the detection rate, only images of the area surrounding the moving target (256*256 pixels) are selected for detection.

[0078] Single-frame small target detection mainly involves finding the difference criteria between the target and the background. Typically, the target area of ​​a flock of birds is significantly different from its surrounding neighborhood, and the target is a continuous, concentrated area of ​​brightness. Therefore, the key to detection is to find the difference between the target pixel and its neighboring pixels.

[0079] Calculating local brightness difference can effectively reveal the difference between the target and the background. Local brightness difference refers to the difference between the average value of pixels surrounding the center pixel and the average value of pixels in its outer perimeter. By using a fixed-size window, the average gray value of a small inner window around the center pixel is calculated, and then the average gray value of pixels in a larger outer window is calculated. The absolute value of the difference between the two is the local brightness difference.

[0080] Since the target size is uncertain, fixed inner and outer window sizes cannot meet practical applications, necessitating the introduction of multi-scale information. Based on the definition of a small target (occupying between 2x2 and 8x8 pixels), we fix the outer window and use the neighborhood of the center pixel of different sizes as the inner window. We then calculate the local brightness difference between the inner and outer windows, and use the largest brightness difference between them as the multi-scale local brightness difference at the center point. Figure 5 As shown.

[0081] The specific formula is as follows:

[0082]

[0083] Where k = 1, 2, ..., K. Set Θ k The set Ω represents the pixels contained within the central window (i.e., the inner window region). k max I(s,t) and I(p,q) represent the pixels contained between the k-th center window and the outer window, respectively. k and Ω k max The grayscale value of the pixel in NΘ k and NΩ k max They respectively represent those contained in set Θ k and Ω k max The number of pixels in the array is represented by K, where K is the number of windows. The maximum difference in grayscale between the inner and outer windows is taken at different central region scales. For an outer window size of (2k+1)x(2k+1), the number of inner window scales can be k-1.

[0084] Because power plants contain numerous heat-generating devices, judging brightness differences can mistakenly identify nodes of these devices as targets. Therefore, after obtaining brighter targets, isolated target identification is necessary. An adaptive thresholding method is used to binarize the range image and calculate the size of the connected component for each target. Since bird flocks generally appear as round dots, targets with larger connected component sizes can be excluded. The detection method flowchart is shown below. Figure 6 As shown.

[0085] The entire system framework (such as) Figure 1 As shown in the figure, the infrared camera and the bird deterrence device are coaxial and installed above the two-dimensional turntable. By adjusting the pitch angle θ1 and the yaw joint angle θ2, the tracking of birds and the directional deterrence with the maximum deterrence intensity can be achieved.

[0086] like Figure 2 As shown, the control system determines the target location (u) based on the bird's current position. i ,v i The difference e between the bird deterrent device and the ideal target position (u0, v0) is used to design a PID control rate based on the camera imaging model, and the pose of the bird deterrent device is adjusted accordingly, so as to achieve the purpose of the bird deterrent device pointing to the bird target.

[0087] When a bird target is detected in an image, its representation in the image coordinate system can be obtained through a detection algorithm. i ,v i To ensure maximum bird deterrence, the pitch and yaw joints should be controlled to reach the desired target point (u0, v0) in the image center. The relationship between the target position and the center position in the image plane is as follows (e.g., ...). Figure 3 As shown in the figure.

[0088] Based on the principles of camera imaging, a control model for pitch-yaw angle and bird target tracking can be established (e.g., Figure 4 As shown in the figure. Since the pitch and yaw motions of the two-dimensional gimbal are completely decoupled and correspond to U and V in the image respectively, a control model can be established according to the following formula.

[0089]

[0090] Similarly, we can find:

[0091]

[0092] in:

[0093] f—Camera focal length;

[0094] Xi, Yi — Horizontal and vertical distances from the bird target point to the camera mounting point;

[0095] θ1θ2 — the camera's angles in the yaw and pitch directions;

[0096] u l v l —The pixel width and height of the image;

[0097] u i —The horizontal coordinate of the image plane;

[0098] v i —The vertical coordinate of the image plane;

[0099] like Figure 7a , Figure 7b As shown, when detecting a single target, this algorithm can track and disperse birds based on the location of that single target.

[0100] like Figure 8a , Figure 8b As shown, this algorithm is also applicable to the detection of multiple targets. When a flock of birds appears, the positions of multiple birds can be detected at the same time. In order to achieve the best dispersal effect, the center position of the flock of birds, that is, the geometric center point of each bird target, can be selected for dispersal.

[0101] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, extensions, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A bird detection and tracking method based on infrared target recognition, characterized in that, It comprises the following steps: The video image sequence is subjected to background difference-based moving target detection to obtain the foreground and background of each frame of the video image sequence, and the foreground is the target; For the target, the local brightness difference of the window is used to obtain the difference between the target and the background to obtain the bird target; The background difference-based moving target detection on the collected image to obtain the foreground and background of each frame of the video image sequence comprises the following steps: Each frame of the input video image sequence is compared with the background model: if the pixel information of the image matches the pixel information of the background model, it is marked as a background point; otherwise, it is classified as a foreground point; The background model is expressed as follows: ； Where N represents the number of samples, express The j-th sample pixel information is used as the sample value; for the first and second sample values, they are directly used as the first frame image. pixel value at The last N-2 samples are in pixel value Add random noise within a set threshold range t to fill the noise. For each frame image starting from the second frame image, the pixel value at the current frame image is calculated the distance difference value of the sample value at in the background model; When the distance difference value is within a threshold range R, and the number of samples in the background model that meet the threshold range is not less than The pixel point is divided into a background pixel; otherwise, it is determined as a foreground; in the following formula, the background is marked as 0, and the foreground is marked as 1: ； wherein, represents the distance difference between the pixel point and the sample.

2. The bird detection and tracking method based on infrared target recognition according to claim 1, characterized in that, The pixel information is a gray value or an infrared image gray value.

3. The bird detection and tracking method based on infrared target recognition according to claim 1, characterized in that, For a sequence of video images, the number of times a pixel is counted as a foreground pixel When the count reaches a fixed value, the pixel is marked as a background pixel.

4. The bird detection and tracking method based on infrared target recognition according to claim 1, characterized in that, The local brightness difference of the window is used to obtain the difference between the target and the background, which comprises the following steps: For the target, a group of multi-scale windows are introduced, including a plurality of inner windows and corresponding outer windows with the same center; The local brightness difference between each inner window and the outer window adjacent thereto is calculated by fixing the outer window and taking different sizes of the center pixel neighborhood as the inner window; For a group of multi-scale windows, the maximum brightness difference is taken as the multi-scale local brightness difference of the window center point: ； wherein, represents a multi-scale local brightness difference, represents a pixel, k = 1, 2,..., K, K is the number of windows; set represents a pixel contained in the kth inner window, set represents a pixel contained between the kth inner window and the outer window, I(s, t) and I (p, q) respectively represent set and the gray value of the pixel point in and N respectively represent the number of pixel points contained in set and ​ When If the multi-scale local brightness difference is greater than the threshold value, the target in the inner window corresponding to the multi-scale local brightness difference is the bird target.

5. The bird detection and tracking method based on infrared target recognition according to claim 1, characterized in that, For the outer window size of (2k+1)×(2k+1), the number of inner windows is k-1.

6. The bird detection and tracking method based on infrared target recognition according to claim 1, characterized in that, When a bird target is detected, a representation of the bird target in the image coordinate system (u i ,v i ) is obtained, and by controlling the turning angles of the pitch and yaw directions of the camera, (u i ,v i ) reaches the target expected point of the image center, i.e. (u0,v0).

7. The bird detection and tracking method based on infrared target recognition according to claim 6, characterized in that, The control camera tilt and yaw direction of the angle, so that (u i ,v i ) to the image center target desired point, that is (u0, v0) It comprises the following steps: The pitch-yaw angle and the bird target tracking control model are used to obtain θ1 and θ2, which control the turning angles of the yaw and pitch directions of the camera; the control model is as follows: ...（1) ... (2) Wherein: f is the focal length of the camera; Xi, Yi respectively are the horizontal and vertical distances from the camera installation point to the bird target point; θ1, θ2 respectively are the turning angles of the camera in the yaw and pitch directions; u l v l are the pixel width and height of the image respectively; u i is the pixel horizontal coordinate of the bird target on the image plane, v i is the pixel vertical coordinate of the bird target on the image plane.

8. A bird detection and tracking system based on infrared target recognition, said system being configured to implement a method of bird detection and tracking based on infrared target recognition according to any one of claims 1 to 7, characterized in that, It comprises: A mechanical subsystem, a control subsystem and a perception subsystem arranged on the mechanical subsystem; The mechanical subsystem comprises a base mounting plate (1), a bird repelling gun base (3) and a bird repelling gun barrel (6); the bird repelling gun base (3) is arranged on the base mounting plate (1); The control subsystem comprises a main control box (2), a two-dimensional electric turntable (5) and a control igniter (7); the main control box (2) is arranged on the bird repelling gun base (3), and the control igniter (7) is arranged on the outer wall of the bird repelling gun barrel (6); The perception subsystem comprises an infrared camera (8); the infrared camera (8) is arranged on the bird repelling gun barrel (6); The two-dimensional electric turntable (5) in the control subsystem is installed on the bird repelling gun base (3) through a flange (4), and the bird repelling gun base mounting plate is arranged on the two-dimensional electric turntable (5); the bird repelling gun barrel (6) is connected with the bird repelling gun base mounting plate through a connecting piece, so that the two-dimensional electric turntable (5) drives the bird repelling gun barrel (6) to rotate in the circumferential and pitch directions, thereby controlling the infrared camera (8) to capture bird images; The main control box (2) is used for bird detection and tracking according to the bird images captured by the infrared camera (8); when a bird target is detected, the turning angles of the pitch and yaw directions of the camera are controlled through the two-dimensional electric turntable (5) to realize the tracking of the bird target; and the control igniter (7) is controlled to shoot the bird.

9. The bird detection and tracking system based on infrared target recognition according to claim 8, characterized in that, The imaging plane normal line of the infrared camera is parallel to the center axis of the barrel.

Images