Multi-mode line iron tower bird repelling system based on image analysis and motion perception

Abstract

The invention discloses a multi-mode line iron tower bird repelling system based on image analysis and motion perception, relates to the technical field of computer vision, and is used for solving the problems of bird flight trajectory tracking and accurate repelling in a complex field environment. Firstly, stable detection and continuous trajectory tracking of flying birds are realized by fusing bimodal visual data of visible light and thermal imaging, and the defect of poor environmental adaptability of a single sensor is overcome; then, motion vector change and thermal radiation characteristics are analyzed based on the track sequence, the flight intention and behavior mode of the birds are intelligently recognized, and the collision risk of the birds and iron tower equipment is predicted; according to different flight modes, different green light driver isolation schemes are generated, and conversion from general deterrence to precise intervention is achieved; finally, the repelling parameters are dynamically adjusted by monitoring the response of the birds to the CD driver, response data are fed back to the prediction module, a closed-loop control system with the self-learning ability is formed, and the continuity of the bird repelling effect and the environmental adaptability are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer vision, and particularly to a multi-modal line tower bird repelling system based on image analysis and motion perception. BACKGROUND

[0002] As the backbone network of power transmission network, the safe and stable operation of high-voltage transmission line is directly related to the lifeline of social economy. However, the overhead transmission line tower provides an ideal habitat and nesting site for birds, and the line faults caused by birds have become one of the main natural factors threatening the safety of the power grid. The excrement of birds on the tower can reduce the insulation performance of the insulator string, cause flashover discharge, and lead to line tripping; the nesting materials such as metal wires and plastics carried by birds may also cause inter-phase short circuit under certain conditions. These sudden failures not only cause equipment damage and economic losses, but also may lead to regional power supply interruption, affecting people's livelihood protection and industrial production. Therefore, developing efficient and reliable intelligent bird repelling means is an urgent need for the power industry to improve the line operation level and ensure the reliability of power supply.

[0003] The existing bird repelling methods mainly rely on sound and light stimulation devices, reflective devices or manual inspection. These methods have significant technical defects in actual operation: sound and light bird repelling devices usually work at a fixed frequency and direction, and cannot make targeted intervention according to the flight posture and motion characteristics of birds, resulting in limited coverage of the repelling range and easy adaptation of birds, thereby reducing the repelling effect; the reflective devices are strongly dependent on light, and are basically ineffective in cloudy, rainy, night or backlight environments; manual inspection method has the problems of insufficient timeliness and high risk. In addition, some automatic bird repelling systems based on a single camera can only respond after the birds approach the tower, and lack the ability of motion perception and trajectory prediction during flight, often causing false triggering, delayed repelling and other phenomena. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a multi-modal line tower bird repelling system based on image analysis and motion perception, which solves the problems in the above background art.

[0005] To achieve the above object, the application is implemented by the following technical solutions: a multi-modal line tower bird repelling system based on image analysis and motion perception, comprising the following modules: a multi-modal trajectory tracking module, configured to synchronously collect video data around the tower through a visible light camera and an infrared thermal imager, extract bird shape contour features and thermal radiation distribution features after image registration of the dual-path video stream, establish a dual-modal feature correspondence relationship and generate a bird trajectory position sequence; a flight intention prediction module, configured to calculate a displacement change between adjacent frames to obtain a motion vector according to the bird trajectory position sequence, identify flight trajectory curvature features according to a direction change rate of the motion velocity vector, and combine thermal radiation intensity changes to identify flight modes of straight-line approach, circling flight or random turning, and predict an intersection probability of the bird and the tower insulator region based on the motion vector and the flight mode; a dynamic light repelling strategy module, configured to generate a differentiated repelling scheme according to the flight mode when the intersection probability exceeds a set threshold: calculate a light spot projection position in advance based on the motion vector for the straight-line approach mode, determine fan-shaped light curtain parameters according to the trajectory curvature for the circling flight mode, and adjust light repelling intensity levels according to the thermal radiation intensity for the random turning mode; and a closed-loop light repelling control module, configured to execute the light repelling scheme and monitor bird responses through the visible light camera, calculate a heading angle deflection degree and a flight speed change rate, adjust a gimbal / optical tracking speed according to the heading angle deflection degree, adjust the light repelling intensity according to the flight speed change rate, and feed back the heading angle deflection degree and the flight speed change rate to the flight intention prediction module for updating flight mode identification rules.

[0006] Further, the specific process of synchronously collecting video data around the tower through the visible light camera and the infrared thermal imager, and extracting bird shape contour features and thermal radiation distribution features after image registration of the dual-path video stream is as follows: performing frame-by-frame image analysis on the visible light video stream, extracting bird shape contour features through edge detection and region segmentation algorithms, calibrating the infrared thermal imaging video stream of the same region, identifying the heat source region of the bird target based on temperature gradient changes, and extracting bird thermal radiation distribution features; achieving spatial registration through dual-modal image feature matching, and completing dual-modal frame alignment based on time stamp synchronization.

[0007] Further, the specific process of establishing a dual-modal feature correspondence relationship and generating a bird trajectory position sequence is as follows: establishing a correspondence relationship between visible light contour features and thermal radiation features according to the registered dual-modal features, predicting the potential position of the target in the next frame through Kalman filtering and performing cross-frame target association, recording the position coordinate data of the target in the continuous video frames for each successfully associated bird target to form an initial position sequence; performing trajectory smoothing processing on the initial position sequence through weighted average method to eliminate trajectory jitter caused by image noise and detection errors, verifying the integrity of the trajectory based on motion continuity constraints, and generating a bird trajectory position sequence containing time stamps and three-dimensional spatial coordinates.

[0008] Further, the specific process of calculating the intersection probability between the bird and the insulator region of the tower based on the motion vector and the flight mode is as follows: a virtual protection region is established around the insulator of the tower, the future flight path of the bird is deduced based on the direction and size of the current motion vector, and the shortest distance between the deduced path and the virtual protection region of the insulator is calculated; the deduced parameters are adjusted in combination with the flight mode: linear deduction is adopted for the straight-line approach mode, circular arc deduction is adopted for the hovering flight mode, and probability diffusion deduction is adopted for the random turning mode; the shortest spatial distance between the deduced trajectory and the boundary of the virtual protection region of the insulator of the tower is calculated, and the basic intersection probability of the inverse proportional relationship is established according to the distance value, the direction stability degree of the motion vector and the fluctuation amplitude of the thermal radiation intensity are combined to correct the basic intersection probability, wherein the direction stability is evaluated by the direction angle variance of the continuous motion vector, and the fluctuation amplitude of the thermal radiation intensity is calculated by the range in the sliding window, and the intersection probability value after dynamic correction is output.

[0009] Further, the specific process of calculating the intersection probability between the bird and the insulator region of the tower based on the motion vector and the flight mode is as follows: a virtual protection region is established around the insulator of the tower, the future flight path of the bird is deduced based on the direction and size of the current motion vector, and the shortest distance between the deduced path and the virtual protection region of the insulator is calculated; the deduced parameters are adjusted in combination with the flight mode: linear deduction is adopted for the straight-line approach mode, circular arc deduction is adopted for the hovering flight mode, and probability diffusion deduction is adopted for the random turning mode; the shortest spatial distance between the deduced trajectory and the boundary of the virtual protection region of the insulator of the tower is calculated, and the basic intersection probability of the inverse proportional relationship is established according to the distance value, the direction stability degree of the motion vector and the fluctuation amplitude of the thermal radiation intensity are combined to correct the basic intersection probability, wherein the direction stability is evaluated by the direction angle variance of the continuous motion vector, and the fluctuation amplitude of the thermal radiation intensity is calculated by the range in the sliding window, and the intersection probability value after dynamic correction is output.

[0010] Furthermore, the specific processes for calculating the pre-projection position of the light spot based on motion vectors in the straight-line approach mode, determining the parameters of the fan-shaped light curtain based on the trajectory curvature in the hovering flight mode, and adjusting the light drive intensity level based on thermal radiation intensity in the random turning mode are as follows: For the straight-line approach mode, the motion vector is decomposed to obtain the horizontal and vertical velocities. Combined with the system response time and green light propagation characteristics, the pre-projection position of the light spot in front of the bird's flight path is calculated to ensure that the green light spot accurately covers the expected flight trajectory within the effective driving distance. For the hovering flight mode, the center position and coverage area of ​​the fan-shaped light curtain are determined based on the trajectory curvature characteristics. The light curtain angle is dynamically adjusted according to the change in the hovering radius so that the fan-shaped light curtain can completely surround the key components of the tower and adapt to the changes in the bird's hovering trajectory. For the random turning mode, by analyzing the instantaneous change amplitude and frequency of thermal radiation intensity, when the change amplitude exceeds the set range and the change frequency is high, the high-intensity green light driving away mode is activated; when the change in thermal radiation intensity is gradual and stable within the normal range, the mode is switched to low-intensity energy-saving driving away mode.

[0011] Furthermore, the specific process of implementing the optical drive scheme and monitoring bird responses using a visible light camera to calculate the heading angle deflection and flight speed change rate, adjusting the gimbal / optical tracking speed based on the heading angle deflection, and adjusting the optical drive intensity based on the flight speed change rate is as follows: While implementing the optical drive scheme, a continuous sequence of bird flight videos is acquired using a visible light camera. By extracting the contour feature changes of the bird's main body in consecutive image frames, the change in the bird's heading angle between adjacent frames is calculated. Based on a preset heading angle change threshold range, the gimbal / optical tracking speed is divided into multiple adjustment levels. When the heading angle... When the change in heading angle is within the first preset range, a fast-moving gimbal / optical tracking strategy is adopted. When the change in heading angle is within the second preset range, a uniform-moving gimbal / optical following strategy is adopted. When the change in heading angle is within the third preset range, a slow and stable gimbal / optical locking strategy is adopted. The rate of change of bird flight speed between adjacent image frames is calculated synchronously. Based on the positive and negative characteristics and the magnitude of the rate of change in speed, a corresponding optical drive intensity adjustment mechanism is established. When the rate of change in speed is continuously negative, the optical drive intensity is gradually reduced. When the rate of change in speed is continuously positive, the optical drive intensity is gradually increased.

[0012] Furthermore, the specific process of feeding back the heading angle deflection and flight speed change rate to the flight intention prediction module to update the flight mode discrimination rules is as follows: Establish a mapping relationship between heading angle deflection data and flight mode discrimination rules; adjust the direction change rate discrimination threshold based on the distribution characteristics of the heading angle deflection; perform correlation analysis between the flight speed change rate and thermal radiation intensity change characteristics; update the speed change benchmark range under different flight modes based on the statistical characteristics of speed change rate recorded during historical expulsion processes; when the detection accuracy of flight mode recognition is consistently lower than the preset standard, update the motion vector direction change rate threshold and thermal radiation intensity fluctuation characteristic parameters in the flight mode discrimination rules.

[0013] The present invention has the following beneficial effects:

[0014] (1) The multimodal bird deterrence system for railway towers based on image analysis and motion perception achieves high-precision identification and dynamic behavior prediction of birds in flight through the collaboration of a multimodal trajectory tracking module and a flight intent prediction module. In the spatiotemporal registration process of visible light and infrared thermal imaging data, it can effectively distinguish between the sky background and flying animal targets, overcoming the problem of insufficient identification accuracy of single optical sensors in strong light, shadow and low illumination environments; by establishing the correspondence between dual-modal features and generating trajectory position sequences, the system can continuously track the bird's movement trajectory, capture subtle changes in attitude and displacement, and provide continuous input for subsequent intent analysis. The flight intent prediction module classifies flight patterns based on the rate of change of motion vector direction and the change of thermal radiation intensity, realizing real-time discrimination of bird approach, hovering or avoidance behavior, so that the potential intersection probability with key components can be calculated before the bird approaches the tower, significantly improving the advance amount and accuracy of the warning.

[0015] (2) A multimodal bird deterrence system for power transmission towers based on image analysis and motion perception is constructed by introducing a dynamic optical drive strategy module and a closed-loop optical drive control module, thereby establishing an adaptive and adjustable intelligent bird deterrence control mechanism. The system implements differentiated optical drive strategies for different flight modes based on prediction results, enabling dynamic optimization of the spot position, illumination angle, and power, ensuring that the deterrence beam always acts on the bird's visually sensitive area, reducing energy consumption and improving deterrence efficiency. The closed-loop optical drive control module determines the effectiveness of the deterrence behavior by calculating the bird's response parameters in real time, and automatically adjusts the gimbal / optical tracking speed and light intensity level, enabling the optical drive intervention process to have self-correction capabilities. The system transmits the response characteristics back to the flight intention prediction module to update the discrimination rules, realizing continuous learning and dynamic optimization of the bird deterrence strategy. This not only improves the targeting and response speed of bird deterrence behavior, but also effectively avoids excessive application of light interference, meeting the requirements of eco-friendliness and energy conservation, and is particularly suitable for long-term automated protection of high-voltage transmission lines in complex climates and multi-period operating environments.

[0016] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0017] Figure 1 This is a flowchart of the multimodal power line tower bird deterrence system based on image analysis and motion perception of the present invention. Detailed Implementation

[0018] This application's embodiment solves the problems of high risk of short circuit discharge caused by birds staying or nesting near power transmission towers, slow response of traditional sound and light bird deterrent devices, and unstable deterrent effect in the prior art through a multimodal power transmission tower bird deterrent system based on image analysis and motion perception.

[0019] The overall concept of the solution in this application embodiment is as follows:

[0020] First, the visible light video stream and infrared thermal imaging video stream around the tower are registered using a multimodal trajectory tracking module. By extracting contour features and fusing thermal radiation features, stable detection and trajectory tracking of bird targets are achieved, ensuring accurate identification of individual birds and their dynamic characteristics even under complex lighting and weather conditions.

[0021] Secondly, by calculating the motion vector between adjacent frames through the flight intention prediction module, extracting the trajectory curvature and thermal radiation change trends, and distinguishing different flight patterns of birds, it is possible to predict whether birds pose a potential risk of approaching critical parts of the tower.

[0022] Subsequently, the dynamic optical drive strategy module adaptively generates differentiated drive-away schemes based on the prediction results. It implements early light spot intervention for approaching trajectories, forms circumferential light curtain disturbances for hovering or loitering trajectories, and dynamically adjusts the optical drive intensity level through infrared thermal radiation feedback for uncertain flight modes, thereby improving the pertinence and real-time performance of the drive-away response.

[0023] Finally, the closed-loop optical drive control module monitors the bird's heading deflection and speed changes through visual feedback, adjusts the gimbal / optical tracking speed in real time according to the deflection angle, and sends the bird's dynamic response parameters back to the flight intention prediction module. This enables self-learning and optical drive strategy optimization based on visual behavior feedback, ensuring the stability and energy efficiency of the system in the long term.

[0024] Please see Figure 1This invention provides a technical solution: a multimodal bird deterrence system for power transmission towers based on image analysis and motion perception, comprising the following modules: a multimodal trajectory tracking module, used to simultaneously acquire video data around the tower using a visible light camera and an infrared thermal imager, extract bird morphological contour features and thermal radiation distribution features after image registration of the dual video streams, establish a dual-modal feature correspondence, and generate a bird trajectory position sequence; a flight intent prediction module, used to calculate the displacement change between adjacent frames based on the bird trajectory position sequence to obtain a motion vector, identify the curvature features of the flight trajectory based on the direction change rate of the motion velocity vector, and determine the flight mode of straight approach, circling flight, or random turning based on the change in thermal radiation intensity, and predict the flight intent based on the motion vector and flight mode. The system measures the probability of birds intersecting with the insulator area of ​​the iron tower; a dynamic optical drive strategy module generates differentiated deterrence schemes based on flight patterns when the probability of intersecting exceeds a set threshold: for straight approach patterns, it calculates the early projection position of the light spot based on motion vectors; for hovering flight patterns, it determines the parameters of the fan-shaped light curtain based on the trajectory curvature; and for random turning patterns, it adjusts the optical drive intensity level based on the thermal radiation intensity. A closed-loop optical drive control module executes the optical drive scheme and monitors bird responses through a visible light camera, calculates the heading angle deflection and flight speed change rate, adjusts the gimbal / optical tracking speed based on the heading angle deflection, adjusts the optical drive intensity based on the flight speed change rate, and feeds back the heading angle deflection and flight speed change rate to the flight intention prediction module to update the flight pattern discrimination rules.

[0025] In this implementation scheme, the visible light camera, infrared thermal imager, and optical drive mechanism adopt a compact layout with coaxial or near-axial parallelism to ensure that the optical axes of the three are highly consistent, enabling precise locking of the tracked target and real-time optical drive. The system further includes a remote data communication module for sending bird information identified by the system (such as species, number, direction, and time) and key data (photos, videos) during the drive-away process to the back-end management system for research. The system also includes a multi-system collaborative communication module for sharing data and coordinating drive-away strategies with other systems on the same or adjacent towers. A multimodal trajectory tracking module is used to simultaneously acquire video data around the tower using the visible light camera and infrared thermal imager. The visible light camera is used to capture the shape, outline, and flight trajectory of birds during the day or in well-lit conditions, while the infrared thermal imager is used to detect the thermal radiation characteristics of birds in low light, at night, or when obstructed. By performing image registration processing on the dual video streams, the data from different sensors are synchronized in the temporal dimension and correspond to the same physical scene location in the spatial dimension, thereby ensuring the accuracy of subsequent feature fusion. Spatiotemporal registration can be performed using a perspective transformation model based on feature point matching for image alignment. Subsequently, the system extracts the bird's morphological contour features and thermal radiation distribution features, establishes a bimodal feature correspondence, and generates a continuous sequence of bird trajectory positions to describe the bird's movement path and spatial position changes over time. This enables robust bird detection and tracking under different lighting and weather conditions, providing accurate dynamic input data for subsequent flight intent analysis. The flight intent prediction module calculates the displacement change between adjacent frames based on the aforementioned bird trajectory position sequence, thereby obtaining the motion vector. The motion vector is a two-dimensional vector representing the direction and speed of an object's movement in the image sequence, reflecting the bird's movement trend. By calculating the rate of change of the motion velocity vector in consecutive frames, the curvature features of the flight trajectory, i.e., the degree of curvature of the flight path, can be identified to distinguish between straight flight, circling flight, or random turning patterns. Simultaneously, by combining the trend of infrared thermal radiation intensity over time, it further determines whether the bird is in an active state of continuously approaching the tower. Based on the motion vector and flight pattern, the probability of the bird intersecting with the tower insulator area is predicted, thereby identifying potential collision or dwelling risks in advance. By combining image motion perception results with behavioral pattern analysis, the functionality is extended from "detection" to "prediction," improving the initiative and intelligence of bird deterrence response. The dynamic optical deterrence strategy module generates differentiated optical deterrence schemes based on flight patterns when the probability of encounter exceeds a set threshold.For birds approaching in a straight line, the system calculates the early projection position of the light spot based on the direction and velocity of the motion vector, creating a warning zone in front of the bird's flight path. For birds circling in a hovering flight pattern, the system determines the fan-shaped coverage angle and flashing frequency of the light curtain based on the trajectory curvature, forming circumferential interference. For birds randomly turning in a turning pattern, the system adjusts the output power or flashing rhythm of the optical drive device based on changes in thermal radiation intensity, achieving flexible bird deterrence. Adaptive optical interference strategy selection based on bird behavior patterns avoids the high energy consumption and poor adaptability problems caused by traditional fixed-mode bird deterrence. A closed-loop optical drive control module executes the above optical drive scheme and monitors the bird's response in real time using a visible light camera. The system extracts the bird's heading angle deflection and flight speed change rate. The gimbal / optical tracking speed is adjusted based on the heading angle deflection to ensure the interference beam continuously acts on the flight area in front of the bird; the optical drive intensity is dynamically adjusted based on the flight speed change rate to ensure significant interference without unnecessary energy waste. Simultaneously, the system feeds back the changing trends of heading angle deflection and flight speed rate of change to the flight intention prediction module, updating the flight mode discrimination rules in real time, thus forming a self-learning closed-loop control mechanism. This achieves dynamic matching between light-driven intervention and bird behavioral responses, improving the system's stability and deterrence accuracy in complex environments.

[0026] Specifically, the process of simultaneously acquiring video data around the tower using a visible light camera and an infrared thermal imager, and then performing image registration on the dual video streams to extract bird morphological contour features and thermal radiation distribution features is as follows: Frame-by-frame image analysis is performed on the visible light video stream; bird morphological contour features are extracted using edge detection and region segmentation algorithms; temperature calibration is performed on the infrared thermal imaging video stream of the same area; the heat source region of the bird target is identified based on temperature gradient changes, and the bird's thermal radiation distribution features are extracted; spatial registration is achieved through dual-modal image feature matching, and dual-modal frame alignment is completed synchronously based on timestamps.

[0027] In this implementation scheme, frame-by-frame image analysis is performed on the visible light video stream, and bird morphological contour features are extracted using edge detection and region segmentation algorithms. Edge detection employs the Canny algorithm based on gradient operators, which detects regions of abrupt gray-level changes between the bird target and the background by calculating the magnitude and direction of the image's gray-level gradient. Based on this, a region growing algorithm is used to merge adjacent pixels into the same target region according to a gray-level similarity threshold, extracting a closed and coherent bird contour. Temperature calibration is performed on the infrared thermal imaging video stream of the same region, using a standard temperature source to linearly correct the infrared sensor response to compensate for the influence of environmental temperature drift. Subsequently, heat source regions of the bird target are identified based on temperature gradient changes; that is, by calculating the local temperature change rate, regions exceeding a set threshold are identified as potential bird heat source regions. Further extraction of the bird's thermal radiation distribution features, including maximum heat intensity, average heat intensity, and heat distribution uniformity, is used for subsequent dual-modal feature fusion. Spatial registration is achieved through dual-modal image feature matching. Specifically, feature point sets from visible light and infrared images are selected respectively, and feature point descriptors are extracted using the Scale Invariant Feature Transform (SIFT) algorithm. Corresponding point pairs are obtained based on the Euclidean distance minimum matching principle. The affine transformation matrix is ​​solved using the least squares method to achieve spatial alignment of the two video streams. Finally, inter-frame time matching is performed based on timestamp synchronization to ensure that the visible light and infrared images at the same moment are consistent in time and space, providing high-precision input data for subsequent trajectory generation and behavior analysis. During dual-modal feature fusion, the system dynamically adjusts the weight ratio of visible light and infrared data in feature extraction and subsequent tracking based on the background brightness collected by the ambient light sensor (to distinguish day / night) and the ambient temperature collected by the ambient temperature sensor (to distinguish winter / summer).

[0028] Specifically, the process of establishing the correspondence between bimodal features and generating a bird trajectory position sequence is as follows: Based on the registered bimodal features, a correspondence between visible light contour features and thermal radiation features is established. The potential position of the target in the next frame is predicted by Kalman filtering and cross-frame target association is performed. For each successfully associated bird target, the position coordinate data of the target in consecutive video frames is recorded to form an initial position sequence. The initial position sequence is smoothed by weighted averaging to eliminate trajectory jitter caused by image noise and detection errors. The integrity of the trajectory is verified based on motion continuity constraints, and a bird trajectory position sequence containing timestamps and three-dimensional spatial coordinates is generated.

[0029] In this implementation scheme, a correspondence between visible light contour features and thermal radiation features is established based on the registered dual-state features. Specifically, for each candidate bird target in each frame of the image, its morphological feature vector is extracted. ;in, Represents the area of ​​the outline; Indicates the length of the spindle; This represents the boundary complexity. Simultaneously, its infrared thermal radiation feature vector is extracted. ;in, Indicates the maximum thermal radiation intensity; Indicates the average thermal radiation intensity; This represents the variance of thermal intensity. The correspondence between bistate features is established using a feature similarity function: ;in, , The weighting coefficients are determined through Bayesian optimization using the feature matching accuracy in the historical sample dataset, and are used to balance the contribution ratios of geometric features and thermal radiation features. The pairings with the highest similarity function values ​​are identified as multimodal correspondence instances of the same bird target. After establishing the correspondence, the system predicts the target's potential location in the next frame using a Kalman filter algorithm. Let the target's state vector in the current frame be... ;in, , This indicates the spatial coordinates of the bird in the current frame; , This represents the velocity components in the horizontal and vertical directions. The state prediction equation is: ;in, This is the state transition matrix; This represents process noise. The observation equation is: ;in, The observation matrix; To mitigate observation noise, the optimal estimation of the target's motion state is achieved by minimizing the trace of the prediction error covariance matrix, thereby obtaining the target's predicted position in consecutive frames. For each successfully associated bird target, its position coordinates in consecutive video frames are recorded to form an initial position sequence. To reduce trajectory jitter caused by image noise and detection errors, a weighted averaging method is used to smooth the initial trajectory. ;in, These are the smoothed trajectory points; These are the trajectory points of historical moments; The weighting coefficients are adaptively adjusted based on the trajectory continuity index. When motion changes drastically, the weight of historical frames is reduced to improve responsiveness. Furthermore, trajectory integrity is verified based on motion continuity constraints; if the position difference between adjacent frames exceeds a set threshold... If no time interruption occurs, the trajectory is deemed valid. Finally, a sequence of bird trajectory positions containing timestamps and three-dimensional spatial coordinates is generated, providing continuous, jitter-free motion input data for the flight intention prediction module.

[0030] Specifically, the process of calculating the displacement change between adjacent frames based on the bird trajectory position sequence to obtain the motion vector, identifying the flight trajectory curvature characteristics based on the direction change rate of the motion velocity vector, and combining this with changes in thermal radiation intensity to determine the flight mode of straight approach, hovering flight, or random turning is as follows: Extract the three-dimensional spatial coordinates of multiple consecutive frames from the bird trajectory position sequence, calculate the difference between the position coordinates of adjacent frames to obtain the displacement vector, obtain the motion vector containing the velocity magnitude and direction through vector synthesis, analyze the changes in the direction angle of multiple consecutive motion vectors, and calculate the change amplitude of the direction angle per unit time as the trajectory curvature feature value; simultaneously monitor the fluctuation of thermal radiation intensity, calculate the standard deviation and change frequency of thermal radiation intensity, and combine the trajectory curvature feature value and thermal radiation intensity change characteristics for comprehensive judgment: when the trajectory curvature feature value is less than a set threshold and the thermal radiation intensity changes smoothly, it is judged as a straight approach mode; when the trajectory curvature feature value changes periodically and the thermal radiation intensity fluctuates synchronously and regularly, it is judged as a hovering flight mode; when the trajectory curvature feature value changes irregularly and the thermal radiation intensity fluctuates violently, it is judged as a random turning mode.

[0031] In this implementation scheme, three-dimensional spatial coordinates of multiple consecutive frames are extracted from the bird trajectory location sequence. Calculate the displacement vector between adjacent frames: ;in, Indicates at time The displacement vector; These are three-dimensional spatial coordinate components. Based on the sampling time interval between adjacent frames. The motion vector can be obtained as follows: ;in, Indicates at time The velocity vector contains information about both the magnitude and direction of the velocity; its direction angle... Calculated from the horizontal component of the vector: To analyze the curvature characteristics of the flight trajectory, the changes in the direction angle of multiple consecutive motion vectors are analyzed, and the rate of change of direction angle per unit time is calculated: ;in, Indicates time The trajectory curvature characteristic value is used to characterize the turning amplitude of the flight path. A smaller curvature value indicates a more stable flight direction; periodic fluctuations in the curvature value indicate the presence of regular circling characteristics. Simultaneously, the thermal radiation intensity sequence in the infrared thermal imaging data... Perform fluctuation analysis and calculate the standard deviation and frequency of change within the sliding window: ;in, This indicates the amplitude of fluctuations in thermal radiation intensity; This represents the frequency of change in thermal radiation intensity; HE is the average thermal intensity within the window. This is a threshold for minute changes, used to suppress the influence of sensor noise; its value is obtained through empirical statistics. The system's comprehensive trajectory curvature characteristic value. With thermal radiation fluctuation parameters Flight mode is determined according to the following rules: When and When, it is determined to be a linear approach pattern; when It changes periodically and When the synchronization exhibits periodic fluctuations, it is determined to be a hovering flight mode; when... Irregular jumps and When this occurs, it is determined to be a random turning mode. Among these, the threshold... The system acquires data through adaptive learning on sample data under different climatic and background conditions, balancing discrimination accuracy and environmental robustness. This process, by combining spatial motion geometry and infrared thermal radiation dynamics, enables the system to accurately identify bird flight intentions at the multimodal perception level, thereby predicting potential approach risks in advance and providing a high-confidence input basis for subsequent bird deterrence decisions.

[0032] Specifically, the process of predicting the probability of birds intersecting with the tower insulator area based on motion vectors and flight patterns is as follows: A virtual protection zone is established around the tower insulator. The future flight path of the birds is inferred based on the direction and magnitude of the current motion vector, and the shortest distance between the inferred path and the virtual protection zone is calculated. The inference parameters are adjusted in combination with the flight pattern: linear inference is used for the straight approach mode, circular inference is used for the circling flight mode, and probability diffusion inference is used for the random turning mode. The shortest spatial distance between the inferred trajectory and the boundary of the virtual protection zone of the tower insulator is calculated. A basic intersection probability with an inverse proportional relationship is established based on the distance value. The basic intersection probability is weighted and corrected by combining the directional stability of the motion vector and the fluctuation range of the thermal radiation intensity. The directional stability is evaluated by the variance of the direction angle of the continuous motion vector, and the fluctuation range of the thermal radiation intensity is calculated by the range within the sliding window. The dynamically corrected intersection probability value is output.

[0033] In this implementation plan, a virtual protective zone is established around the insulators of the transmission tower to simulate the critical protection space of the power equipment. Let the current location of the bird be... The current motion vector is Then in the prediction time window The flight path within can be represented as: ;in, ; This is a path offset correction term used to describe the motion differences under different flight modes: for the straight-line approach mode, Linear extrapolation was used; for the hovering flight mode, For the arc offset term, it is defined as follows: ,in The radius of the trajectory, For angular velocity; for random steering mode, The term is a Gaussian diffusion term, which follows the rules of Gaussian diffusion. ,in Let be the variance matrix, representing the uncertainty distribution of the trajectory. This is achieved by extrapolating the trajectory. Boundary surface of virtual protection zone of tower insulator Perform spatial distance calculations to obtain the shortest distance: The basic intersection probability is established based on an inverse proportional relationship according to the distance value: ;in, The proportionality coefficient, determined by fitting proximity events from field sample data, is used to control the sensitivity of the intersection probability. The basic intersection probability is further weighted and corrected by combining the degree of directional stability and the amplitude of thermal radiation fluctuations. The directional stability metric is defined as the variance of the direction angle of the continuous motion vector: The amplitude parameter of thermal radiation fluctuation is defined as the range within the sliding window: The final intersection probability after comprehensive correction is: ;in, , These are the directional stability correction coefficient and the thermal radiation fluctuation correction coefficient, determined through multi-objective optimization to balance overfitting and robustness. By introducing multi-mode path modeling and multi-source feature joint correction in the motion prediction layer, the system can accurately quantify the potential intersection risk between birds and tower insulators in non-deterministic flight environments, achieving an intelligent transition from "visual detection" to "motion prediction," and providing real-time, quantifiable risk input for the dynamic optical drive strategy module. When multiple birds are present within the field of view, the system will calculate their respective intersection probabilities and, combined with their flight speed and size, comprehensively determine the threat level, prioritizing the driving away of the bird target with the highest threat level.

[0034] Specifically, the process for calculating the early projection position of the light spot based on motion vectors in the straight-line approach mode, determining the parameters of the fan-shaped light curtain based on the trajectory curvature in the hovering flight mode, and adjusting the light drive intensity level based on thermal radiation intensity in the random turning mode is as follows: For the straight-line approach mode, the motion vector is decomposed to obtain the horizontal and vertical velocities. Combining the system response time and green light propagation characteristics, the early projection position of the light spot in front of the bird's flight path is calculated to ensure that the green light spot accurately covers the expected flight trajectory within the effective driving distance. For the hovering flight mode, the center position and coverage area of ​​the fan-shaped light curtain are determined based on the trajectory curvature characteristics. The light curtain angle is dynamically adjusted according to the change in the hovering radius to ensure that the fan-shaped light curtain can completely surround the key components of the tower and adapt to the changes in the bird's hovering trajectory. For the random turning mode, by analyzing the instantaneous change amplitude and frequency of thermal radiation intensity, when the change amplitude exceeds the set range and the change frequency is high, the high-intensity green light driving away mode is activated; when the change in thermal radiation intensity is gradual and stable within the normal range, the mode is switched to low-intensity energy-saving driving away mode.

[0035] In this implementation scheme, for the linear approach mode, the specific process of calculating the early projection position of the light spot based on the motion vector is as follows: the motion vector is decomposed into a horizontal component and a vertical component, denoted as , ... and Combined with the system response time constant and green light propagation delay coefficient Calculate the position of the light spot in advance Specifically, it can be expressed as: ;in, The coordinates of the bird's current location; The horizontal component of the bird's velocity is used to reflect the bird's displacement trend along the ground. The vertical component of velocity is used to describe its vertical rise and fall. This is the system's response time from detecting the CD-ROM drive startup, used to ensure that the position of the light spot dynamically matches the bird's trajectory; The light propagation delay coefficient is used to compensate for the time difference between the emission of green light and its coverage of the target. Through the above calculations, a light spot projection point can be generated in advance ahead of the bird's flight path, ensuring the beam accurately covers the bird's direction of motion before it enters the effective dispersal distance, thereby improving the real-time performance and effectiveness of the dispersal response. For the circling flight mode, the specific process for determining the fan-shaped light curtain parameters based on the trajectory curvature is as follows: based on the trajectory curvature characteristic value... Calculate the turning radius And based on this, determine the center point of the fan-shaped light curtain. and Zhang Jiao Among them, Zhang Jiao The calculation formula is: ;in, The lateral dimensions of key components of the tower are used to determine the coverage area of ​​the light curtain; The radius of the circling trajectory determines the spatial range within which birds can fly. This is achieved through real-time updates. The system can dynamically adjust the coverage angle and position of the fan-shaped light curtain, ensuring that the light curtain always completely surrounds the critical area of ​​the tower and adapts to changes in bird circling trajectories, thereby achieving continuous, low-energy-consumption regional protection. For the random turning mode, the specific process of adjusting the light drive intensity level based on thermal radiation intensity is as follows: by calculating the thermal radiation intensity sequence... instantaneous change amplitude and frequency of change ,like Exceeding the set range threshold and If so, a high-intensity expulsion mode will be activated; if and Then it switches to a low-intensity energy-saving mode. Among them, Determined based on the average fluctuation range of historical monitoring data, taking... ,in This represents the average intensity of thermal radiation. Its standard deviation. This strategy allows for the matching and adjustment of the repelling light intensity with the degree of bird disturbance. The optical drive light source has continuously or stepped adjustable brightness, adjustable light pulse width, and the maximum brightness can be set to a continuous high-brightness mode. The optical drive intensity is automatically adjusted to the baseline based on the background brightness.

[0036] Specifically, the process of implementing the optical drive scheme and monitoring bird responses using a visible light camera to calculate the heading angle deflection and flight speed change rate, adjusting the gimbal / optical tracking speed based on the heading angle deflection, and adjusting the optical drive intensity based on the flight speed change rate is as follows: While implementing the optical drive scheme, a continuous sequence of bird flight videos is acquired using a visible light camera. By extracting the contour feature changes of the bird's main body in consecutive image frames, the change in the bird's heading angle between adjacent frames is calculated. Based on a preset heading angle change threshold range, the gimbal / optical tracking speed is divided into multiple adjustment levels. When the heading angle changes... When the change in heading angle is within the first preset range, a fast-moving gimbal / optical tracking strategy is adopted. When the change in heading angle is within the second preset range, a uniform-moving gimbal / optical following strategy is adopted. When the change in heading angle is within the third preset range, a slow and stable gimbal / optical locking strategy is adopted. The rate of change of bird flight speed between adjacent image frames is calculated synchronously. Based on the positive and negative characteristics and the magnitude of the change in speed, a corresponding optical drive intensity adjustment mechanism is established. When the rate of change in speed is continuously negative, the optical drive intensity is gradually reduced. When the rate of change in speed is continuously positive, the optical drive intensity is gradually increased.

[0037] In this implementation scheme, while executing the optical drive strategy, the system continuously acquires video sequences of bird flight using a visible light camera, performs angular analysis on the directional changes of the bird's body outline in consecutive frames, and obtains the change in heading angle. .when Within the first threshold range At that time, the gimbal / optical tracking speed is set to fast tracking mode; when In the second interval When, a constant speed following mode is used; when At this time, a slow locking mode is used. The above threshold... The system uses empirical statistics on bird turn angle distribution to distinguish three typical behaviors: slight disturbance, normal avoidance, and sharp turn. Simultaneously, the system calculates the rate of change in bird flight speed between adjacent frames. And establish a CD-ROM drive intensity adjustment function: ;in, This is the initial light intensity value; This is the rate of change adjustment coefficient, calibrated in dispersal experiments based on different bird sizes and reaction sensitivities. When When the value remains negative (indicating deceleration or turning around), the light intensity is gradually reduced; when... When the light intensity remains positive (indicating acceleration or approach), it gradually increases. The gimbal / optical tracking speed is divided into multiple adjustment levels: the first preset range indicates a large change in heading angle, the second preset range indicates a medium change in heading angle, and the third preset range indicates a small change in heading angle. When the change in heading angle is within the first preset range, a fast-moving gimbal / optical tracking strategy is adopted; when the change in heading angle is within the second preset range, a uniform-speed gimbal / optical following strategy is adopted; and when the change in heading angle is within the third preset range, a slow and stable gimbal / optical locking strategy is adopted. This closed-loop adjustment process creates real-time feedback between the optical drive behavior and the bird's dynamic response, improving the overall deterrence efficiency.

[0038] Specifically, the process of feeding back the heading angle deflection and flight speed change rate to the flight intention prediction module to update the flight mode discrimination rules is as follows: Establish a mapping relationship between heading angle deflection data and flight mode discrimination rules; adjust the direction change rate discrimination threshold based on the distribution characteristics of the heading angle deflection; perform correlation analysis between the flight speed change rate and thermal radiation intensity change characteristics; update the speed change benchmark range for different flight modes based on the statistical characteristics of speed change rate recorded during historical clearance processes; when the detection accuracy of flight mode identification is consistently lower than the preset standard, update the motion vector direction change rate threshold and thermal radiation intensity fluctuation characteristic parameters in the flight mode discrimination rules.

[0039] In this implementation plan, the specific process of feeding back the heading angle deflection and the rate of change of flight speed to the flight intention prediction module to update the flight mode discrimination rules is as follows: Establishing the heading angle deflection distribution function The mapping relationship with flight mode, when the system detects a straight line approaching the sample The variance exceeds the historical mean threshold. At the same time, the threshold for the rate of change of direction is automatically increased to prevent false judgments; simultaneously, the rate of change of velocity is... With thermal radiation intensity fluctuation parameters Perform correlation analysis and calculate the Pearson correlation coefficient. ,like The significant decrease indicates a weakening of the stability of the current pattern recognition, prompting the system to adjust the flight pattern classification weights accordingly. When the recognition accuracy rate is lower than the preset standard for multiple consecutive periods. At that time, the rule adaptive update mechanism is automatically triggered to re-optimize the combination range of the motion vector direction change rate threshold and the thermal radiation intensity fluctuation parameter, so that the flight mode discrimination model can maintain long-term stability and self-learning ability.

[0040] In summary, this application has at least the following effects:

[0041] The multimodal bird deterrence system for power transmission towers, based on image analysis and motion perception, achieves all-weather, all-time detection and identification of bird targets by introducing visible light and infrared dual-modal visual fusion, significantly improving detection accuracy in complex environments. Through an intent prediction model based on motion vectors and flight curvature characteristics, it can predict potential collision risks between birds and key tower components in advance, demonstrating strong foresight and intelligent decision-making capabilities. By combining flight mode-adaptive optical deterrence strategies with real-time feedback closed-loop control, it achieves dynamic matching between optical deterrence intensity and bird behavior responses, ensuring deterrence effectiveness while reducing energy consumption and false triggering rates. The system as a whole possesses the comprehensive advantages of multimodal fusion perception, intelligent prediction and discrimination, and closed-loop self-learning control, effectively improving the safety protection level and intelligent operation and maintenance of power transmission towers.

[0042] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0043] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0044] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0045] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0046] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0047] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A multimodal bird-repelling system for power transmission towers based on image analysis and motion perception, characterized in that, Includes the following modules: The multimodal trajectory tracking module is used to simultaneously acquire video data around the tower using a visible light camera and an infrared thermal imager. After image registration of the two video streams, it extracts bird morphological contour features and thermal radiation distribution features, establishes the correspondence between the two modal features, and generates a sequence of bird trajectory positions. The flight intent prediction module is used to calculate the displacement change between adjacent frames based on the bird trajectory position sequence to obtain the motion vector, identify the curvature characteristics of the flight trajectory based on the direction change rate of the motion velocity vector, and determine the flight mode of straight approach, circling flight or random turning based on the change of thermal radiation intensity. Based on the motion vector and flight mode, the probability of the bird intersecting with the tower insulator area is predicted. The dynamic optical drive strategy module is used to generate differentiated driving away schemes based on flight mode when the intersection probability exceeds a set threshold: for the straight approach mode, the light spot projection position is calculated based on the motion vector; for the hovering flight mode, the fan-shaped light curtain parameters are determined based on the trajectory curvature; and for the random turning mode, the optical drive intensity level is adjusted based on the thermal radiation intensity. The closed-loop optical drive control module is used to execute the optical drive scheme and monitor bird responses through a visible light camera. It calculates the heading angle deflection and the rate of change of flight speed, adjusts the gimbal / optical tracking speed according to the heading angle deflection, adjusts the optical drive intensity according to the rate of change of flight speed, and feeds back the heading angle deflection and the rate of change of flight speed to the flight intention prediction module to update the flight mode discrimination rules.

2. The multimodal power line tower bird deterrent system based on image analysis and motion perception according to claim 1, characterized in that: The specific process of extracting bird morphological contour features and thermal radiation distribution features by simultaneously acquiring video data around the iron tower using a visible light camera and an infrared thermal imager, and then performing image registration on the dual video streams, is as follows: Frame-by-frame image analysis of visible light video streams is performed, and bird morphological contour features are extracted through edge detection and region segmentation algorithms. Temperature calibration is performed on infrared thermal imaging video streams of the same region, and heat source regions of bird targets are identified based on temperature gradient changes, and bird thermal radiation distribution features are extracted. Spatial registration is achieved through bimodal image feature matching, and bimodal frame alignment is completed synchronously based on timestamps.

3. The multimodal power line tower bird deterrent system based on image analysis and motion perception according to claim 2, characterized in that: The specific process of establishing the correspondence between bimodal features and generating bird trajectory location sequences is as follows: Based on the registered dual-modal features, a correspondence between visible light contour features and thermal radiation features is established. The potential position of the target in the next frame is predicted by Kalman filtering and cross-frame target association is performed. For each successfully associated bird target, the position coordinate data of the target in consecutive video frames is recorded to form an initial position sequence. The initial position sequence is smoothed using a weighted average method, and the integrity of the trajectory is verified based on motion continuity constraints, generating a bird trajectory position sequence containing timestamps and three-dimensional spatial coordinates.

4. The multimodal power line tower bird deterrent system based on image analysis and motion perception according to claim 1, characterized in that: The specific process of calculating the displacement change between adjacent frames based on the bird trajectory position sequence to obtain the motion vector, identifying the curvature characteristics of the flight trajectory based on the rate of change of the motion velocity vector, and combining this with changes in thermal radiation intensity to determine the flight mode of straight approach, circling flight, or random turning is as follows: The three-dimensional spatial coordinates of multiple consecutive frames are extracted from the bird trajectory position sequence. The difference between the position coordinates of adjacent frames is calculated to obtain the displacement vector. The motion vector containing the velocity magnitude and direction is obtained through the vector synthesis method. The change of the direction angle of multiple consecutive motion vectors is analyzed, and the change amplitude of the direction angle per unit time is calculated as the trajectory curvature feature value. The system synchronously monitors fluctuations in thermal radiation intensity, calculates the standard deviation and frequency of change of thermal radiation intensity, and makes a comprehensive judgment based on the trajectory curvature characteristic value and the thermal radiation intensity change characteristics: when the trajectory curvature characteristic value is less than a set threshold and the thermal radiation intensity changes steadily, it is judged as a straight-line approach mode; when the trajectory curvature characteristic value changes periodically and the thermal radiation intensity fluctuates synchronously and regularly, it is judged as a hovering flight mode; when the trajectory curvature characteristic value changes irregularly and the thermal radiation intensity fluctuates violently, it is judged as a random turning mode.

5. The multimodal power line tower bird deterrent system based on image analysis and motion perception according to claim 4, characterized in that: The specific process of predicting the probability of birds intersecting with the insulator area of ​​the iron tower based on motion vectors and flight patterns is as follows: Establish a virtual protection zone, predict the future flight path of birds based on the direction and magnitude of the current motion vector, and calculate the shortest distance between the predicted path and the virtual protection zone; Adjust the simulation parameters according to the flight mode: use linear simulation for the straight approach mode, use circular simulation for the hovering flight mode, and use probability diffusion simulation for the random turning mode; The shortest spatial distance between the simulated trajectory and the boundary of the virtual protection zone of the tower insulator is calculated. A basic intersection probability with an inverse relationship is established based on the distance value. The basic intersection probability is weighted and corrected by combining the directional stability of the motion vector and the fluctuation range of the thermal radiation intensity. The directional stability is evaluated by the variance of the direction angle of the continuous motion vector, and the fluctuation range of the thermal radiation intensity is calculated by the range within the sliding window. The intersection probability value after dynamic correction is output.

6. The multimodal power line tower bird deterrent system based on image analysis and motion perception according to claim 1, characterized in that: The specific processes for calculating the early projection position of the light spot based on the motion vector in the straight-line approach mode, determining the parameters of the fan-shaped light curtain based on the trajectory curvature in the hovering flight mode, and adjusting the light drive intensity level based on the thermal radiation intensity in the random turning mode are as follows: For the straight-line approach mode, the motion vector is decomposed to obtain the horizontal and vertical velocities. Combined with the system response time and green light propagation characteristics, the advance projection position of the light spot in front of the bird's flight path is calculated. For the hovering flight mode, the center position and coverage of the fan-shaped light curtain are determined based on the trajectory curvature characteristics, and the angle of the light curtain is dynamically adjusted according to the change of the hovering radius. For the random steering mode, by analyzing the instantaneous change amplitude and frequency of thermal radiation intensity, when the change amplitude exceeds the set range and the change frequency is high, the high-intensity green light drive-away mode is activated; when the change of thermal radiation intensity is gradual and stable within the normal range, the mode is switched to low-intensity energy-saving drive-away mode.

7. The multimodal power line tower bird deterrent system based on image analysis and motion perception according to claim 1, characterized in that: The specific process of implementing the optical drive scheme and monitoring bird responses using a visible light camera, calculating the heading angle deflection and flight speed change rate, adjusting the gimbal / optical tracking speed based on the heading angle deflection, and adjusting the optical drive intensity based on the flight speed change rate is as follows: While implementing the optical drive scheme, a visible light camera continuously acquires video sequences of bird flight. By extracting the contour feature changes of the bird's main body in consecutive image frames, the change in the bird's heading angle between adjacent frames is calculated. Based on the preset heading angle change threshold range, the gimbal / optical tracking speed is divided into multiple adjustment levels. When the heading angle change is within the first preset range, a fast-moving gimbal / optical tracking strategy is adopted. When the heading angle change is within the second preset range, a uniform-moving gimbal / optical following strategy is adopted. When the heading angle change is within the third preset range, a slow and stable gimbal / optical locking strategy is adopted. The rate of change of bird flight speed between adjacent image frames is calculated synchronously. Based on the positive and negative characteristics and the magnitude of the rate of change of speed, a corresponding optical drive intensity adjustment mechanism is established. When the rate of change of speed is continuously negative, the optical drive intensity is gradually reduced; when the rate of change of speed is continuously positive, the optical drive intensity is gradually increased.

8. The multimodal power line tower bird deterrent system based on image analysis and motion perception according to claim 7, characterized in that: The specific process of feeding back the heading angle deflection and the rate of change of flight speed to the flight intention prediction module to update the flight mode discrimination rules is as follows: Establish a mapping relationship between heading angle deflection data and flight mode discrimination rules, and adjust the direction change rate discrimination threshold according to the distribution characteristics of heading angle deflection. The correlation analysis between the rate of change of flight speed and the characteristics of change of thermal radiation intensity was conducted, and the benchmark range of speed change under different flight modes was updated based on the statistical characteristics of the rate of change of speed recorded in the historical expulsion process. When the accuracy of flight mode recognition is consistently lower than the preset standard, the threshold for the rate of change of motion vector direction and the characteristic parameters of thermal radiation intensity fluctuation in the flight mode discrimination rule are updated.

Images