Substation bird three-dimensional positioning method and system based on double thermal infrared imagers

Abstract

The invention relates to a transformer substation bird three-dimensional positioning method and system based on double thermal infrared imagers, and belongs to the technical field of transformer substation safety monitoring. The system comprises a data acquisition module, a background heat source filtering module, a three-dimensional coordinate fusion module, a deep learning recognition module, a multi-modal meteorological compensation module, a trajectory prediction and early warning module and a data output and linkage control module. The method comprises the following steps: acquiring data cooperatively through double thermal infrared imagers, accurately identifying a bird target by using a deep learning model after background filtering and temperature threshold preliminary screening, calculating a three-dimensional space position of the bird target through coordinate fusion, and dynamically optimizing detection parameters in combination with real-time meteorological data. And finally, an improved Kalman filtering algorithm is adopted to predict a bird movement track, and early warning and linkage with a bird repelling device are realized. The problems of inaccurate positioning, poor environmental adaptability, early warning lag and the like in the prior art are solved, and the accuracy, the reliability and the active defense capability of bird monitoring of the transformer substation are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of substation safety monitoring, more particularly, to a substation bird three-dimensional positioning method and system based on dual infrared thermal imagers. BACKGROUND

[0002] As a key node of the power system, the safe and stable operation of the substation is of great importance. The activities of birds in the substation, such as nesting, building nests or flying, may cause equipment short circuit, insulator pollution and other faults, which seriously threaten the safety of the power grid. Therefore, an effective bird monitoring and driving system is an important part of intelligent operation and maintenance of the substation. At present, the existing bird monitoring technology of the substation mainly has the following limitations and defects: Inaccurate positioning based on monocular vision: Traditional monitoring methods mostly rely on monocular visible light or infrared thermal imagers. Monocular systems can only provide two-dimensional plane information of the target, and cannot obtain the accurate height (Z coordinate) and real three-dimensional position of the bird. This leads to the inability to accurately determine the real distance between the bird and the live equipment (such as conductors and insulators), making it difficult to achieve precise early warning and linkage control based on spatial position, and the false positive rate and false negative rate are high.

[0003] Target recognition is difficult in complex thermal background: There are many heat-emitting devices such as transformers and conductor joints in the substation, forming a complex static thermal background. A single temperature threshold method or a simple background difference method is easily disturbed by these fixed heat sources, environmental temperature fluctuations and sunlight reflection, which may misjudge the heat devices as birds or submerge the bird targets in the thermal background, resulting in poor recognition reliability.

[0004] Poor environmental adaptability: The detection sensitivity parameters (such as temperature threshold, background model update rate) of existing systems are usually fixed and preset. However, different time of day, different seasons and changing weather conditions (such as temperature and humidity, light, wind speed) can significantly affect the infrared imaging effect and bird activity characteristics. Fixed parameters are difficult to adapt to all-weather and all-season changes, and the system performance decreases dramatically in bad weather.

[0005] Passive response rather than active early warning: Most systems only trigger the bird driving device when the bird has entered the dangerous area, which belongs to the passive response mode. Lack of prediction ability of bird movement trajectory, unable to predict its flight path and the dangerous area it will enter in advance, the early warning timeliness is insufficient, and it is difficult to effectively prevent bird damage accidents.

[0006] Lack of multi-modal information fusion: Existing solutions often process image data in isolation and fail to effectively fuse environmental and meteorological information to dynamically optimize detection algorithms. The system performance is greatly affected by the environment, lacks adaptive adjustment ability, and the degree of intelligence is limited.

[0007] Therefore, there is an urgent need in this field for a substation bird monitoring and removal solution that can achieve accurate three-dimensional positioning, highly reliable identification, strong environmental adaptability, and proactive early warning capabilities. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention proposes a method and system for three-dimensional bird localization in substations based on dual infrared thermal imagers.

[0009] The technical solution of this invention is as follows: This invention proposes a three-dimensional bird positioning system for substations based on dual infrared thermal imagers, comprising: The data acquisition module includes two infrared thermal imagers, one horizontally mounted and one vertically mounted, as well as a time synchronization unit for synchronizing the two external thermal imagers. The background heat source filtering module is used to construct a static heat source map of the substation, extract dynamic targets, and preliminarily filter bird targets through temperature thresholds. The 3D coordinate fusion module is used to fuse the 2D observation data of two thermal imagers into 3D spatial coordinates through coordinate system I, target association judgment and 3D calculation unit; The data output and linkage control module is used to output three-dimensional coordinate data and link it with the bird deterrent device. The deep learning recognition module uses a convolutional neural network model to perform real-time recognition and classification of thermal imaging targets, distinguishing between birds and non-bird targets. The multimodal weather compensation module connects to temperature, humidity, light intensity, and wind speed sensors to dynamically adjust the temperature threshold and background modeling sensitivity of the detection system. The trajectory prediction and early warning module uses an improved Kalman filter algorithm to predict bird movement trajectories in the short term and trigger bird deterrence devices in advance.

[0010] Preferably, the background heat source filtering module includes: Static heat source map construction unit: used to establish a static background heat source model of the substation monitoring area by statistically analyzing multiple frames of infrared images collected during periods of no bird activity; the static background heat source model includes the fixed thermal radiation characteristics of transformers, conductors, insulator strings, and metal frame equipment in the substation and their positions in the images; Dynamic target extraction unit: During real-time monitoring, the current frame is compared with the static heat source map by using frame difference method, Gaussian mixture model or background subtraction algorithm such as ViBe to quickly identify the moving pixel area in the image; Temperature threshold filtering unit: Filters candidate bird targets by setting the average or highest temperature of the extracted moving pixel region and the bird's body temperature range.

[0011] Preferably, the deep learning recognition module adopts a lightweight convolutional neural network model and is deployed on an edge computing device or cloud server to further identify and classify candidate bird targets screened by the temperature threshold filtering unit.

[0012] Preferably, the multimodal meteorological compensation module is connected to temperature, humidity, light, and wind speed sensors in the substation, and dynamically adjusts the temperature threshold of the temperature threshold filtering unit and the sensitivity of static heat source map background modeling based on sensor data.

[0013] Preferably, the improved Kalman filter algorithm includes: An adaptive noise covariance adjustment mechanism is added to the Kalman filter to adaptively adjust the process noise covariance according to the bird target movement, thereby improving the confidence in the observation values. Introducing constraints from bird kinematics models can improve the accuracy of bird flight trajectory prediction.

[0014] Preferably, the adaptive noise covariance adjustment mechanism includes: The process noise covariance matrix and measurement noise covariance matrix are dynamically adjusted based on the motion state of the bird target and the prediction error. When the speed of the bird target changes beyond a set threshold, the process noise covariance is automatically increased to adapt to the uncertainty of the motion model.

[0015] Preferably, in the three-dimensional coordinate fusion module, the target association judgment is based on the X-coordinate consistency criterion, that is, if the absolute difference of the X-coordinates of the infrared images acquired by the two external thermal imagers is less than a preset threshold, they are determined to be the same target.

[0016] On the other hand, the present invention also provides a three-dimensional bird localization method for substations based on dual infrared thermal imagers, comprising the following steps: Infrared image data of the substation area were acquired using two infrared thermal imagers, one installed horizontally and the other vertically. A static heat source map of the substation is constructed based on infrared image data, dynamic targets are extracted, and bird targets are initially screened by setting temperature thresholds. A convolutional neural network model is used to identify and classify bird targets in real time, distinguishing between bird and non-bird targets. By integrating data from temperature, humidity, light intensity, and wind speed sensors, the temperature threshold and background modeling sensitivity of the detection system are dynamically adjusted. Two-dimensional observation data from two thermal imagers are fused into three-dimensional spatial coordinates through coordinate system I, target association judgment, and three-dimensional calculation unit; and an improved Kalman filter algorithm is used to predict bird movement trajectories in the short term and trigger bird deterrence devices in advance.

[0017] In another aspect, the present invention also provides an electronic device having a computer program stored thereon, wherein the computer program, when executed by a processor, implements a three-dimensional bird localization method for substations based on a dual infrared thermal imager as described in any embodiment of the present invention.

[0018] In another aspect, the present invention also provides a computer-readable medium for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement a three-dimensional bird localization method for substations based on dual infrared thermal imagers as described in any embodiment of the present invention.

[0019] The present invention has the following beneficial effects: 1. By employing dual infrared thermal imagers installed horizontally and vertically in tandem, and combining precise time synchronization and coordinate system fusion technology, the two two-dimensional observation surfaces are intersected and merged to directly calculate the three-dimensional spatial coordinates (X, Y, Z) of bird targets, including their altitude above the ground. This provides accurate data support for determining the true distance between birds and high-risk equipment in substations, laying a solid foundation for precise and coordinated bird control.

[0020] 2. A multi-level filtering mechanism of "background heat source filtering + deep learning recognition" is adopted. By constructing a static heat source map and temperature threshold for preliminary screening, most of the interference from fixed heat sources is effectively eliminated; then, a deep learning model based on convolutional neural networks is used to perform fine recognition of candidate targets, which can effectively distinguish between birds and non-bird targets, significantly reducing the false alarm rate and false negative rate of the system.

[0021] 3. Through a multimodal meteorological compensation module, the system senses changes in ambient temperature, humidity, light intensity, and wind speed in real time and dynamically adjusts the core parameters of the detection algorithm (such as temperature threshold and background modeling sensitivity). This enables the system to adapt to different seasons, time periods, and complex meteorological conditions, ensuring the stability of monitoring performance in all-weather scenarios and solving the problem of poor adaptability of fixed-parameter systems.

[0022] 4. A trajectory prediction and early warning module based on an improved Kalman filter algorithm is introduced. This algorithm, through adaptive noise adjustment and kinematic model constraints, can accurately predict the short-term flight trajectories of birds. Based on the prediction results, the system can trigger bird deterrence devices in advance before birds actually enter the danger zone, transforming passive response into active intervention, greatly improving the timeliness of early warning and the effectiveness of bird deterrence, and effectively preventing problems before they occur. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0024] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0025] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0026] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0027] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0028] Example 1: To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of this application.

[0029] To address the problems in existing technologies, this embodiment provides a three-dimensional bird positioning system for substations based on dual infrared thermal imagers, comprising: The data acquisition module includes two infrared thermal imagers, one horizontally mounted and one vertically mounted, and a time synchronization unit for synchronizing the two external thermal imagers; wherein: Infrared thermal imager A: Horizontally mounted on the substation frame or the top of a building, with its optical axis roughly parallel to the ground, it is mainly used to cover the horizontal (XY) plane of the monitoring area and obtain the horizontal coordinates of birds. .

[0030] Infrared thermal imager B: Installed vertically or at an angle on a standalone tower, light pole, or dedicated pole, with its optical axis at a significant angle to the ground (preferably approximately perpendicular). It is mainly used to cover the vertical (XZ or YZ) plane of the monitoring area and acquire the horizontal and vertical coordinates of birds. .

[0031] Timing Unit: BeiDou / GPS timing is preferred, either integrated into the two thermal imagers or used as an external synchronization controller. This unit ensures that every frame of image data acquired by the two thermal imagers has a unified and accurate timestamp, controlling the data time error between the two devices to within milliseconds (e.g., ≤10ms) to meet the real-time requirements of moving target association.

[0032] The background heat source filtering module is used to construct a static heat source map of the substation, extract dynamic targets, and preliminarily filter bird targets through temperature thresholds. The background heat source filtering module includes: Static heat source map construction unit: During system initialization or periodic self-learning phases, a static background heat source model of the substation monitoring area is established by statistically analyzing infrared image sequences taken over a period of time (e.g., 30 minutes) when there is no bird activity. This model records the fixed thermal radiation characteristics of equipment such as transformers, conductors, insulator strings, and metal frames, as well as their locations in the images.

[0033] Dynamic target extraction unit: During real-time monitoring, the current frame is compared with the static heat source map by using background subtraction algorithms such as frame difference method, Gaussian mixture model (GMM) or ViBe to quickly identify the moving pixel region in the image.

[0034] Temperature threshold filtering unit: Calculates the average or maximum temperature of the extracted moving pixel region. Based on the biological characteristic that birds' body temperature is typically maintained between 35°C and 42°C, a temperature filtering threshold is set. Only moving regions with temperatures falling within this preset range are retained, initially identifying them as candidate bird targets, thereby effectively filtering out interference caused by changes in ambient temperature, cloud cover, or non-biological heat sources (such as hotspots from operating equipment).

[0035] The 3D coordinate fusion module is used to fuse the 2D observation data of two thermal imagers into 3D spatial coordinates through coordinate system I, target association judgment and 3D calculation unit; The 3D coordinate fusion module includes: Coordinate System 1 and Calibration Unit: During system deployment, joint calibration technology is used to determine the intrinsic parameters (focal length, principal point, distortion coefficient) and extrinsic parameters (rotation matrix and translation vector) of the two thermal imagers, thereby establishing a unified world coordinate system for both.

[0036] Target Association Judgment Unit: After background heat source filtering, candidate bird targets detected by thermal imager A and thermal imager B near the same timestamp are evaluated for association. The core association logic is based on the principle that the X-coordinate of the same target should be consistent in both views. Specifically, it calculates the absolute difference between the horizontal coordinate X1 of the target detected by thermal imager A and the horizontal coordinate X2 of the target detected by thermal imager B. If the difference is less than or equal to a preset error tolerance threshold δ (which can be set according to the on-site calibration accuracy and sensor resolution, for example, 0.3 meters), then the two two-dimensional observation results are considered to originate from the same bird in space.

[0037] 3D coordinate calculation unit: For a successfully associated target pair, its 3D spatial coordinates (X, Y, Z) are calculated as follows: To reduce error, the X-coordinate is averaged.

[0038] The Y-coordinate is taken directly from the observation value of thermal imager A:

[0039] The Z-coordinate (height) is taken directly from the observations of thermal imager B:

[0040] Exception handling unit: when If the target association fails, the system is considered to have failed. The system can trigger a target re-identification mechanism, such as combining information about the target's size, shape, and movement trends for secondary matching. If re-identification still fails, the location result for that frame is temporarily discarded as abnormal data, and a log is recorded for analysis to avoid outputting incorrect location information.

[0041] The data output and linkage control module is used to output three-dimensional coordinate data and link it with the bird deterrent device; specifically: Data interface: Outputs the final 3D coordinates (X, Y, Z), target ID, timestamp, confidence level, and other information in real time through standard communication protocols (such as TCP / IP, MQTT, Modbus).

[0042] Bird deterrence linkage unit: Receives location data and links with active bird deterrence devices deployed within the substation, including sound, light, and electricity devices (such as ultrasonic bird deterrents, high-intensity flashing lights, and voice-activated bird deterrent speakers). Based on preset rules, such as when birds enter an "electronic fence" area with a radius of R centered on high-risk equipment, it automatically triggers the bird deterrence device at the designated location for directional and precise deterrence.

[0043] The deep learning recognition module employs a convolutional neural network (CNN) model for real-time identification and classification of thermal imaging targets, distinguishing between birds and non-bird targets. Specifically, the CNN model preferably uses a lightweight network structure, such as MobileNetV2, SqueezeNet, or a specially designed CNN, to ensure real-time inference performance on edge computing devices. The input to the CNN is an infrared image patch of candidate bird target regions preprocessed by the background heat source filtering module.

[0044] The model was trained using a dataset of infrared bird images with extensive annotations. This dataset covers different bird species (such as sparrows, pigeons, and eagles) and their thermal imaging features under various postures, distances, and environments. It also includes a wealth of negative samples (such as flying insects, fallen leaves, cloud shadows, and interference from device heat sources). Data augmentation techniques such as rotation, scaling, and brightness adjustment were employed during training to improve the model's generalization ability.

[0045] In real-time monitoring, this module receives candidate target image patches from the background heat source filtering module, extracts depth features using a trained convolutional neural network model, and outputs the probability value of whether the target belongs to "bird" or "non-bird" through a final classifier such as Softmax. A classification confidence threshold is set; in this embodiment, it is set to 0.9. When the confidence of the "bird" category is higher than this threshold, the target is determined to be a bird, and its information (such as target ID, location, and confidence) is passed to subsequent modules; otherwise, it is filtered out as an interference target, effectively reducing the false alarm rate.

[0046] The deep learning recognition module can be deployed on edge computing devices at substation sites to achieve low-latency localized recognition; it can also upload image data to cloud servers for recognition, making it suitable for complex models with higher computing resource requirements.

[0047] The multimodal weather compensation module integrates temperature, humidity, light intensity, and wind speed sensors to dynamically adjust the temperature threshold and background modeling sensitivity of the detection system; specifically: The system reads the measured values ​​of temperature and humidity sensors, light intensity sensors, and wind speed sensors deployed in the substation in real time via industrial bus (such as RS485) or wireless communication.

[0048] Dynamic temperature threshold adjustment: Based on changes in ambient temperature, the threshold range of the temperature threshold filtering unit in the background heat source filtering module is dynamically adjusted. For example, in hot summer weather, when the ambient temperature is close to or exceeds the bird's body temperature, the temperature threshold range is appropriately narrowed (e.g., adjusted to 38℃ to 41℃) to enhance the suppression of high-temperature interference; in cold winter environments, when the difference between the ambient temperature and the bird's body temperature is significant, the threshold range can be appropriately widened (e.g., adjusted to 34℃ to 42℃) to ensure effective bird capture. The adjustment logic can be based on a preset temperature-threshold lookup table or calculated using empirical formulas.

[0049] Background modeling sensitivity adjustment: Based on changes in ambient light intensity (affecting the noise level of infrared imaging) and wind speed (which may cause slight background swaying), dynamically adjust the parameters of the static heat source map construction unit and the dynamic target extraction unit, such as the background model update rate and the sensitivity threshold for foreground detection. In strong light or windy weather, appropriately increase the update frequency of the background model and adjust the foreground detection threshold to reduce false alarms caused by changes in light or vegetation movement.

[0050] Compensation logic: Through the built-in compensation algorithm, the influence weight of various meteorological factors is comprehensively evaluated, and adjustment instructions for temperature threshold and background modeling parameters are output to realize the adaptive response of the detection system to environmental changes and improve the stability of all-weather monitoring.

[0051] The trajectory prediction and early warning module uses an improved Kalman filter algorithm to predict bird movement trajectories in the short term and trigger bird deterrence devices in advance. Specifically: State vector definition: The state vector is composed of the three-dimensional spatial coordinates (X, Y, Z) of the bird target and its velocity components (Vx, Vy, Vz) in the X, Y, and Z directions.

[0052] Improved Kalman Filter Algorithm: Based on the standard Kalman filter prediction and update framework, the following improvements are made: Adaptive noise covariance adjustment mechanism: Based on the real-time motion state of the bird target (such as the rate of change of acceleration) and the prediction error covariance matrix of the previous moment, the process noise covariance matrix Q and the measurement noise covariance matrix R are dynamically adjusted. When the target's motion is drastic and the velocity change exceeds a set threshold, the process noise covariance is automatically increased to respond more quickly to changes in the target's motion state; when the observation data is continuous and stable, the measurement noise covariance is appropriately reduced to increase the confidence in the observation values.

[0053] Bird kinematic model constraints: Introducing simplified bird kinematic models (such as uniform acceleration or uniform velocity models) as constraints into the state prediction equations better reflects the actual flight patterns of birds, thereby improving the accuracy of trajectory prediction.

[0054] Trajectory Prediction and Early Warning: Utilizing an improved Kalman filter, the system predicts the target's short-term trajectory and potential spatial location based on its current and historical status information. Combined with substation equipment layout information, if the predicted trajectory indicates that the bird will enter a pre-defined danger zone or core equipment protection zone in the near future, an early warning signal is generated.

[0055] Early warning and linkage control: The early warning signal is sent to the data output and linkage control module, which instructs it to trigger the bird deterrent device (such as sound, light, or electric bird deterrent) at the corresponding location in advance, realizing the transformation from passive detection to active early warning and early intervention, effectively improving the timeliness and accuracy of bird deterrence, and preventing birds from approaching high-risk equipment.

[0056] Example 2: This embodiment provides a three-dimensional bird localization method for substations based on dual infrared thermal imagers, including the following steps: Infrared image data of the substation area were acquired using two infrared thermal imagers, one installed horizontally and the other vertically. A static heat source map of the substation is constructed based on infrared image data, dynamic targets are extracted, and bird targets are initially screened by setting temperature thresholds. A convolutional neural network model is used to identify and classify bird targets in real time, distinguishing between bird and non-bird targets. By integrating data from temperature, humidity, light intensity, and wind speed sensors, the temperature threshold and background modeling sensitivity of the detection system are dynamically adjusted. Two-dimensional observation data from two thermal imagers are fused into three-dimensional spatial coordinates through coordinate system I, target association judgment, and three-dimensional calculation unit; and an improved Kalman filter algorithm is used to predict bird movement trajectories in the short term and trigger bird deterrence devices in advance.

[0057] Example 3: This embodiment provides an electronic device that stores a computer program. When the computer program is executed by a processor, it implements a three-dimensional bird localization method for substations based on dual infrared thermal imagers, as described in any embodiment of the present invention.

[0058] Example 4: This embodiment provides a computer-readable medium for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement a three-dimensional bird localization method for substations based on dual infrared thermal imagers as described in any embodiment of the present invention.

[0059] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0060] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0061] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0062] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0063] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A three-dimensional bird positioning system for substations based on dual infrared thermal imagers, characterized in that, include: The data acquisition module includes two infrared thermal imagers, one horizontally mounted and one vertically mounted, as well as a time synchronization unit for synchronizing the two external thermal imagers. The background heat source filtering module is used to construct a static heat source map of the substation, extract dynamic targets, and preliminarily filter bird targets through temperature thresholds. The 3D coordinate fusion module is used to fuse the 2D observation data of two thermal imagers into 3D spatial coordinates through coordinate system I, target association judgment and 3D calculation unit; The data output and linkage control module is used to output three-dimensional coordinate data and link it with the bird deterrent device. The deep learning recognition module uses a convolutional neural network model to perform real-time recognition and classification of thermal imaging targets, distinguishing between birds and non-bird targets. The multimodal weather compensation module connects to temperature, humidity, light intensity, and wind speed sensors to dynamically adjust the temperature threshold and background modeling sensitivity of the detection system. The trajectory prediction and early warning module uses an improved Kalman filter algorithm to predict bird movement trajectories in the short term and trigger bird deterrence devices in advance.

2. The three-dimensional bird positioning system for substations based on dual infrared thermal imagers according to claim 1, characterized in that: The background heat source filtering module includes: Static heat source map construction unit: used to establish a static background heat source model of the substation monitoring area by statistically analyzing multiple frames of infrared images collected during periods of no bird activity; the static background heat source model includes the fixed thermal radiation characteristics of transformers, conductors, insulator strings, and metal frame equipment in the substation and their positions in the images; Dynamic target extraction unit: During real-time monitoring, the current frame is compared with the static heat source map by using frame difference method, Gaussian mixture model or background subtraction algorithm such as ViBe to quickly identify the moving pixel area in the image; Temperature threshold filtering unit: Filters candidate bird targets by setting the average or highest temperature of the extracted moving pixel region and the bird's body temperature range.

3. A three-dimensional bird positioning system for substations based on dual infrared thermal imagers according to claim 2, characterized in that: The deep learning recognition module employs a lightweight convolutional neural network model, deployed on edge computing devices or cloud servers, to further identify and classify candidate bird targets screened by the temperature threshold filtering unit.

4. A three-dimensional bird positioning system for substations based on dual infrared thermal imagers according to claim 2, characterized in that: The multimodal meteorological compensation module is connected to temperature, humidity, light, and wind speed sensors in the substation, and dynamically adjusts the temperature threshold of the temperature threshold filtering unit and the sensitivity of static heat source map background modeling based on sensor data.

5. A three-dimensional bird positioning system for substations based on dual infrared thermal imagers according to claim 1, characterized in that: The improved Kalman filter algorithm includes: An adaptive noise covariance adjustment mechanism is added to the Kalman filter to adaptively adjust the process noise covariance according to the bird target movement, thereby improving the confidence in the observation values. Introducing constraints from bird kinematics models can improve the accuracy of bird flight trajectory prediction.

6. A three-dimensional bird positioning system for substations based on dual infrared thermal imagers according to claim 5, characterized in that: The adaptive noise covariance adjustment mechanism includes: The process noise covariance matrix and measurement noise covariance matrix are dynamically adjusted based on the motion state of the bird target and the prediction error. When the speed of the bird target changes beyond a set threshold, the process noise covariance is automatically increased to adapt to the uncertainty of the motion model.

7. A three-dimensional bird positioning system for substations based on dual infrared thermal imagers according to claim 1, characterized in that: In the three-dimensional coordinate fusion module, the target association judgment is based on the X-coordinate consistency criterion, that is, if the absolute difference of the X-coordinate of the infrared images acquired by two external thermal imagers is less than a preset threshold, they are determined to be the same target.

8. A three-dimensional bird localization method for substations based on dual infrared thermal imagers, used to implement the system described in any one of claims 1-7, characterized in that, Includes the following steps: Infrared image data of the substation area were acquired using two infrared thermal imagers, one installed horizontally and the other vertically. A static heat source map of the substation is constructed based on infrared image data, dynamic targets are extracted, and bird targets are initially screened by setting temperature thresholds. A convolutional neural network model is used to identify and classify bird targets in real time, distinguishing between bird and non-bird targets. By integrating data from temperature, humidity, light intensity, and wind speed sensors, the temperature threshold and background modeling sensitivity of the detection system are dynamically adjusted. Two-dimensional observation data from two thermal imagers are fused into three-dimensional spatial coordinates through coordinate system I, target association judgment, and three-dimensional calculation unit; and an improved Kalman filter algorithm is used to predict bird movement trajectories in the short term and trigger bird deterrence devices in advance.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the three-dimensional bird localization method for substations based on dual infrared thermal imagers as described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the three-dimensional bird localization method for substations based on dual infrared thermal imagers as described in claim 8.