Bird interference prediction and protection method based on big data analysis

Abstract

The invention discloses a bird interference prediction and protection method based on big data analysis, and relates to the technical field of ecological monitoring, and the method comprises the steps: continuously collecting video images and environment data of a target region, and carrying out the integration according to a timestamp, and obtaining a risk clue set; mapping the interference risk file to a specific execution sequence by adopting a non-intrusive priority strategy, and determining a field execution instruction; performing validity judgment on the non-intrusive priority strategy based on the real-time effect record and the adjustment track, updating the interference risk file, and generating an updated strategy and file; and according to the updated strategy and file, setting a cruise and light interference window, and configuring a grading gear and a pointing sequence to form a next-cycle prevention plan and an equipment instruction set. According to the method, the interference risk file is mapped to the specific execution sequence by adopting the non-intrusive priority strategy, the field execution instruction is determined, and balance control between the protection intensity and the bird sensitivity is realized.

Description

Technical Field

[0001] This invention relates to the field of ecological monitoring technology, and in particular to a method for predicting and protecting against bird disturbances based on big data analysis. Background Technology

[0002] In scenarios such as ecological protection, aviation operations, and wind power generation, bird activity can cause equipment interference, flight risks, or environmental imbalances. Existing methods typically rely on video surveillance, acoustic recognition, or manual patrols to detect and record bird activity trajectories. By deploying multi-source sensors and analyzing time-series data, bird activity frequency, flight direction, and aggregation characteristics can be obtained, enabling basic regional risk assessment and early warning. In recent years, integrated monitoring systems combining computer vision and meteorological data have been widely applied, providing high-precision bird activity identification capabilities for target areas and assisting protective devices in initial response.

[0003] However, conventional methods based on sensor perception and image analysis have limitations in terms of the spatiotemporal continuity of data and the accuracy of behavior prediction. On the one hand, conventional methods often focus on static analysis using single-moment or short-cycle signals, failing to fully reflect the migration patterns of bird behavior over time. On the other hand, the dynamic coupling between on-site protection strategies and monitoring feedback is insufficient, making it difficult to update protection instructions in a timely manner, resulting in weak proactiveness in interference response. Therefore, further optimization is needed in terms of continuous monitoring and closed-loop strategy adjustment. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a bird disturbance prediction and protection method based on big data analysis to solve the problem of insufficient dynamic coupling between on-site protection strategies and monitoring feedback.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for predicting and protecting against bird disturbances based on big data analysis. The method includes: continuously collecting video footage and environmental data of the target area and integrating them by timestamp to obtain a risk clue set; performing spatiotemporal prediction based on the risk clue set and completing behavioral learning classification to generate a disturbance risk profile; mapping the disturbance risk profile to specific execution sequences using a non-intrusive priority strategy to determine on-site execution instructions; executing on-site execution instructions in the target area, dynamically correcting the gimbal's pointing and intensity, and collecting on-site feedback data in real time to obtain real-time performance records and adjustment trajectories; determining the effectiveness of the non-intrusive priority strategy based on the real-time performance records and adjustment trajectories, updating the disturbance risk profile, and generating an updated strategy and profile; and setting patrol and mild disturbance windows based on the updated strategy and profile, configuring graded levels and pointing sequences to form the next cycle's prevention plan and equipment instruction set.

[0007] As a preferred embodiment of the bird disturbance prediction and protection method based on big data analysis described in this invention, the steps of continuously collecting video footage and environmental data of the target area and integrating them according to timestamps to obtain a risk clue set are as follows. The system continuously collects video footage and environmental data of the target area using a cruise-style framing method. The video footage and environmental data are preprocessed and integrated according to timestamps to obtain a set of risk clues.

[0008] As a preferred embodiment of the bird disturbance prediction and protection method based on big data analysis described in this invention, the specific steps for generating a disturbance risk profile by performing spatiotemporal prediction based on a risk cue set and completing behavioral learning classification are as follows. Spatiotemporal correlation and coherence checks were performed on risk cues within the same time window to obtain sequences of bird activity signs and environmental characteristics. Based on the comparison and difference annotation of similar scenarios using bird activity sign sequences and environmental feature sequences, a candidate list of regions is generated. The candidate regional list is assessed for probability and non-intrusive priorities are determined, and then integrated into an interference risk profile.

[0009] As a preferred embodiment of the bird disturbance prediction and protection method based on big data analysis described in this invention, the step of mapping the disturbance risk profile to a specific execution sequence using a non-intrusive priority strategy to determine the on-site execution instructions involves the following specific steps. A non-intrusive priority strategy is adopted to perform non-intrusive priority expansion on each record in the interference risk file to obtain an initial strategy sequence; The initial strategy sequence is matched sequentially based on the rhythm of sound wave frequency changes, light intensity level and gimbal pointing to generate an execution orchestration list; The feasibility of the execution schedule is checked and refined to form on-site execution instructions.

[0010] As a preferred embodiment of the bird disturbance prediction and protection method based on big data analysis described in this invention, the step of executing on-site execution instructions in the target area refers to sending on-site execution instructions to the equipment in the target area and executing the on-site execution instructions.

[0011] As a preferred embodiment of the bird disturbance prediction and protection method based on big data analysis described in this invention, the specific steps of dynamically correcting the gimbal's pointing and intensity and collecting real-time on-site feedback data to obtain real-time performance records and adjustment trajectories are as follows. Based on the on-site execution commands, the gimbal's real-time attitude, audio-visual output intensity and target position are synchronously compared and deviation analyzed to obtain the pointing correction amount and intensity adjustment amount; The gimbal angle and output energy level are dynamically corrected based on the pointing correction amount and intensity adjustment amount, and video footage and environmental data are collected simultaneously to generate real-time performance records and adjustment trajectories.

[0012] As a preferred embodiment of the bird disturbance prediction and protection method based on big data analysis described in this invention, the specific steps for determining the effectiveness of the non-intrusive priority strategy based on real-time performance records and trajectory adjustments are as follows: By comparing and analyzing the interference process based on real-time effectiveness records and adjusted trajectories, the similarity matching degree and the effectiveness of the expulsion can be obtained. The non-intrusion priority strategy is comprehensively judged based on similarity matching degree and expulsion effectiveness, and a conclusion on the effectiveness of the strategy is generated.

[0013] As a preferred embodiment of the bird disturbance prediction and protection method based on big data analysis described in this invention, the steps for updating the disturbance risk profile and generating updated strategies and profiles are as follows: Based on the conclusions of the effectiveness of the strategy, the interference risk profile is compared and weighted item by item to obtain updated elements. The non-intrusive priority strategy is integrated and revised based on the updated elements, generating updated strategies and files.

[0014] As a preferred embodiment of the bird interference prediction and protection method based on big data analysis described in this invention, the step of setting the cruise and mild interference windows according to the updated strategy and files refers to preset the cruise path and mild interference duration based on the updated strategy and files, and generating the cruise and mild interference windows.

[0015] As a preferred embodiment of the bird disturbance prediction and protection method based on big data analysis described in this invention, the specific steps for configuring graded levels and pointing sequences to form the next cycle's prevention plan and equipment instruction set are as follows: Based on the cruise and light interference windows, the target area is correlated to obtain the grade level and corresponding pointing range; Based on the grade level and target range, the equipment is combined and configured to form the next cycle prevention plan and equipment instruction set.

[0016] The beneficial effects of this invention are as follows: by performing spatiotemporal prediction based on the risk cue set and completing behavioral learning classification, an interference risk profile is generated, which realizes the dynamic identification and risk classification of bird activity patterns in the target area, providing a continuous predictive basis for subsequent protection decisions, thereby enabling the early construction of a protection response framework in the temporal dimension; by adopting a non-intrusive priority strategy to map the interference risk profile to a specific execution sequence and determine the on-site execution instructions, a balance control between protection intensity and bird sensitivity is achieved, maintaining the continuity of protection effect while reducing ecological disturbance. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a bird disturbance prediction and protection method based on big data analysis.

[0019] Figure 2 A flowchart for integrating environmental data.

[0020] Figure 3 A flowchart for categorizing behavioral learning.

[0021] Figure 4 The flowchart shows the execution of a non-intrusive priority strategy. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for predicting and protecting against bird disturbances based on big data analysis, including the following steps: S1. Continuously collect video footage and environmental data of the target area and integrate them according to timestamps to obtain a set of risk clues; The system continuously collects video footage and environmental data of the target area using a cruise-style framing method. Furthermore, the route of the wind farm area is distributed to drones for patrol and scouting. By rotating the camera equipment and continuously changing the shooting angle around the wind turbine tower and turbine blades, a full-range video image and environmental data of the wind farm are obtained. The video image is captured by the camera in a continuous recording mode to capture the dynamic changes of birds' activities near the wind turbine. Environmental data, including light, temperature, humidity and wind speed, are recorded by environmental sensors deployed in key locations of the wind farm (such as the turbine base and blade projection area). The video footage and environmental data are preprocessed and integrated according to timestamps to obtain a set of risk clues; Furthermore, the acquired images are denoised and enhanced using median filtering, with particular optimization for motion blur caused by the rotating blades of the wind turbine. For environmental data, such as illumination, temperature, humidity, and wind speed, linear interpolation is used to handle missing values ​​and ensure data integrity. The processed data is then integrated according to timestamps to generate a risk clue set.

[0026] S2. Based on the risk clue set, perform spatiotemporal prediction and complete behavioral learning classification to generate an interference risk profile; Spatiotemporal correlation and coherence checks were performed on risk cues within the same time window to obtain sequences of bird activity signs and environmental characteristics. Furthermore, for risk clue sets within the same time window, spatiotemporal correlation is performed using timestamp matching to obtain spatiotemporal correlated data. Missing values ​​in the spatiotemporal correlated data are processed using linear interpolation, and the continuity of video image sequences in the spatiotemporal correlated data is checked using median filtering to obtain smooth correlated data. The frame difference method is used to separate the video image portion from the smooth correlated data as a sequence of bird activity signs (e.g., bird gathering behavior near wind turbines), and the frame difference method is used to separate the environmental data portion as an environmental feature sequence, generating bird activity sign sequences and environmental feature sequences. It should be noted that the frame difference method refers to selecting two adjacent frames of images, calculating the absolute difference of the grayscale or RGB values ​​of pixels at the same position to obtain a difference image, and then using the difference binarization method to binarize the difference image to extract the contour and trajectory of the moving target; the bird activity sign sequence is essentially a numerical sequence extracted from the motion changes between consecutive video frames. Based on the comparison and difference annotation of similar scenarios using bird activity sign sequences and environmental feature sequences, a candidate list of regions is generated. Furthermore, based on bird activity sign sequences and environmental feature sequences, motion detection is performed on the bird activity sign sequences using the frame difference method to obtain motion sequences. The motion sequences are then grouped into similar pattern groups using the mean clustering method to obtain activity pattern groups (e.g., clustering patterns near wind turbine blades, turbine stress avoidance patterns, and flight patterns around towers). Gap filling is performed on the environmental feature sequences using linear interpolation to obtain complete environmental feature sequences. The average deviation of the complete environmental feature sequences is calculated using the sliding window method to obtain environmental variation groups. The activity pattern groups and environmental variation groups are aligned using a timestamp matching method to obtain a control scenario. Visual differences are calculated from the bird activity sign sequence portion of the control scenario using the frame difference method, and environmental differences are calculated from the environmental feature sequence portion of the control scenario using the frame difference method. Visual differences and environmental differences are used as difference labels. A candidate list of regions is generated by extracting spatial locations with difference labels from the control scenario (e.g., high-risk wind turbine areas). It should be noted that mean clustering is a method that groups similar patterns by representing motion sequences as feature vectors and iteratively minimizing the intra-cluster variance. First, initial cluster centers are randomly selected. Then, each motion sequence is assigned to the nearest cluster center based on Euclidean distance. The average value of the sequences in each cluster is then calculated to update the cluster centers. This assignment and update process is repeated until the cluster centers are stable, thereby aggregating motion sequences with similar dynamic changes into groups. The contrasting scenario refers to strategies and files that have been persistently stored. In contrast, existing technologies often rely on a single sensor or simple computer vision for isolated detection, which cannot distinguish between birds and distracting objects and is prone to misjudgment. This step combines video (frame difference method) and environmental data (sliding window average deviation) to achieve non-isolated detection, and quantifies the degree of environmental interference through difference labeling to avoid misjudgment. The candidate regional list is assessed for probability and non-intrusive priorities are determined, and then integrated into an interference risk profile. Furthermore, the bird activity sign sequences and environmental characteristic sequences of each entry in the regional candidate list were standardized in terms of dimensions with the high bird activity sequences in the risk cue set using the linear normalization method. The similarity was then calculated using Euclidean distance to obtain a probability assessment score, expressed as: ; in, As a probability assessment score, This represents the total number of entries. For the entry Sequences of bird activity signs For the entry High-altitude bird activity sequences; The environmental adaptation score is obtained by calculating the deviation between the environmental characteristic sequence of each item in the regional candidate list and the standard environmental sequence using the differential calculation method. The expression is as follows: ; in, To score environmental adaptability, This represents the total number of entries. For the current entry, For the entry Environmental feature sequence, For the entry Standard environmental sequence; Based on the environmental adaptation score and the probability assessment score, the non-intrusive priority score is calculated by weighted summation. The regional candidate list items, the probability assessment score, and the non-intrusive priority score are integrated to generate an interference risk profile. It should be noted that the standard environmental sequence is obtained by aligning the environmental feature sequences corresponding to the persistently stored strategies and historical successful eviction sequences in the archives using a timestamp matching method, and then calculating the average deviation using the sliding window method. When calculating the average deviation of the environmental feature sequences based on the sliding window method, environmental parameters such as light, temperature, humidity, and wind speed in the environmental feature sequences have been converted into dimensionless deviation representations. Therefore, when calculating the deviation value between the environmental feature sequences and the standard environmental sequences using the difference calculation method, the environmental adaptation score remains consistent in dimensions, and no additional dimension unification processing is required.

[0027] S3. A non-intrusive priority strategy is adopted to map the interference risk profile to the specific execution sequence to determine the on-site execution instructions; A non-intrusive priority strategy is adopted to perform non-intrusive priority expansion on each record in the interference risk file to obtain an initial strategy sequence; Furthermore, the non-intrusive priority strategy specifically involves: calculating the vector distance between the probability assessment score and the environmental adaptation score in each record of the interference risk profile using the Euclidean distance method to obtain a priority vector; summing the non-intrusive related scores in the priority vector using a weighted summation method to obtain a non-intrusive priority ranking; matching the non-intrusive priority ranking with the records in the interference risk profile using a timestamp matching method to obtain expanded records; filling the missing priorities in the expanded records using a linear interpolation method to obtain a complete expanded sequence; calculating the average deviation of the complete expanded sequence using a sliding window method to obtain sequence stability; calculating the difference in sequence stability using a frame difference method to obtain an adjustment sequence; and grouping the adjustment sequence into similar pattern groups using a mean clustering method to obtain an initial strategy sequence. Ideally, this step dynamically quantifies priorities based on historical bird deterrence effectiveness data, and collaboratively optimizes the pairing of sound wave frequency, light intensity level and gimbal pointing through linear interpolation, sliding window stability calculation and frame difference method; ensuring minimal ecological disturbance, compared with traditional bird deterrence methods that default to high intensity stimulation (such as popping sounds), it can reduce the intensity of disturbance at the same bird deterrence success rate, while reducing the equipment coordination error rate and alleviating bird stress response; The initial strategy sequence is matched sequentially based on the rhythm of sound wave frequency changes, light intensity level and gimbal pointing to generate an execution orchestration list; Furthermore, based on the rhythm of sound wave frequency changes, light intensity levels, and gimbal pointing, the initial strategy sequence is paired using a timestamp matching method to obtain rhythm pairing sequences, level pairing sequences, and pointing pairing sequences. Linear interpolation is then used to fill these pairs to obtain complete rhythm, level, and pointing pairing sequences. A sliding window method is used to calculate the average deviation of the complete rhythm, level, and pointing pairing sequences to obtain a stability sequence. The frame difference method is used to calculate the differences in the stability sequence to obtain rhythm adjustment sequences, level adjustment sequences, and pointing adjustment sequences. Euclidean distance is used to calculate the vector distance between the rhythm adjustment sequences, level adjustment sequences, and pointing adjustment sequences to obtain pairing distances. A weighted summation method is used to sum the pairing distances to obtain a comprehensive pairing score. Mean clustering is used to group the comprehensive pairing scores into similar pattern groups to obtain orchestration groups. Finally, a timestamp matching method is used to integrate the orchestration groups with the initial strategy sequence to generate an execution orchestration list. The feasibility of the execution schedule is checked and refined to generate on-site execution instructions; Furthermore, the execution orchestration list is processed using Euclidean distance to calculate the vector distance between the rhythm adjustment sequence, level adjustment sequence, and direction adjustment sequence in the execution orchestration list. A feasibility score is obtained by summing the distance vectors using a weighted summation method. A checklist is obtained by matching the feasibility score with the execution orchestration list using a timestamp matching method. A complete checklist is obtained by filling in the missing scores in the checklist using linear interpolation. A check stability is obtained by calculating the average deviation of the complete checklist using a sliding window method. A check adjustment sequence is obtained by calculating the difference in check stability using a frame difference method. A refinement group is obtained by grouping the check adjustment sequence into similar patterns using a mean clustering method. Finally, the refinement group is integrated with the execution orchestration list using a timestamp matching method to form the field execution instructions.

[0028] S4. Execute on-site instructions in the target area, dynamically correct the gimbal's direction and intensity, and collect on-site feedback data in real time to obtain real-time results records and adjust the trajectory. The on-site execution instructions are issued to the equipment in the target area and executed. Furthermore, the on-site execution instructions are transmitted to the target area equipment (including the bird deterrent device installed on the wind turbine tower) via wireless communication. The target area equipment matches the timestamp in the on-site execution instructions with the equipment clock using a timestamp matching method. When the timestamp match is successful, the gimbal pointing adjustment operation (e.g., pointing to the turbine blade area), the sound wave frequency change rhythm output operation, and the light intensity level adjustment operation are executed to complete the execution of the on-site execution instructions. Based on the on-site execution commands, the gimbal's real-time attitude, audio-visual output intensity and target position are synchronously compared and deviation analyzed to obtain the pointing correction amount and intensity adjustment amount; Furthermore, the target position is obtained by detecting moving targets in real-time video footage using the frame difference method; the real-time attitude of the gimbal is time-aligned with the on-site execution command using the timestamp matching method to obtain synchronous comparison data; the difference between the real-time attitude of the gimbal and the on-site execution command in the synchronous comparison data is calculated using the differential calculation method to obtain the gimbal pointing deviation; the difference between the audio-visual output intensity in the synchronous comparison data and the on-site execution command is calculated using the differential calculation method to obtain the audio-visual intensity deviation; the distance between the target position in the synchronous comparison data and the on-site execution command is calculated using the Euclidean distance calculation method to obtain the position deviation; the gimbal pointing deviation and position deviation are weighted and summed based on historical expulsion effectiveness data using the weighted summation method to obtain the pointing correction amount; and the audio-visual intensity deviation is used as the intensity adjustment amount using the differential calculation method. It should be noted that historical eviction validity data is obtained through persistently stored strategies and archives; The gimbal angle and output energy level are dynamically corrected by using pointing correction and intensity adjustment, and video footage and environmental data are collected simultaneously to generate real-time performance records and adjustment trajectories. Furthermore, the gimbal angle is adjusted according to the pointing correction amount, and the device output energy level is adjusted according to the intensity adjustment amount; the camera is simultaneously triggered to collect video images and the environmental sensor collects environmental data through a timestamp matching method; missing values ​​in the synchronously collected video images and environmental data are processed through linear interpolation; and the processed video images, environmental data, pointing correction amount, and intensity adjustment amount are integrated by a timestamp integration method to generate a real-time performance record and adjustment trajectory.

[0029] S5. Based on real-time performance records and adjustment trajectories, determine the effectiveness of the non-intrusive priority strategy, update the interference risk profile, and generate updated strategies and profiles. By comparing and analyzing the interference process based on real-time effectiveness records and adjusted trajectories, the similarity matching degree and the effectiveness of the expulsion can be obtained. Furthermore, moving targets are detected from the video footage of real-time performance records using the frame difference method to obtain bird activity sign sequences; the bird activity sign sequences are then time-aligned with historical successful expulsion sequences in the persistently stored strategy and archive using the timestamp matching method to obtain aligned activity data; the similarity between the aligned activity data and the historical successful expulsion sequences is calculated using the Euclidean distance method to obtain the expulsion effectiveness; the real-time performance records are then time-aligned with the on-site execution instructions using the timestamp matching method to obtain aligned execution data; and the similarity between the actual sequence and the expected sequence in the aligned execution data is calculated using the Euclidean distance method to obtain the similarity matching degree. The non-intrusion priority strategy is comprehensively judged based on similarity matching degree and expulsion effectiveness, and a conclusion on the effectiveness of the strategy is generated. Furthermore, when comprehensively judging the non-intrusive priority strategy based on similarity matching degree and expulsion effectiveness, firstly, the similarity between the bird activity sign sequence in the real-time performance record and the historical successful expulsion sequence is compared using the Euclidean distance calculation method to obtain the expulsion effectiveness score; then, the timestamp matching method is used to align the real-time performance record with the expected execution data in the on-site execution instructions, and the similarity between the actual sequence and the expected sequence is calculated to obtain the similarity matching degree score; finally, the similarity matching degree score and the expulsion effectiveness score are used as the conclusion of the strategy effectiveness. The effectiveness conclusions of the strategy are used to compare and adjust the weights of each item in the interference risk file to obtain updated elements. Furthermore, the strategy effectiveness conclusions are time-aligned with each entry in the interference risk profile using a timestamp matching method. The difference between the expulsion effectiveness score in the strategy effectiveness conclusions and the probability assessment score in the entries is calculated using a differential calculation method. The difference is then combined with the non-intrusion priority score of the entries using a weighted summation method to calculate a new priority score. The updated elements are obtained by integrating the new priority scores of all entries. The non-intrusive priority strategy is integrated and revised using updated elements to generate updated strategies and files, which are then persistently stored. Furthermore, the updated elements are time-aligned with the new priority score and non-intrusive priority strategy using the timestamp matching method to obtain the alignment strategy. The alignment strategy is then vector-distance calculated using the Euclidean distance method to obtain the integration distance. The integration distance is then difference-calculated using the difference method to obtain the revision difference. The missing parts in the revision difference are filled using the linear interpolation method to obtain the complete revision sequence. The complete revision sequence is average-biased using the sliding window method to obtain the revision stability sequence. The revision stability sequence is then difference-calculated using the frame difference method to obtain the revision adjustment sequence. The revision adjustment sequence is then weighted-summed using the weighted sum method to obtain the revision value. The revision value is then grouped into revision groups using the mean clustering method based on similar patterns. The revision groups are then integrated with the interference risk profile using the timestamp matching method to generate an updated strategy and profile including historical gimbal pointing sequences, spatial location risk sequences, cruise path data, and historical mild interference duration sequences, which are then persistently stored.

[0030] S6. Based on the updated strategy and files, set up cruise and mild interference windows, configure grade levels and pointing sequences, and form the next cycle prevention plan and equipment instruction set.

[0031] Based on the updated strategy and files, the cruise path and duration of minor interference are preset, and cruise and minor interference windows are generated. Furthermore, cruise path data is extracted from the updated strategy and archives using a timestamp matching method to obtain the cruise path; the duration of mild interference is obtained by averaging the historical mild interference duration sequence in the updated strategy and archives using a sliding window method; and the cruise path and mild interference duration are integrated using a timestamp matching method to generate a cruise and mild interference window. Based on the cruise and light interference windows, the target area is correlated to obtain the grade level and corresponding pointing range; Furthermore, the distance between the cruise path sequence and the spatial location risk sequence in the cruise and light interference windows is calculated using the Euclidean distance method to obtain the risk distance sequence; the regional risk score is calculated by combining the risk distance sequence and the light interference duration sequence using the weighted summation method; the regional risk score is grouped into similar patterns using the mean clustering method to obtain the grade level; the historical gimbal pointing sequence is extracted from the updated strategy and archives using the timestamp matching method, and the grade level is matched with the historical gimbal pointing sequence to obtain the corresponding pointing range (exemplary value range: azimuth 0 degrees to 30 degrees and elevation 10 degrees to 20 degrees). In contrast, existing bird deterrence methods often involve reactive, strong-interference triggers. This step, based on the updated strategy profile, uses a sliding window method to preset the duration of mild interference. It combines the risk score calculated by Euclidean distance to generate graded levels and gimbal pointing ranges, enabling dynamic configuration of equipment combinations. By intervening at low intensity, the risk is mitigated before birds gather, thus protecting bird habitat behavior. Based on the grade level and the target range, the equipment is combined and configured to form the next cycle prevention plan and equipment instruction set; Furthermore, based on the grade level and the pointing range, the grade level and the pointing range are time-aligned using a timestamp matching method to obtain the aligned grade level. The aligned grade level is then vector-distance calculated using Euclidean distance to obtain the combined distance. The combined distance is then difference-calculated using a difference method to obtain the configuration difference. The configuration difference is then filled with gaps using a linear interpolation method to obtain the complete configuration sequence. The complete configuration sequence is then average-deviation-calculated using a sliding window method to obtain the configuration stability sequence. The difference in the configuration stability sequence is then calculated using a frame difference method to obtain the configuration adjustment sequence. The configuration adjustment sequence is then summed using a weighted summation method to obtain the configuration value. The configuration value is then grouped into similar patterns using a mean clustering method to obtain the device combination. Finally, the device combination is integrated with the cruise and mild interference windows using a timestamp matching method to form the next cycle prevention plan and device instruction set.

[0032] In summary, this invention achieves dynamic identification and risk classification of bird activity patterns within a target area by generating a disturbance risk profile through spatiotemporal prediction based on a risk cue set and behavioral learning classification. This provides continuous predictive basis for subsequent protection decisions, enabling the early construction of a protection response framework in the temporal dimension. By employing a non-intrusive priority strategy to map the disturbance risk profile to specific execution sequences and determine on-site execution instructions, a balance between protection intensity and bird sensitivity is achieved, reducing ecological disturbance while maintaining the continuity of protection effectiveness. This results in a comprehensive benefit of early warning, dynamic adjustment, and eco-friendliness.

[0033] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting and protecting against bird disturbances based on big data analysis, characterized in that: include, Continuously collect video footage and environmental data of the target area and integrate them according to timestamps to obtain a set of risk clues; Based on the risk cues set, spatiotemporal predictions are made and behavioral learning classification is completed to generate interference risk profiles. A non-intrusive priority strategy is adopted to map interference risk profiles to specific execution sequences to determine on-site execution instructions; Execute on-site instructions in the target area, dynamically correct the gimbal's direction and intensity, and collect on-site feedback data in real time to obtain real-time results records and adjust the trajectory; The effectiveness of the non-intrusive priority strategy is determined based on real-time performance records and adjustment trajectories, and the interference risk profile is updated to generate updated strategies and profiles. Based on the updated strategy and files, set up cruise and mild interference windows, configure grade levels and direction sequences, and form the next cycle prevention plan and equipment instruction set.

2. The bird disturbance prediction and protection method based on big data analysis as described in claim 1, characterized in that: The process of continuously collecting video footage and environmental data of the target area and integrating them according to timestamps to obtain a risk clue set involves the following steps: The system continuously collects video footage and environmental data of the target area using a cruise-style framing method. The video footage and environmental data are preprocessed and integrated according to timestamps to obtain a set of risk clues.

3. The bird disturbance prediction and protection method based on big data analysis as described in claim 2, characterized in that: The specific steps for generating an interference risk profile by performing spatiotemporal prediction based on the risk clue set and completing behavioral learning classification are as follows. Spatiotemporal correlation and coherence checks were performed on risk cues within the same time window to obtain sequences of bird activity signs and environmental characteristics. Based on the comparison and difference annotation of similar scenarios using bird activity sign sequences and environmental feature sequences, a candidate list of regions is generated. The candidate regional list is assessed for probability and non-intrusive priorities are determined, and then integrated into an interference risk profile.

4. The bird disturbance prediction and protection method based on big data analysis as described in claim 3, characterized in that: The non-intrusive priority strategy is used to map interference risk profiles to specific execution sequences to determine on-site execution instructions. The specific steps are as follows. A non-intrusive priority strategy is adopted to perform non-intrusive priority expansion on each record in the interference risk file to obtain an initial strategy sequence; The initial strategy sequence is matched sequentially based on the rhythm of sound wave frequency changes, light intensity level and gimbal pointing to generate an execution orchestration list; The feasibility of the execution schedule is checked and refined to form on-site execution instructions.

5. The bird disturbance prediction and protection method based on big data analysis as described in claim 4, characterized in that: The execution of on-site execution instructions in the target area refers to sending on-site execution instructions to the equipment in the target area and executing the on-site execution instructions.

6. The bird disturbance prediction and protection method based on big data analysis as described in claim 5, characterized in that: The process of dynamically correcting the gimbal's direction and intensity, and collecting real-time feedback data from the field to obtain real-time performance records and adjustment trajectories, involves the following specific steps. Based on the on-site execution commands, the gimbal's real-time attitude, audio-visual output intensity and target position are synchronously compared and deviation analyzed to obtain the pointing correction amount and intensity adjustment amount; The gimbal angle and output energy level are dynamically corrected by using pointing correction and intensity adjustment, and video footage and environmental data are collected simultaneously to generate real-time performance records and adjustment trajectories.

7. The bird disturbance prediction and protection method based on big data analysis as described in claim 6, characterized in that: The specific steps for determining the effectiveness of the non-intrusive priority strategy based on real-time performance records and adjustment trajectories are as follows. By comparing and analyzing the interference process based on real-time effectiveness records and adjusted trajectories, the similarity matching degree and the effectiveness of the expulsion can be obtained. The non-intrusion priority strategy is comprehensively judged based on similarity matching degree and expulsion effectiveness, and a conclusion on the effectiveness of the strategy is generated.

8. The bird disturbance prediction and protection method based on big data analysis as described in claim 7, characterized in that: The process of updating the interference risk profile and generating updated strategies and profiles involves the following steps. The effectiveness conclusions of the strategy are used to compare and adjust the weights of each item in the interference risk file to obtain updated elements. The updated elements are used to integrate and revise the non-intrusive priority strategy, generating updated strategies and files, which are then persistently stored.

9. The bird disturbance prediction and protection method based on big data analysis as described in claim 8, characterized in that: The setting of cruise and mild interference windows based on the updated strategy and files refers to preset the cruise path and mild interference duration based on the updated strategy and files, and generating cruise and mild interference windows.

10. The bird disturbance prediction and protection method based on big data analysis as described in claim 9, characterized in that: The configuration of hierarchical levels and instruction sequences forms the next cycle's prevention plan and equipment instruction set. The specific steps are as follows: Based on the cruise and light interference windows, the target area is correlated to obtain the grade level and corresponding pointing range; Based on the grade level and target range, the equipment is combined and configured to form the next cycle prevention plan and equipment instruction set.

Images