Bird repelling device with self-recognition function

Abstract

The invention relates to the technical field of bird repelling devices, in particular to a bird repelling device with a self-recognition function, and the system comprises a target extraction module, a maloperation screening module, a response pause module, an interference classification module and a recognition rule updating module. According to the method, the change combination of multi-channel signals is constructed, time alignment is realized, collaborative features of image brightness edges, audio dominant frequency intervals and infrared heat source distribution are extracted, inconsistent combination contents are identified and screened out, and behavior validity is judged according to interruption and abrupt change signals. Frequently appearing similar disturbances are classified and calibrated, an interference characteristic pattern is formed, edge judgment and a time parameter range are adjusted in combination with identification deviation, a response coordination mechanism among multiple channels is further improved, and the device has dynamic adaptive capacity aiming at different interference modes; the judgment accuracy in the target extraction process and the stability of the expelling response are enhanced, and continuous optimization of recognition and reaction logic in a complex scene is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bird repelling devices, in particular to a bird repelling device with self-recognition function. BACKGROUND

[0002] The technical field of bird repelling devices involves monitoring and controlling bird activities, aiming to reduce the harm caused by birds to human facilities, especially power systems, airports, farmlands and other key areas. This technical field covers various means for bird identification and repelling, including physical barrier setting, sound interference, optical stimulation, electromagnetic repulsion, and intelligent bird repelling devices combining image, sound, infrared and other multi-source information. With the development of artificial intelligence, multi-modal recognition and edge computing, the bird repelling system gradually changes from traditional single repelling means to a complex system with identification, judgment and response capabilities, forming a comprehensive prevention and control technology system integrating identification, judgment, repelling and adaptability optimization. Among them, the traditional bird repelling device refers to a system that mainly relies on physical devices such as bird spikes, bird nets or devices based on single modal such as visual sensors or sound sensors, passively intervenes or implements fixed mode repelling on birds. This kind of system usually identifies targets through fixed cameras or implements bird repelling through continuous emission of ultrasonic waves, biological enemy sounds, etc. Image recognition mostly uses fixed threshold segmentation method or classification method based on traditional feature extraction, sound recognition relies on spectrum analysis and template matching method, and target detection often uses infrared induction single trigger without fusing multi-source features, which cannot accurately judge the bird species, behavior state or its relative position with the device, and lacks data back transmission and self-learning mechanism, resulting in low recognition accuracy in variable environment, single intervention strategy and lack of differentiated repelling response to different birds.

[0003] Traditional bird repelling devices rely on single sensing channel for information perception, and the recognition logic is mainly based on static threshold or template matching, lacking the processing ability of time sequence association between image, audio and infrared signals, which cannot extract target features with correlation in multi-signal interactive change scene, leading to difficulty in associating and identifying abnormal fluctuations between different channels, and further forming problems such as false triggering or response omission. In addition, such systems cannot perform clustering analysis and attribution judgment on continuously appearing non-target interference features, lack the comparison mechanism between behavior state and response condition, and are difficult to correct recognition rules and response boundaries in time when signal interruption, mutation or interference occurs repeatedly, causing imbalance in repelling instruction execution, lag in adaptive adjustment and repeated response to invalid stimulation. SUMMARY

[0004] In order to solve the technical problems existing in the prior art, the embodiments of the present application provide a bird repelling device with self-recognition function. On the one hand, a bird repelling device with self-recognition function is provided, which comprises: The target extraction module collects image signals, audio signals, and infrared signals generated by the camera, microphone, and thermal imager, and extracts edge brightness changes in the image signals, main frequency range changes in the audio signals, and heat source area distribution changes in the infrared signals to construct channel change combinations. The false triggering filtering module analyzes the channel change combinations and filters out feature combinations including edge interruption in image signals, abnormal main frequency in audio signals and heat source drift in infrared signals, and removes false triggering combination information with a consistency matching coefficient lower than the preset consistency judgment benchmark value. The response pause module, based on the edge interruption and main frequency change signals in the false trigger combination information, combined with the interruption time and action response window, determines whether there is a lack of target support and generates a pause command. The interference classification module extracts edge interruption and main frequency change signals from the pause command, calculates the image change frequency and audio change duration, classifies time segments with the same combination pattern, and constructs a list of interference trigger features. The recognition rule update module analyzes the image frequency range and audio duration range in the interference trigger feature list, adjusts the edge judgment conditions of the camera channel and the audio frequency detection range of the microphone channel, and generates the recognition execution content of the bird deterrent device.

[0005] As a further aspect of the present invention, the channel change combination includes the image edge brightness change amplitude, the audio main frequency range change range, and the infrared heat source area distribution difference; the false trigger combination information includes edge brightness interruption feature parameters, main frequency range abnormal fluctuation values, and heat source area drift offset; the pause command includes image edge interruption signal features, audio main frequency sudden change signal features, and response interruption time stamp; the interference trigger feature list includes image change frequency range values, audio burst duration length, and combination mode occurrence frequency; and the bird deterrent device identifies and executes content including edge trigger response threshold, audio detection time window range, and channel parameter adjustment configuration.

[0006] As a further aspect of the present invention, the failure to meet multi-channel consistency means that the feature information extracted from the sensor channel image, audio, and infrared does not have synchronicity in terms of time, space, and change trend, and the correlation coefficient of the change trend is lower than a preset threshold.

[0007] As a further aspect of the present invention, the determination of whether there is a lack of target support refers to the failure of edge interruption and main frequency change to continuously correspond to effective target actions within the target time window.

[0008] As a further aspect of the present invention, the target extraction module includes: The data synchronization receiving submodule acquires camera image frames, microphone audio waveforms and thermal imager infrared thermal images, performs timestamp alignment processing on the three types of signals, unifies the time axis index, and generates a time-aligned signal matrix. Based on the time-aligned signal matrix, the signal feature analysis submodule extracts the image frame edge brightness change value, the audio main frequency range change amplitude and the heat source region distribution change rate of the heat map, and generates the image brightness change factor, audio frequency change factor and heat source distribution change factor. The channel variation combination construction submodule calls the image brightness variation factor, the audio frequency variation factor, and the heat source distribution variation factor to perform combination analysis on a unified time axis, obtain the interactive characteristics of channel variation, and generate channel variation combinations.

[0009] As a further aspect of the present invention, the erroneous screening module includes: Based on the channel change combination, the anomaly feature detection submodule identifies interrupted segments in the image edge brightness change sequence, calculates the deviation range of the audio main frequency interval within the time window, detects the positional shift of the infrared heat source region between consecutive frames, and generates a set of channel anomaly feature indicators. The multi-channel consistency judgment submodule calls the set of channel abnormal feature indicators, performs joint comparison of abnormal changes in image, audio and infrared channels on the same time axis, determines the degree of matching between channels within the time segment based on the channel consistency judgment condition threshold, and generates a multi-channel consistency matching coefficient sequence. The false trigger information filtering submodule filters out combination records with consistency coefficients lower than the consistency judgment benchmark value based on the multi-channel consistency matching coefficient sequence, extracts the time period and feature information corresponding to the non-consistent events, and obtains false trigger combination information.

[0010] As a further aspect of the present invention, the response pause module includes: Based on the false triggering combination information, the signal interruption judgment submodule extracts the edge brightness interruption time point in the image channel, detects the time segment of the main frequency change in the audio channel, determines whether there is a target support missing situation in the channel, and generates a time index sequence without target support. The response window comparison submodule calls the targetless support time index sequence, matches the erroneous trigger time segment with the current action response window one by one, filters the time overlap area, calculates the time overlap rate and concurrent duration of signal missing and response behavior in the area, and obtains the behavior interference conflict interval table. The instruction generation and output submodule extracts the conflict period between signal interruption and behavior response based on the behavior interference conflict interval table, converts the conflict period into control signal output content, sets the status of the pause instruction control flag, and generates a pause instruction.

[0011] As a further aspect of the present invention, the interference classification module includes: The image frequency extraction submodule extracts the inter-frame change time interval of the image based on the image edge interruption signal recorded in the pause instruction, calculates the number of edge state transitions within the time interval, forms an image change rate curve, and obtains the image change frequency interval value. The audio duration calculation submodule calls the corresponding audio change signal in the pause instruction, marks the start and end time points of the change, calculates the continuous duration, and statistically analyzes the distribution range of the duration of similar events to obtain the audio change duration interval value. The combination mode classification submodule classifies time segments with the same frequency range and the duration range of audio changes into the same type based on the frequency range of the image changes and the duration range of the audio changes. It also organizes the triggering time period and structure of each type of combination and establishes a list of interference triggering features.

[0012] As a further aspect of the present invention, the identification rule update module includes: The image threshold adjustment submodule, based on the image change frequency range value in the interference trigger feature list, calls the existing edge trigger condition parameters in the camera channel, resets the response judgment threshold corresponding to the frequency range, and overwrites the adjusted parameters according to the frame sequence index to generate an image edge response judgment parameter set. The audio range revision submodule compares the audio burst duration interval value in the interference trigger feature list with the original audio frequency detection time threshold of the pickup channel, corrects the detection window boundary according to the upper and lower limits of the time interval, updates the event trigger time range configuration, and obtains the audio time judgment parameter set. The recognition instruction generation submodule calls the image edge response judgment parameter set and the audio time judgment parameter set, sorts out the applicable range of the two types of parameters in the response stage of the bird deterrence device, and writes them into the control instruction in a formatted manner according to the channel logic mapping relationship to generate a bird deterrence recognition execution configuration instruction set.

[0013] As a further aspect of the present invention, the response determination threshold refers to a boundary value used to determine whether the sensing signal has reached the triggering condition, and to decide whether to respond to the target change.

[0014] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following: By constructing a combination of changes in multi-channel signals and achieving time alignment, the system extracts the collaborative features of image brightness edges, audio frequency ranges, and infrared heat source distribution. It identifies and filters out inconsistent combinations, judges the validity of behavior based on interruption and abrupt change signals, classifies and calibrates frequently occurring similar disturbances to form interference feature patterns, and adjusts the edge judgment and time parameter range based on recognition deviations. This further improves the response coordination mechanism between multiple channels, enabling the device to have dynamic adaptability to different interference modes, enhancing the accuracy of judgment and the stability of the drive-away response during the target extraction process, and achieving continuous optimization of recognition and reaction logic in complex scenarios. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0016] Figure 1 This is an overall schematic diagram of the invention; Figure 2 This is a schematic diagram of the framework of the present invention; Figure 3 This is a flowchart of the target extraction module in this invention; Figure 4 This is a flowchart of the erroneous screening module in this invention; Figure 5 This is a flowchart of the response pause module in this invention; Figure 6 This is a flowchart of the interference classification module in this invention; Figure 7 This is a flowchart of the identification rule update module in this invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] This invention provides a bird deterrent device with self-identification function, such as... Figures 1-2 The diagram shown illustrates a bird deterrent device with self-identification capabilities. The system includes: The target extraction module collects images, audio, and infrared signals generated by the camera, microphone, and thermal imager in the bird deterrence device. It extracts changes in image edge brightness, audio frequency range, and infrared heat source distribution. It performs time alignment processing on the changes in images, audio, and infrared signals within the same time segment and constructs channel change combinations. The false triggering removal module is based on channel change combinations to screen for combinations of features such as interrupted brightness changes at the image edge, audio main frequency range deviating from the normal range, and drifting phenomenon in the infrared heat source area. It removes combinations that do not meet the multi-channel consistency condition and filters out false triggering combination information. The response pause module determines whether the current behavior triggering condition lacks target support based on the image edge interruption and audio main frequency change signal in the false triggering combination information, and compares and analyzes the signal interruption time point with the action response window to generate a pause command; The interference classification module extracts the feature intervals of image change frequency and audio burst duration based on the combination of image edge interruption signals and audio abrupt change signals recorded in the pause command, classifies the time segments in which the same combination pattern appears, and constructs an interference trigger feature list. The recognition rule update module adjusts the response judgment conditions for edge triggering in the camera channel and the time judgment range for audio detection in the microphone channel based on the image change frequency range and audio burst duration range in the interference trigger feature list, thereby generating the recognition execution content of the bird deterrent device.

[0023] The channel variation combination includes the amplitude of image edge brightness variation, the range of audio main frequency interval variation, and the distribution difference of infrared heat source area. The false trigger combination information includes edge brightness interruption feature parameters, abnormal fluctuation value of main frequency interval, and heat source area drift offset. The pause command includes image edge interruption signal characteristics, audio main frequency sudden change signal characteristics, and response interruption time label. The interference trigger feature list includes image change frequency interval value, audio burst duration length, and frequency of occurrence of combination mode. The bird deterrent device identification execution content includes edge trigger response threshold, audio detection time window range, and channel parameter adjustment configuration.

[0024] Specifically, such as Figure 2 , 3 As shown, the target extraction module includes: The data synchronization receiving submodule acquires camera image frames, microphone audio waveforms and thermal imager infrared thermal images, performs timestamp alignment processing on the three types of signals, unifies the time axis index, and generates a time-aligned signal matrix. First, a multi-dimensional sensor array deployed in the core monitoring area of ​​the substation is activated. This array consists of a high-definition visible light camera, a high-sensitivity microphone, and an infrared thermal imager. The high-definition visible light camera is configured to capture images at a rate of 25 frames per second, with a resolution of 1920 pixels by 1080 pixels, meaning the theoretical time interval between each frame is 40 milliseconds. The microphone is set to a sampling rate of 44.1 kHz, continuously acquiring ambient sound waves. The thermal imager outputs 320-pixel by 240-pixel infrared thermal images at a rate of 9 frames per second, with a frame interval of approximately 111.11 milliseconds. To address the issue of inconsistent sampling frequencies among heterogeneous sensors, the data synchronization receiving submodule introduces the Network Time Protocol (NTP) as a unified clock source, performing microsecond-level timestamp calibration on each sensor every 60 seconds. During data acquisition, the module establishes a unified reference time axis with a time granularity set to 100 milliseconds (i.e., 10 synchronization points per second). The module performs strict alignment operations: for image data, it reads the timestamp at the beginning of each frame and selects the frame with the smallest absolute deviation from the reference time point; for audio data, it slices the continuous waveform stream in 100-millisecond intervals, with each slice containing 4410 sampling points (i.e., 44100 Hz multiplied by 0.1 seconds); for thermal infrared data, it uses nearest-neighbor interpolation to map the thermal image frames to this time scale. For example, at the reference time point 14:00:00:100, the system finds the nearest image frame with a deviation of +2 milliseconds, the audio slice time is strictly aligned, and the nearest thermal image frame has a deviation of +11 milliseconds (because a 111-millisecond frame period corresponds to a 100-millisecond scale). The module encapsulates the processed image frame buffer address, audio waveform slice, and infrared thermal image data to generate a time-aligned signal matrix. In the records of this matrix, the synchronization errors of the three types of signals in the time dimension are all controlled within the system's allowable range, ensuring the consistency of the physical reference for subsequent fusion analysis.

[0025] The signal feature analysis submodule extracts the image frame edge brightness change value, the audio main frequency range change amplitude and the heat source region distribution change rate based on the time-aligned signal matrix, and generates the image brightness change factor, audio frequency change factor and heat source distribution change factor. The module reads data line by line and extracts physical features. For image data, the module first converts the color image into an 8-bit grayscale image and uses the Sobel operator to perform pixel-level difference operations between the current frame and the previous frame. The system sets the grayscale change threshold to 15 (approximately 5.8% of 255 grayscale levels) and counts the total number of pixels in the difference image whose grayscale value changes exceed this threshold. In this embodiment, the system detects 103,680 pixels with significant changes. The module divides this number of changed pixels by the total number of pixels (1920 multiplied by 1080 equals 2073600) to calculate a normalized image brightness change factor of 0.05. For audio data, the module performs a Fast Fourier Transform (FFT) on a 100-millisecond audio slice, setting the FFT point count to 4096 and the frequency resolution to approximately 10.7 Hz. The module extracts the highest-energy frequency peak in the spectrum and calculates the absolute value of the difference between it and the dominant frequency of the previous time slice. Assuming the current slice's dominant frequency abruptly changes to 50 Hz (a typical power frequency interference frequency), while the previous moment was a silent state with a dominant frequency of 0 Hz, the calculated audio frequency change factor is 50 Hz. For thermal imaging data, the module uses Otsu's method to perform threshold segmentation on the heat map to extract the centroid of the high-temperature region and calculates the Euclidean distance between this centroid and the previous frame. Assuming the centroid coordinates move from (100, 100) to (100, 160), the calculated heat source distribution change factor is 60 pixels. Finally, the module outputs a feature vector containing the values ​​0.05, 50 Hz, and 60 pixels.

[0026] The channel variation combination construction submodule calls the image brightness variation factor, audio frequency variation factor, and heat source distribution variation factor to perform combination analysis on a unified time axis, obtain the interactive characteristics of channel variation, and generate channel variation combinations. The module calls upon the three variation factors mentioned above and performs standardized combined analysis on a unified time axis. To eliminate the influence of different physical dimensions, the module presets the maximum variation amplitude benchmarks for each sensor: the maximum variation ratio for the image is 0.5 (i.e., half-screen variation), the maximum frequency jump for the audio is 1000 Hz, and the maximum displacement of the heat source is 100 pixels. The module performs normalization calculations: for the image channel, 0.05 is divided by 0.5, resulting in a normalized value of 0.1; for the audio channel, 50 is divided by 1000, resulting in a normalized value of 0.05; for the heat source channel, 60 is divided by 100, resulting in a normalized value of 0.6. The module combines these three values ​​to generate a normalized channel state vector (0.1, 0.05, 0.6). This vector quantitatively reveals the non-equilibrium characteristics at the current moment: the variation amplitude of the heat source channel is significant (0.6), while the variation amplitudes of the image and audio channels are relatively low. This combination of features provides a data basis for subsequent consistency determination.

[0027] Specifically, such as Figure 2 , 4 As shown, the erroneous screening module includes: The anomaly detection submodule identifies interrupted segments in the image edge brightness change sequence based on channel change combinations, calculates the deviation range of the audio main frequency interval within the time window, detects the positional shift of the infrared heat source region between consecutive frames, and generates a set of channel anomaly feature indicators. Identify potential logical conflicts. For the image channel, the module scans the brightness change sequence. Although the detected change factor is 0.1, in this embodiment, the system initially determines that it has detected "valid motion," and therefore marks the image state as "normal" (logic value recorded as 1). For the audio channel, the module sets the effective biological sound source frequency range to 1000 Hz to 6000 Hz. Since the currently detected frequency is 50 Hz, this value falls outside the effective range and conforms to the characteristics of power frequency noise. Therefore, the module determines the audio feature as "abnormal," and the state is marked with a logic value of 0. For the infrared channel, the module detects a heat source displacement of 60 pixels. The system sets the maximum inter-frame displacement threshold of biological targets within 100 milliseconds to 50 pixels (exceeding this value usually means non-biological mechanical motion or sensor noise). Since 60 is greater than 50, the module determines the heat source movement feature as "abnormal," and the state is marked with a logic value of 0. Finally, the module generates a set of abnormal feature indicators for the channels, which are: image channel normal (1), audio channel abnormal (0), and heat source channel abnormal (0).

[0028] The multi-channel consistency judgment submodule calls the channel abnormal feature index set, performs joint comparison of abnormal changes in image, audio and infrared channels on the same time axis, determines the degree of matching between channels within the time segment based on the channel consistency judgment condition threshold, and generates a multi-channel consistency matching coefficient sequence. The system invokes a set of abnormal feature indicators for each channel to perform a joint weighted score on the effectiveness of each channel. The system presets the confidence weights for each channel in the comprehensive judgment: image channel weight is 0.4, audio channel weight is 0.4, and infrared thermal image channel weight is 0.2, with a sum of 1.0. The module calculates the multi-channel consistency matching coefficient using a weighted summation formula: 0.4 multiplied by 1 (image state), plus 0.4 multiplied by 0 (audio state), plus 0.2 multiplied by 0 (thermal source state). The final calculated matching coefficient is 0.4. This coefficient of 0.4 is far lower than the ideal consistency value of 1.0, intuitively reflecting the serious conflicts between the current multi-source sensor data, indicating extremely low reliability of the system's perception and suggesting a potential risk of false triggering.

[0029] The false trigger information filtering submodule filters out combination records with consistency coefficients lower than the consistency judgment benchmark value based on the multi-channel consistency matching coefficient sequence, extracts the time period and feature information corresponding to the non-consistent events, and obtains false trigger combination information; False alarm filtering is performed based on the multi-channel consistency matching coefficient sequence. The module sets the consistency judgment benchmark value to 0.55, which is the optimal segmentation point selected based on ROC curve analysis of a large number of historical samples. The module traverses the sequence and determines that the matching coefficient of the current moment is 0.4, which is less than the benchmark value of 0.55. Based on this, the system classifies this time segment (e.g., 14:05:00 to 14:05:05) as a "false trigger event". The module extracts the original feature data within this time segment, namely the slight flicker of the image, the 50 Hz noise of the audio, and the abnormal jump of the heat source, encapsulates them into false trigger combination information, and immediately blocks the alarm request for this time segment to prevent the system from performing erroneous expulsion actions.

[0030] Specifically, such as Figure 2 , 5 As shown, the response pause module includes: The signal interruption judgment submodule extracts the edge brightness interruption time point in the image channel based on the false trigger combination information, detects the time segment of the main frequency change in the audio channel, determines whether there is a missing target support in the channel, and generates a time index sequence without target support. Based on the false triggering information, the physical root causes of the low confidence level were analyzed in depth. The module executed the following logic: Although the image channel status was marked as 1, the audio and heat source channels were both marked as 0, and the overall consistency score did not meet the standard, indicating that the system lacked effective multidimensional data support during this period. Specifically, the module detected that the audio channel was in the "invalid spectrum region" (50 Hz), which is equivalent to the interruption of the effective acoustic signal; the heat source channel was in the "non-biological motion region", which is equivalent to the interruption of the effective thermal signal. The module determined that this constituted a "target support missing" state and generated a time index sequence of no target support for the identified invalid period (14:05:00 to 14:05:05), clearly identifying that the system was in a blind spot of perception capability.

[0031] The response window comparison submodule calls the time index sequence without target support, matches the erroneous trigger time segment with the current action response window one by one, filters the time overlap area, calculates the time overlap rate and concurrent duration of signal missing and response behavior in the area, and obtains the behavior interference conflict interval table. The module reads the preset action plan of the current bird deterrent device. Assume the system originally planned to activate the loud bird deterrent (i.e., the action response window) between 14:05:00 and 14:05:10. The module performs a time-axis overlap comparison between the target-free time series (14:05:00 to 14:05:05) and this action response window. The comparison results show a 5-second complete overlap. This means that during the first 5 seconds of action execution, the system will completely lack closed-loop support from effective sensor data. The module defines this overlapping time period as the behavioral interference conflict interval and records the conflict type as "multi-dimensional feature verification failed."

[0032] The instruction generation and output submodule extracts the conflict period between signal interruption and behavior response based on the behavior interference conflict interval table, converts the conflict period into control signal output content, sets the status of the pause instruction control flag, and generates a pause instruction. A suppression control strategy is generated for the identified conflict period (14:05:00 to 14:05:05). The module converts this period into a specific pause command, sets the pause control flag to "valid" (logic 1), and calculates the required pause duration as 5000 milliseconds. The module generates a hexadecimal command packet according to the industrial control protocol, for example: AA55031405001388FF, where "1388" represents 5000 milliseconds in decimal. This command is sent to the controller, forcing the device to remain silent during the conflict period, avoiding blind expulsion in the event of sensor failure.

[0033] Specifically, such as Figure 2 , 6 As shown, the interference classification module includes: The image frequency extraction submodule extracts the inter-frame change time interval of the image based on the image edge interruption signal recorded in the pause instruction, calculates the number of edge state transitions within the time interval, forms the image change rate curve, and obtains the image change frequency interval value. Frequency domain backtracking analysis was performed on the image data that caused the false triggering. The module extracted the time series of edge brightness changes between image frames during the specified period (14:05:00 to 14:05:01). The system statistically determined that the overall image brightness underwent 100 "bright-dark" alternations within a 1-second sampling window. By calculating (100 times divided by 2), the fundamental frequency of the image change was found to be 50 Hz. This frequency characteristic perfectly matches the flicker characteristics of an AC-powered artificial light source. Based on this statistical result, the module determined the image change frequency range to be 48 Hz to 52 Hz. This range covers the normal range of power grid frequency fluctuations, quantifying the physical characteristics of the visual interference source.

[0034] The audio duration calculation submodule calls the corresponding audio change signal in the pause command, marks the start and end time points of the change, calculates the continuous duration, and statistically analyzes the distribution range of the duration of similar events to obtain the range value of the audio change duration. The system performs time-domain length statistics on abnormal audio segments. It employs the short-time energy method, setting the energy threshold to three times the background noise level (i.e., the three-standard-deviation principle). The system detects a sudden burst of impulse noise with a start time of 100 milliseconds and an end time of 350 milliseconds. Calculation (350 minus 100) yields a continuous duration of 250 milliseconds for this single noise event. Further analysis of historical data reveals that the duration of such noise (e.g., the sound of a mechanical switch striking) has a 95% probability of falling between 200 and 300 milliseconds. Based on this, the system establishes this statistical range as the audio variation duration interval, i.e., 200 to 300 milliseconds, providing precise time parameters for subsequent filter design.

[0035] The combination mode classification submodule classifies time segments with the same frequency range and audio duration range based on the frequency range of image changes and the duration range of audio changes into the same type, organizes the trigger time period and structure of each type of combination, and establishes a list of interference trigger features. Pattern matching is performed based on the extracted physical features. The module combines the features "image change frequency between 48 and 52 Hz" and "audio duration between 200 and 300 milliseconds" and classifies them as "power frequency flicker accompanied by short-term mechanical noise" interference mode (ID: INT_MODE_01). The system establishes a list of interference triggering features, clearly recording the typical triggering time of this mode (e.g., 18:00 to 06:00 the next day) and its physical description. This classification enables the system to distinguish between changes in the natural environment and specific man-made electromagnetic or mechanical interference.

[0036] Specifically, such as Figure 2 , 7 As shown, the identification rule update module includes: The image threshold adjustment submodule, based on the image change frequency range value in the interference trigger feature list, calls the existing edge trigger condition parameters in the camera channel, resets the response judgment threshold corresponding to the frequency range, and overwrites the adjusted parameters according to the frame sequence index to generate an image edge response judgment parameter set. The camera channel parameters are adaptively corrected. The module adds a band-stop filter logic to the image processing algorithm, with a center frequency of 50 Hz and a bandwidth of 4 Hz. Simultaneously, for brightness changes falling within the 48 Hz to 52 Hz frequency range, after detecting the brightness change and determining that its frequency continuously and stably falls within this range, the system dynamically raises the trigger response threshold for that frequency band to 255 to mask false alarm responses caused by power frequency flicker. The adjusted parameters are written to the hardware register by frame sequence index, generating a new set of image edge response judgment parameters, thus eliminating false alarms caused by flicker at the source.

[0037] The audio range revision submodule compares the audio burst duration interval value in the interference trigger feature list with the original audio frequency detection time threshold of the microphone channel, corrects the detection window boundary according to the upper and lower limits of the time interval, updates the event trigger time range configuration, and obtains the audio time judgment parameter set. Based on the audio interference duration range (200 to 300 milliseconds), the microphone detection logic was optimized. The original detection logic was set to "trigger if the duration is greater than 100 milliseconds". To effectively avoid the aforementioned mechanical noise, the module revised the detection lower limit: taking the upper limit of the interference range as 300 milliseconds, and increasing the safety margin by 50 milliseconds, raising the minimum trigger duration threshold to 350 milliseconds. This means that the module sets the time-determining band-stop interval based on the interference duration characteristics (concentrated between 200 and 300 milliseconds), shielding audio signals falling within this time period, while retaining abnormal sounds with durations shorter than 200 milliseconds or longer than 300 milliseconds as potential trigger criteria, thereby effectively avoiding short-term mechanical noise interference and avoiding false rejection of genuine bird calls. The module updates the audio time judgment parameter set accordingly, significantly reducing the false alarm rate caused by short-term impulse noise.

[0038] The recognition instruction generation submodule calls the image edge response judgment parameter set and the audio time judgment parameter set, sorts out the applicable range of the two types of parameters in the response stage of the bird deterrence device, and writes them into the control instruction in a formatted manner according to the channel logic mapping relationship to generate the bird deterrence recognition execution configuration instruction set; The above parameters are adjusted and translated into low-level hardware control instructions. The module generates two core configuration instructions: the first is a video filtering instruction, with the code set to 0x32 (representing 50 Hz in decimal), used to activate the notch filter; the second is an audio duration instruction, with the code set to 0x015E (representing 350 milliseconds in decimal), used to set a new duration threshold. The module writes these instructions into the system's non-volatile memory (EEPROM), generating the final bird deterrence recognition execution configuration instruction set. When the system runs in the next cycle or restarts, it will automatically load these configurations, achieving immunity to specific environmental interference and completing a closed-loop control process from fault identification to parameter self-healing.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A bird-repelling device with self-identification function, characterized in that, include: The target extraction module collects image signals, audio signals, and infrared signals generated by the camera, microphone, and thermal imager, and extracts edge brightness changes in the image signals, main frequency range changes in the audio signals, and heat source area distribution changes in the infrared signals to construct channel change combinations. The false triggering filtering module analyzes the channel change combinations and filters out feature combinations including edge interruption in image signals, abnormal main frequency in audio signals and heat source drift in infrared signals, and removes false triggering combination information with a consistency matching coefficient lower than the preset consistency judgment benchmark value. The response pause module, based on the edge interruption and main frequency change signals in the false trigger combination information, combined with the interruption time and action response window, determines whether there is a lack of target support and generates a pause command. The interference classification module extracts edge interruption and main frequency change signals from the pause command, calculates the image change frequency and audio change duration, classifies time segments with the same combination pattern, and constructs a list of interference trigger features. The recognition rule update module analyzes the image frequency range and audio duration range in the interference trigger feature list, adjusts the edge judgment conditions of the camera channel and the audio frequency detection range of the microphone channel, and generates the recognition execution content of the bird deterrent device.

2. The bird deterrent device with self-identification function according to claim 1, characterized in that: The channel change combination includes the image edge brightness change amplitude, the audio main frequency range change range, and the infrared heat source area distribution difference. The false trigger combination information includes edge brightness interruption feature parameters, main frequency range abnormal fluctuation values, and heat source area drift offset. The pause command includes image edge interruption signal features, audio main frequency sudden change signal features, and response interruption time stamp. The interference trigger feature list includes image change frequency range values, audio burst duration length, and combination mode occurrence frequency. The bird deterrent device identifies and executes content including edge trigger response threshold, audio detection time window range, and channel parameter adjustment configuration.

3. The bird deterrent device with self-identification function according to claim 1, characterized in that: The failure to meet multi-channel consistency refers to the fact that the feature information extracted from the sensor channel image, audio, and infrared does not have synchronicity in time and space and the correlation coefficient of the change trend is lower than a preset threshold.

4. The bird deterrent device with self-identification function according to claim 1, characterized in that: The determination of whether there is a lack of target support refers to the failure of edge interruption and main frequency change to continuously correspond to effective target actions within the target time window.

5. The bird deterrent device with self-identification function according to claim 1, characterized in that, The target extraction module includes: The data synchronization receiving submodule acquires camera image frames, microphone audio waveforms and thermal imager infrared thermal images, performs timestamp alignment processing on the three types of signals, unifies the time axis index, and generates a time-aligned signal matrix. Based on the time-aligned signal matrix, the signal feature analysis submodule extracts the image frame edge brightness change value, the audio main frequency range change amplitude and the heat source region distribution change rate of the heat map, and generates the image brightness change factor, audio frequency change factor and heat source distribution change factor. The channel variation combination construction submodule calls the image brightness variation factor, the audio frequency variation factor, and the heat source distribution variation factor to perform combination analysis on a unified time axis, obtain the interactive characteristics of channel variation, and generate channel variation combinations.

6. The bird deterrent device with self-identification function according to claim 1, characterized in that, The erroneous screening module includes: Based on the channel change combination, the anomaly feature detection submodule identifies interrupted segments in the image edge brightness change sequence, calculates the deviation range of the audio main frequency interval within the time window, detects the positional shift of the infrared heat source region between consecutive frames, and generates a set of channel anomaly feature indicators. The multi-channel consistency judgment submodule calls the set of channel abnormal feature indicators, performs joint comparison of abnormal changes in image, audio and infrared channels on the same time axis, determines the degree of matching between channels within the time segment based on the channel consistency judgment condition threshold, and generates a multi-channel consistency matching coefficient sequence. The false trigger information filtering submodule filters out combination records with consistency coefficients lower than the consistency judgment benchmark value based on the multi-channel consistency matching coefficient sequence, extracts the time period and feature information corresponding to the non-consistent events, and obtains false trigger combination information.

7. The bird deterrent device with self-identification function according to claim 1, characterized in that, The response pause module includes: Based on the false triggering combination information, the signal interruption judgment submodule extracts the edge brightness interruption time point in the image channel, detects the time segment of the main frequency change in the audio channel, determines whether there is a target support missing situation in the channel, and generates a time index sequence without target support. The response window comparison submodule calls the targetless support time index sequence, matches the erroneous trigger time segment with the current action response window one by one, filters the time overlap area, calculates the time overlap rate and concurrent duration of signal missing and response behavior in the area, and obtains the behavior interference conflict interval table. The instruction generation and output submodule extracts the conflict period between signal interruption and behavior response based on the behavior interference conflict interval table, converts the conflict period into control signal output content, sets the status of the pause instruction control flag, and generates a pause instruction.

8. The bird deterrent device with self-identification function according to claim 1, characterized in that, The interference classification module includes: The image frequency extraction submodule extracts the inter-frame change time interval of the image based on the image edge interruption signal recorded in the pause instruction, calculates the number of edge state transitions within the time interval, forms an image change rate curve, and obtains the image change frequency interval value. The audio duration calculation submodule calls the corresponding audio change signal in the pause instruction, marks the start and end time points of the change, calculates the continuous duration, and statistically analyzes the distribution range of the duration of similar events to obtain the audio change duration interval value. The combination mode classification submodule classifies time segments with the same frequency range and the duration range of audio changes into the same type based on the frequency range of the image changes and the duration range of the audio changes. It also organizes the triggering time period and structure of each type of combination and establishes a list of interference triggering features.

9. The bird deterrent device with self-identification function according to claim 1, characterized in that, The identification rule update module includes: The image threshold adjustment submodule, based on the image change frequency range value in the interference trigger feature list, calls the existing edge trigger condition parameters in the camera channel, resets the response judgment threshold corresponding to the frequency range, and overwrites the adjusted parameters according to the frame sequence index to generate an image edge response judgment parameter set. The audio range revision submodule compares the audio burst duration interval value in the interference trigger feature list with the original audio frequency detection time threshold of the pickup channel, corrects the detection window boundary according to the upper and lower limits of the time interval, updates the event trigger time range configuration, and obtains the audio time judgment parameter set. The recognition instruction generation submodule calls the image edge response judgment parameter set and the audio time judgment parameter set, sorts out the applicable range of the two types of parameters in the response stage of the bird deterrence device, and writes them into the control instruction in a formatted manner according to the channel logic mapping relationship to generate a bird deterrence recognition execution configuration instruction set.

10. The bird deterrent device with self-identification function according to claim 1, characterized in that: The response determination threshold refers to the boundary value used to determine whether the sensing signal has reached the trigger condition, and to decide whether to respond to the change in the target.

Images