Grain pile inspection structure, grain pile pest identification method and electronic equipment

Abstract

The application provides a grain pile inspection structure, a grain pile pest identification method and an electronic device, and relates to the technical field of granary pest monitoring. The method comprises the following steps: after known pests are placed at preset positions near at least one grain pile inspection unit in a grain pile, historical audio signals of different vertical depths during movement of the grain pile inspection unit along a vertical direction are collected, and a threshold parameter combination is determined based on the historical audio signals; real-time audio signals of different vertical depths during movement of the grain pile inspection unit along the vertical direction are collected; whether the real-time audio signals meet a pest determination condition is judged based on the threshold parameter combination; and when it is determined that the real-time audio signals meet the pest determination condition, pest event information is generated. The application can improve the identification accuracy of grain pile pests.

Description

Technical Field

[0001] This invention relates to the field of grain storage pest monitoring technology, and in particular to a grain pile inspection structure, a grain pile pest identification method, and an electronic device. Background Technology

[0002] In the field of grain storage pest inspection and monitoring, existing technologies mainly include four methods: manual inspection, grain surface cameras, grain pile trapping, and grain pile sound collection and recognition. Among these methods, manual inspection is inefficient, labor-intensive, and prone to missed inspections because individual grain storage silos typically exceed 50 square meters while adult grain pests are only about 1 millimeter long. While grain surface cameras can automate inspections, their monitoring range is limited to the surface of the grain pile and cannot penetrate the grain layer to detect internal pest activity, since pests often breed and multiply inside the grain pile. Although grain pile traps can penetrate deep into the grain pile for monitoring, due to cost and deployment limitations, they cannot achieve full coverage and can only effectively monitor an area of ​​about 1 cubic meter around the trap, which is insufficient to meet the overall pest awareness needs of the grain pile. Grain pile sound acquisition solutions often place the acquisition units on the grain silo walls or the surface of the grain pile, analyzing sound signals from different locations to detect and locate pests. However, due to the large space of the grain silo and the complex environmental noise, surface deployment alone cannot obtain high-quality internal acoustic signals, resulting in low recognition accuracy and coarse location positioning. It is evident that existing technologies for monitoring pests inside grain piles all suffer from low accuracy in identification. Summary of the Invention

[0003] This invention provides a grain pile inspection structure, a grain pile pest identification method, and an electronic device to solve the problem of low identification accuracy in existing technologies for monitoring pests inside grain piles.

[0004] This invention provides a grain pile inspection structure, comprising: Multiple grain pile inspection units are respectively wrapped around multiple temperature measuring cables pre-embedded inside the grain pile, and each grain pile inspection unit moves up and down along the temperature measuring cables; The main controller is communicatively connected to each of the grain pile inspection units, and is used to issue movement commands or data collection commands to each of the grain pile inspection units, and to receive the detection data returned by each of the grain pile inspection units.

[0005] According to the grain pile inspection structure provided by the present invention, the multiple temperature measuring cables are arranged in an array inside the grain pile, and each temperature measuring cable is wrapped with a grain pile inspection unit. The multiple grain pile inspection units constitute an inspection unit array inside the grain pile.

[0006] According to the grain pile inspection structure provided by the present invention, the detection data includes audio data, image data, and location data; Each of the aforementioned grain pile inspection units includes: The body is wrapped around the temperature measuring cable; A camera, mounted on the wrapping body, is used to collect image data of the inside of the grain pile; An audio acquisition sensor is installed on the wrapping body to collect audio data inside the grain pile; A power unit, mounted on the wrapped body, is used to drive the grain pile inspection unit to move up and down along the temperature measuring cable; A grid photoelectric sensor is installed on the sleeve body to count the equidistant grids on the temperature measuring cable, determine the current vertical position based on the grid counting results, and generate position data. A communication module, disposed on the sleeve body, is communicatively connected to the main controller and is used to receive the movement command or acquisition command and send the audio data, image data and location data to the main controller.

[0007] The present invention also provides a method for identifying pests in grain piles, applied to the main controller in the grain pile inspection structure described in any of the above claims, comprising: After placing known pests at a predetermined point near at least one grain pile inspection unit inside the grain pile, historical audio signals at different vertical depths are collected as the grain pile inspection unit moves along the vertical direction, and a combination of threshold parameters is determined based on the historical audio signals. Real-time audio signals at different vertical depths are collected as the grain pile inspection unit moves along the vertical direction. Based on the combination of threshold parameters, determine whether the real-time audio signal meets the pest determination criteria; When the real-time audio signal is determined to meet the pest detection criteria, pest event information is generated.

[0008] According to the grain pile pest identification method provided by the present invention, the step of collecting real-time audio signals at different vertical depths during the vertical movement of the grain pile inspection unit includes: The grain pile inspection unit is controlled to move vertically at a fixed step size. It remains stationary when it reaches a certain vertical depth and collects audio data for a preset duration at that vertical depth. The audio data collection at each vertical depth is completed sequentially to obtain the historical audio signal corresponding to each vertical depth.

[0009] According to the present invention, a method for identifying pests in grain piles, wherein determining a combination of threshold parameters based on the historical audio signal includes: Define the frequency range set and the threshold fine-tuning parameter set. A preset frequency range is randomly selected from the set of frequency ranges, and a preset threshold fine-tuning parameter is randomly selected from the set of threshold fine-tuning parameters to form a candidate threshold parameter combination; Based on the preset frequency range, the historical audio signal is filtered to obtain the target historical audio signal; Based on the target historical audio signal and the preset threshold fine-tuning parameter, a dynamic signal threshold is set, and the predicted pest location is determined according to the dynamic signal threshold. Based on the preset locations and the predicted pest locations, the detection accuracy of the candidate threshold parameter combinations is calculated, and the candidate threshold parameter combination with the highest detection accuracy is determined as the threshold parameter combination.

[0010] According to the grain pile pest identification method provided by the present invention, the step of setting a dynamic signal threshold based on the target historical audio signal and the preset threshold fine-tuning parameter includes: The target historical audio signal is subjected to regularization processing to obtain a regularized historical audio signal; Based on the regularized historical audio signal, a numerical sequence is generated, and the upper quartile is determined according to the numerical sequence. The dynamic signal threshold is determined based on the upper quartile and the preset threshold fine-tuning parameters.

[0011] According to the grain pile pest identification method provided by the present invention, the step of performing regularization processing on the target historical audio signal to obtain a regularized historical audio signal includes: Extract the amplitude data of the target historical audio signal across all time periods to form the original amplitude sequence; Based on the original amplitude sequence, determine the lower quartile, minimum value, and maximum value; Based on the lower quartile, the minimum value, and the maximum value, the target historical audio signal is regularized to obtain a regularized historical audio signal.

[0012] According to the grain pile pest identification method provided by the present invention, the step of determining the predicted pest location based on the dynamic signal threshold includes: The audio feature values ​​corresponding to different time periods in the regularized historical audio signal are compared one by one with the dynamic signal threshold. If there is at least one historical acquisition period whose corresponding audio feature value is greater than the dynamic signal threshold, then the historical audio signal is determined to meet the pest determination condition, and the historical height of the pest is determined based on the historical acquisition period and the corresponding vertical depth, and the predicted pest location is determined based on the historical height of the pest.

[0013] According to the grain pile pest identification method provided by the present invention, the threshold parameter combination includes a target preset frequency range and a target preset threshold fine-tuning parameter; The step of determining whether the real-time audio signal meets the pest detection criteria based on the threshold parameter combination includes: Based on the target preset frequency range, the real-time audio signal is filtered to obtain the target real-time audio signal; The target real-time audio signal is subjected to regularization processing to obtain a regularized real-time audio signal; Based on the regularized real-time audio signal, a real-time numerical sequence is generated, and the real-time upper quartile is determined according to the real-time numerical sequence. Based on the real-time upper quartile and the target preset threshold fine-tuning parameters, the real-time dynamic signal threshold is determined; Based on the real-time dynamic signal threshold, determine whether the real-time audio signal meets the pest identification criteria.

[0014] According to the grain pile pest identification method provided by the present invention, the step of generating pest event information when determining that the real-time audio signal meets the pest identification conditions includes: The audio feature values ​​corresponding to different acquisition time periods in the regularized real-time audio signal are compared one by one with the real-time dynamic signal threshold. If at least one target acquisition time period has an audio feature value greater than the real-time dynamic signal threshold, the real-time audio signal is determined to meet the pest determination condition. Based on the target acquisition time period and the corresponding vertical depth, the real-time height of the pest is determined, and pest event information is generated based on the real-time height of the pest.

[0015] According to the grain pile pest identification method provided by the present invention, the step of generating pest event information when determining that the real-time audio signal meets the pest identification conditions includes: When the real-time audio signal meets the pest determination criteria, initial pest event information is generated; The grain pile inspection unit collects image data corresponding to the real-time audio signal and determines the image recognition result based on the image data. When the image recognition result and the initial pest event information indicate the presence of pests at the same vertical depth, the initial pest event information is confirmed as pest event information.

[0016] The method for identifying grain pile pests according to the present invention further includes: Acquire multiple pest event information confirmed by multiple grain pile inspection units inside the grain pile, wherein each pest event information includes at least horizontal position coordinates and vertical depth coordinates; Based on a preset horizontal interval, the horizontal plane of the grain pile is divided into multiple grid units; Based on the vertical depth coordinates, pest event information at the same vertical depth is grouped into the same depth group data. Based on the horizontal position coordinates, determine the grid cell to which each pest event information belongs within each depth group of data; The horizontal coordinates of all pest events within the same depth group and belonging to the same grid cell are fused to obtain the corrected target horizontal coordinates.

[0017] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the grain pile pest identification method described above.

[0018] This invention provides a grain pile inspection structure, a grain pile pest identification method, and an electronic device. After placing a known pest at a preset point near at least one grain pile inspection unit inside the grain pile, historical audio signals at different vertical depths are collected as the grain pile inspection unit moves vertically. Based on these historical audio signals, a combination of threshold parameters is determined. Real-time audio signals at different vertical depths are also collected as the grain pile inspection unit moves vertically. Based on the threshold parameter combination, it is determined whether the real-time audio signals meet pest identification criteria. When the real-time audio signals meet the pest identification criteria, pest event information is generated. This invention addresses the problem of low accuracy in monitoring pests inside grain piles using existing technologies. Compared to existing technologies, this invention utilizes a movable grain pile inspection unit wrapped around a pre-embedded temperature-measuring cable inside the grain pile, enabling intensive inspection of the vertical profile of the grain pile. By placing known pests at preset locations and collecting historical audio signals using this structure, adaptive calibration of threshold parameter combinations is achieved, ensuring that the filtering frequency and judgment threshold accurately match the acoustic propagation characteristics inside the grain pile. Based on this, real-time audio signals are processed and dynamically thresholded using the calibrated threshold parameter combinations, effectively filtering out complex environmental noise interference inside the grain pile and significantly improving the accuracy of pest activity identification. Simultaneously, by collecting real-time audio signals at different vertical depths along the temperature-measuring cable and generating pest events containing three-dimensional location information, the inspection unit achieves automated monitoring and precise location of pests inside the grain pile, providing reliable support for early warning and targeted control of pests within grain piles. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the grain pile inspection structure provided by the present invention.

[0021] Figure 2 This is a schematic diagram of the deployment of the grain pile inspection unit array provided by the present invention.

[0022] Figure 3 This is a structural diagram of the grain pile inspection unit provided by the present invention.

[0023] Figure 4 This is a sectional view of the grain pile inspection unit provided by the present invention.

[0024] Figure 5 This is one of the flowcharts illustrating the grain pile pest identification method provided by the present invention.

[0025] Figure 6 This is a schematic diagram of the grid sensor used in the grain pile pest identification method provided by the present invention.

[0026] Figure 7 This is the second flowchart of the grain pile pest identification method provided by the present invention.

[0027] Figure 8 This is an example diagram of the audio data sequence of the grain pile pest identification method provided by the present invention.

[0028] Figure 9 This is a schematic diagram of the vertical intensity of the audio signal in the grain pile pest identification method provided by the present invention.

[0029] Figure 10 This is the third flowchart of the grain pile pest identification method provided by the present invention.

[0030] Figure 11 This is a schematic diagram of the horizontal intensity of the audio signal in the grain pile pest identification method provided by the present invention.

[0031] Figure 12 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0032] Figure label: 1-1: Temperature measuring cable; 1-2: Grain pile inspection unit; 1-3: Grain silo wall; 2-2: Camera; 2-3: Audio acquisition sensor; 2-4: Power component; 2-5: Grid photoelectric sensor; 2-6: Internal cable pulley; 2-7: Temperature measuring cable grid; 2-8: Temperature measuring cable slide rail; 2-9: Rotary motor. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0034] Figure 1 This is a schematic diagram of the grain pile inspection structure provided by the present invention, as shown below. Figure 1 As shown, the grain pile inspection structure specifically includes: Multiple grain pile inspection units 1-2 are respectively wrapped around multiple temperature measuring cables 1-1 pre-embedded inside the grain pile, and each of the grain pile inspection units 1-2 moves up and down along the temperature measuring cables 1-1.

[0035] The main controller is communicatively connected to each of the grain pile inspection units 1-2, and is used to issue movement commands or data collection commands to each of the grain pile inspection units 1-2, and to receive the detection data returned by each of the grain pile inspection units 1-2.

[0036] It should be noted that multiple temperature measuring cables 1-1 are pre-embedded inside the grain silo enclosed by the grain silo walls 1-3. These cables are vertical cables extending from the top to the bottom of the grain pile, primarily used for routine monitoring of the internal temperature of the grain pile. This invention fully utilizes the existing deployment of the temperature measuring cables 1-1 in the grain depot, configuring a grain pile inspection unit 1-2 for each temperature measuring cable 1-1. This inspection unit is wrapped around the outside of the temperature measuring cable 1-1, and its overall dimensions are designed with a length between 300mm and 800mm and a cross-sectional dimension between 20mm and 50mm to facilitate movement among the grain grains and reduce disturbance to the original structure of the grain pile.

[0037] It should be noted that the central controller is wirelessly connected to each grain pile inspection unit 1-2, and is responsible for the unified scheduling of the entire inspection array. During data acquisition, the central controller first issues a reset command, controlling all inspection units to move to the top of the temperature measuring cable 1-1 to complete position calibration; then it issues a movement command, controlling each unit to move downwards according to a preset step size, and confirming the arrival at the target position in real time through grid counting; after being ready, the central controller issues an acquisition command, controlling each unit to synchronously activate the camera and audio sensor to acquire image and audio data of the current location; finally, it receives and stores the detection data returned by each unit, providing a data foundation for subsequent pest detection and location. Through centralized control, the central controller can achieve coordinated scheduling of multiple inspection units, ensuring the synchronization and consistency of array-style data acquisition.

[0038] Based on any of the above embodiments, the multiple temperature measuring cables 1-1 are arranged in an array inside the grain pile, and each of the temperature measuring cables 1-1 is wrapped with a grain pile inspection unit 1-2. The multiple grain pile inspection units 1-2 constitute an inspection unit array inside the grain pile.

[0039] It should be noted that, as Figure 2 As shown, multiple temperature measuring cables 1-1 are arranged in an array at preset intervals inside the grain silo, for example, at equal intervals along the x and y directions in a horizontal plane, forming a grid-like layout covering the entire cross-section of the grain pile. Each temperature measuring cable 1-1 is wrapped with a grain pile inspection unit 1-2, and multiple grain pile inspection units 1-2 together constitute an inspection unit array in three-dimensional space inside the grain pile. Figure 2 In the diagram, solid black spheres represent the current position of each inspection unit, while dashed spheres represent other vertical positions that each inspection unit can reach by moving up and down along temperature measuring cable 1-1. Through unified scheduling by the central controller, each grain pile inspection unit 1-2 can independently move to any vertical depth along its respective temperature measuring cable 1-1, thereby forming a dynamically adjustable three-dimensional sampling grid inside the grain pile.

[0040] Understandably, when collecting audio signals from pests, the signal strength decreases with distance, so the closer the inspection unit is to the pest's location, the stronger the signal it collects. Utilizing this characteristic, by comparing the signal strength differences between inspection units in the grain pile array through differential analysis, and combining this with the unit's spatial coordinates, the location of the pest can be deduced, achieving precise three-dimensional positioning of pests inside the grain pile.

[0041] Based on any of the above embodiments, the detection data includes audio data, image data, and location data; Each of the aforementioned grain pile inspection units 1-2 includes: The body is wrapped around the temperature measuring cable 1-1; Camera 2-2 is mounted on the wrapping body and is used to collect image data of the inside of the grain pile; Audio acquisition sensors 2-3 are installed on the wrapping body and are used to collect audio data inside the grain pile; Power assembly 2-4 is mounted on the wrapping body and is used to drive the grain pile inspection unit 1-2 to move up and down along the temperature measuring cable 1-1. A grid photoelectric sensor is installed on the sleeve body to count the equidistant grids on the temperature measuring cable 1-1, determine the current vertical position based on the grid counting results, and generate position data. A communication module is installed on the sleeve body and is connected to the main controller for receiving the motion control command or data acquisition command, and sending the audio data, image data and location data to the main controller.

[0042] It should be noted that, as Figure 3 as well as Figure 4 As shown, the encasing body has a symmetrical structure to maintain balance during movement, and its interior is equipped with installation space to accommodate various functional components. The temperature measuring cable 1-1 runs through the encasing body, ensuring the entire grain pile inspection unit is tightly wrapped around the outside of the temperature measuring cable 1-1, providing a stable support foundation for each functional module. Based on this, the grain pile inspection unit 1-2 integrates the following functional components to achieve data acquisition, autonomous movement, and communication functions: (1) Camera 2-2, set on the side of the wrapped body, is used to capture the internal image of the grain pile at the current location; (2) Preferably, there are multiple audio acquisition sensors 2-3, for example, two are set on the front and two on the back of the body to achieve all-round acquisition of ambient sound.

[0043] (3) The power assembly 2-4 further includes an external moving device and a rotary motor. The external moving device is divided into upper and lower parts, which are responsible for driving the inspection unit to move up or down on the temperature measuring cable 1-1, respectively. The rotary motor provides driving force for the external moving device.

[0044] (4) The grid photoelectric sensor 2-5 works in conjunction with the equidistant grid on the temperature measuring cable 1-1 to accurately locate the current position of the inspection unit in the vertical direction by detecting the number of grids.

[0045] In addition, the casing also includes internal cable pulleys 2-6 to assist the inspection unit in moving smoothly on the temperature measuring cable 1-1, and to provide power through contact with the temperature measuring cable 1-1; it also includes a temperature measuring cable grid 2-7, which is composed of equally spaced scale units, and can achieve accurate vertical position counting in conjunction with the grid photoelectric sensor on the inspection unit; the temperature measuring cable slide rail 2-8 provides guidance for the movement of the inspection unit, ensuring its smooth sliding on the temperature measuring cable; and the rotary motor 2-9 serves as a power source to drive the external moving device to move the inspection unit up and down.

[0046] Based on the above structure, the main working process of the grain pile inspection unit 1-2 is as follows: First, the communication module receives the movement control command issued by the main controller, and the rotary motor in the power component starts, driving the external moving device to move the entire inspection unit upward or downward along the temperature measuring cable 1-1. During the movement, the internal cable pulley cooperates with the slide rail of the temperature measuring cable 1-1 to ensure the smoothness of the movement and the guiding accuracy. At the same time, the grid photoelectric sensor counts the equidistant grids on the temperature measuring cable 1-1 in real time, determines the current vertical position by the number of grids, and generates position data. When the inspection unit reaches the target position, according to the data acquisition command issued by the main controller, the camera and audio acquisition sensor are activated to collect image data of the grain pile inside the current position and audio data of the surrounding environment, respectively. Finally, the communication module sends the collected image data, audio data, and corresponding position data to the main controller for storage and subsequent analysis and processing. Through the above workflow, the grain pile inspection unit 1-2 can move autonomously along the temperature measuring cable 1-1 inside the grain pile and collect multimodal data, providing a reliable data source for pest detection and location.

[0047] Figure 5 This is one of the flowcharts illustrating the grain pile pest identification method provided by the present invention, applied to a central controller, such as... Figure 5 As shown, the method includes the following: Step 101: After placing known pests at a preset point near at least one grain pile inspection unit 1-2 inside the grain pile, collect historical audio data at different vertical depths during the vertical movement of the grain pile inspection unit 1-2, and determine the threshold parameter combination based on the historical audio data. It should be noted that the grain pile inspection units 1-2 refer to detection devices that can move vertically and collect audio signals; the preset points are spatial coordinates pre-selected to simulate real pest occurrence scenarios. These points are usually representative and can cover different horizontal positions and vertical depths to ensure the comprehensiveness of calibration data. By placing known pests at these points, the system obtains the standard answer regarding the true location of the pests, namely the set of true values ​​(PGTX, PGTY, PGTZ), which will serve as the evaluation benchmark for subsequent parameter optimization.

[0048] Understandably, after placing known pests, the grain pile inspection unit 1-2 is controlled to move at a constant speed or in steps along the vertical direction, collecting audio signals at different vertical depths during the movement. These signals constitute the "historical audio data." Since the collection process covers the depth layer containing the known pests, this historical audio data includes characteristic signals generated by pest activity and their background noise. Subsequently, the system processes and analyzes this historical audio data, with the core objective of determining the "threshold parameter combination." This threshold parameter combination is a set of adjustable parameters used to control signal processing and pest detection, typically including, but not limited to, the cutoff frequency of the bandpass filter (used to extract pest activity sounds in specific frequency bands) and the signal threshold fine-tuning coefficient (used to adjust the sensitivity of the detection).

[0049] In the specific implementation, active pests are placed at predetermined locations inside the grain pile, and the pest locations are marked as follows. By identifying the location of pests within the deployment range of the temperature-sensing cable, it is determined that pests are present near the current temperature-sensing cable, and the pests are grouped together. .

[0050] Step 102: Collect real-time audio signals at different vertical depths as the grain pile inspection unit 1-2 moves along the vertical direction; In the specific implementation, the grain pile inspection unit 1-2 is controlled to move vertically along the temperature measuring cable 1-1, descending layer by layer according to a preset fixed step length (e.g., 10 cm). When it moves to a preset vertical depth position, the grain pile inspection unit 1-2 is controlled to remain stationary. After the mechanical vibration is eliminated, the audio acquisition sensor 2-3 is activated to collect the real-time audio signal at that vertical depth for a preset duration (e.g., 10 seconds). At the same time, the grid photoelectric sensor 2-5 records the current vertical depth position information in real time. After completing the acquisition of the current vertical depth, it continues to descend to the next vertical depth, repeating the above process until the real-time audio signal acquisition of all preset vertical depths is completed.

[0051] Understandably, real-time audio signals are relative to "historical audio signals." They represent the dynamic data stream acquired by the system in actual operation for the current state of the grain pile, reflecting the real acoustic environment inside the grain pile at the current moment.

[0052] Step 103: Based on the threshold parameter combination, determine whether the real-time audio signal meets the pest determination conditions; It should be noted that the threshold parameter combination is the optimal set of parameters obtained through calibration with known pests. It typically includes core parameters such as the cutoff frequency of the bandpass filter and the signal threshold fine-tuning coefficient. Its function is to map the original audio signal to a feature space that can effectively characterize pest activity. The pest determination criteria are a set of logical rules based on statistical characteristics or empirical thresholds, used to distinguish normal background noise from abnormal pest activity sounds.

[0053] In its implementation, the real-time audio signal is first frequency-domain filtered using the filtering parameters in the threshold parameter combination to extract characteristic frequency band signals related to pest activity and filter out environmental noise and mechanical interference. Then, the processed real-time audio signal is compared with a dynamic threshold, which is typically determined based on the upper quartile of historical data sequences at that vertical depth and adaptively corrected using threshold fine-tuning coefficients in the threshold parameter combination. The core of this process is that the system does not use a fixed threshold for judgment, but rather dynamically adjusts the judgment criteria based on the acoustic characteristics of different depths within the grain pile using calibrated parameters, thereby adapting to the complex acoustic environment inside the grain pile.

[0054] Step 104: When the real-time audio signal is determined to meet the pest determination conditions, pest event information is generated.

[0055] It should be noted that pest event information refers to a digital description used to characterize a pest activity, which includes at least the following core elements: the horizontal coordinates of the pest's location, the vertical depth coordinates of the pest's location, and the timestamp of the event. The horizontal coordinates are uniquely determined by the planar position of the cable connecting the grain pile inspection units 1-2, the vertical depth coordinates are calculated based on the acquisition time of the real-time audio signal, combined with the inspection unit's movement speed and cycle, and the timestamp records the precise moment the event occurred.

[0056] The grain pile pest identification method provided in this invention achieves intensive inspection of the vertical profile inside the grain pile by wrapping a vertically movable grain pile inspection unit around a temperature measuring cable pre-embedded inside the grain pile. Using this structure, known pests are placed at preset points and historical audio signals are collected, enabling adaptive calibration of threshold parameter combinations. This ensures that the filtering frequency and judgment threshold accurately match the acoustic propagation characteristics inside the grain pile. Based on this, real-time audio signals are processed and dynamically thresholded according to the calibrated threshold parameter combinations, effectively filtering out complex environmental noise interference inside the grain pile and significantly improving the accuracy of pest activity identification. Simultaneously, by collecting real-time audio signals at different vertical depths along the temperature measuring cable and generating pest events containing three-dimensional positional information, automated monitoring and precise location of pests inside the grain pile are achieved, providing reliable support for early warning and targeted control of pests inside the grain pile.

[0057] Based on any of the above embodiments, the acquisition of historical audio signals at different vertical depths during the vertical movement of the grain pile inspection unit 1-2 includes: The grain pile inspection unit 1-2 is controlled to move vertically with a fixed step size. It remains stationary when it reaches a certain vertical depth and collects audio data for a preset duration at that vertical depth. The audio data collection at each vertical depth is completed in sequence to obtain the historical audio signal corresponding to each vertical depth.

[0058] It should be noted that a fixed step length means that the interval between each vertical movement of the inspection unit is preset and equal, for example, moving once every 10cm. This uniform sampling method ensures that the vertical profile of the grain pile is divided into several monitoring layers at equal intervals, and each monitoring layer has an independent sampling opportunity, avoiding the problem of some depth layers being missed or oversampled due to uneven movement speed or irregular sampling intervals. The specific value of the step length can be set according to the height of the grain pile, the activity habits of pests, and the monitoring accuracy requirements, and is usually in the range of 5cm to 20cm.

[0059] Understandably, once the inspection unit reaches the preset vertical depth (e.g., 30cm), the control system instructs its drive motor to stop, allowing the grain pile inspection units 1-2 to remain stably stationary at that depth. This eliminates interference from mechanical vibrations and airflow disturbances caused by movement on audio acquisition, ensuring that the acquired audio signal reflects only the true acoustic environment inside the grain pile at that depth, without incorporating additional noise from the movement process. Simultaneously, the stationary state provides a stable physical platform for subsequent long-term continuous data acquisition.

[0060] It should be noted that the preset duration refers to the length of time that the grain pile inspection unit 1-2 continuously collects audio at each stopping depth, such as 10 seconds, 30 seconds, or 1 minute. Within this duration, the system continuously records the audio signal at a high sampling frequency (e.g., 100 times per second), thereby obtaining a time-series dataset (e.g., containing 1000 data points) at that depth. Compared to the limitation of only being able to collect a single instantaneous value at each depth in the uniform speed movement mode, this implementation method, through preset duration collection, allows a large number of data samples to be accumulated at each depth. These samples reflect the randomness, intermittency, and fluctuation characteristics of pest activity and background noise at that depth.

[0061] In the specific implementation, after receiving the movement command issued by the main controller, the grain pile inspection unit 1-2 starts the rotary motor 2-9 to move downward along the temperature measuring cable 1-1. At the same time, the grid photoelectric sensor 2-5 monitors the number of jumps of the temperature measuring cable grid 2-7 in real time. Since the grid is evenly distributed, the movement distance can be accurately calculated by accumulating the number of jumps and multiplying it by the preset step size. When the detected movement distance reaches a preset step size (e.g., 10cm), the controller instructs the motor to stop, keeping the grain pile inspection unit 1-2 stationary at the current depth. After the mechanical vibration is eliminated, the audio acquisition sensor 2-3 and camera 2-2 are immediately activated to synchronously acquire audio and image data for a preset duration TS (e.g., 10 seconds), obtaining the audio signal and image data P(x,y,t) at the current depth. Here, x and y are the horizontal coordinates of the temperature measuring cable 1-1 corresponding to the grain pile inspection unit 1-2 within the grain silo. A coordinate system is established with the right angle from the grain silo entrance door to the left rear of the silo as the origin, and t is the time variable for acquisition. After acquisition, the data, along with the current depth position information, is sent to the main controller for storage in real time. TS is the acquisition duration hyperparameter, which can be set to 10 seconds. A longer time results in better detection but lower efficiency.

[0062] In specific implementations, such as Figure 6 The schematic diagram of the grid sensor shows that the grid photoelectric sensor 2-5 monitors the transition status of the temperature measuring cable grid 2-7 in real time by transmitting carrier photoelectric signals and receiving reflected signals. When the grain pile inspection unit 1-2 moves, the transmission and reception time of the sensor output signal changes as the grid passes. By setting a time threshold and using transition detection technology, a transition count is generated each time a grid passes the sensor. Figure 6 In the example, there were 3 jumps, corresponding to a movement of 3 grid squares. Since the grid squares are evenly distributed, the cumulative number of jumps multiplied by the grid spacing is the actual movement distance, thus enabling accurate measurement of the vertical position of the grain pile inspection unit.

[0063] The grain pile pest identification method provided in this invention achieves layered and precise monitoring of pest activity inside the grain pile by vertically moving at a fixed step length and collecting audio at various depths while remaining stationary. This effectively avoids interference from moving noise and ensures the purity of the audio signal. The method can comprehensively acquire pest activity information at different depths, improve the accuracy and reliability of identification, and provide a scientific basis for early warning and location control of grain pile pests.

[0064] Figure 7 This is the second flowchart illustrating the grain pile pest identification method provided by the present invention, as shown below. Figure 7 As shown, step 101 further includes steps 1011 to 1015: Step 1011: Set the frequency range set and the threshold fine-tuning parameter set. It should be noted that the frequency range set is a set of multiple bandpass filter frequency intervals pre-set based on the statistical characteristics of historical audio data. It is used to effectively filter out background noise such as mechanical noise and airflow interference in the grain pile environment in the frequency domain, while accurately retaining the characteristic frequency components generated by pest activities. The threshold fine-tuning parameter set is a set of fine-tuning coefficient steps preset for each grain pile inspection unit 1-2. It is used to dynamically adjust the judgment threshold of audio intensity in the time domain to adapt to the changes in acoustic characteristics caused by differences in grain pile medium and propagation distance at different locations.

[0065] Step 1012: Randomly select a preset frequency range from the set of frequency ranges, and randomly select a preset threshold fine-tuning parameter from the set of threshold fine-tuning parameters to form a candidate threshold parameter combination; It should be noted that the candidate threshold parameter combination is a pair of parameters consisting of a randomly selected set of frequency ranges and a set of threshold fine-tuning coefficients.

[0066] Step 1013: Based on the preset frequency range, filter the historical audio signal to obtain the target historical audio signal; It should be noted that, based on a preset frequency range randomly selected from the set of frequency ranges, a corresponding bandpass filter is constructed to filter the collected historical audio signal. The purpose is to effectively filter out mechanical operating noise, airflow disturbance, and other low-frequency or high-frequency interference components in the grain pile environment, while retaining the characteristic sound waves of pest activity that may be contained in the frequency range. After filtering, the target historical audio signal has a significantly improved signal-to-noise ratio, and the characteristic frequency components of pests are more prominent.

[0067] Step 1014: Based on the target historical audio signal and the preset threshold fine-tuning parameter, set a dynamic signal threshold, and determine the predicted pest location according to the dynamic signal threshold; It should be noted that, based on the filtered target historical audio signal, and combined with preset threshold fine-tuning parameters randomly selected from the threshold fine-tuning parameter set, the judgment threshold of the current audio data sequence is dynamically calculated. The dynamic signal threshold is usually set based on the statistical characteristics of the audio signal, such as by using the 3 / 4 quantile of the sequence superimposed with the threshold fine-tuning coefficient, to adaptively reflect the acoustic background intensity of the current location. Locations in the target historical audio signal that are higher than this dynamic threshold are identified as candidate locations where pests may be active, and their corresponding spatial coordinates and time information are recorded to form a predicted pest location sequence. This dynamic threshold setting mechanism can effectively adapt to the differences in sound wave attenuation caused by different depths and different grain pile media, avoiding the problems of missed detection or false detection caused by using a fixed threshold.

[0068] In the specific implementation, the amplitude of each time t in the regularized historical audio signal is compared with the dynamic signal threshold one by one. ,in, Indicates the dynamic signal threshold. For all of history The set of data; Q3(·) is the 3 / 4 digit (i.e., the upper quartile) in the sequence. This represents the threshold fine-tuning coefficient, and I(·) represents the integrated values ​​that meet the requirements. This represents the regularized historical audio signal. If pest activity is suspected at a given moment, the location corresponding to that moment is recorded as the predicted pest location; all locations meeting the conditions together constitute the predicted pest location sequence. .

[0069] Step 1015: Based on the preset location and the predicted pest location, calculate the detection accuracy of the candidate threshold parameter combination, and determine the candidate threshold parameter combination with the highest detection accuracy as the threshold parameter combination.

[0070] It should be noted that the preset points refer to the actual coordinates of active pests that are artificially placed inside the grain pile during the initialization phase, serving as the baseline true value for verifying the effectiveness of the parameters. The predicted pest location is the sequence of possible pest locations determined by processing historical audio signals based on the current candidate threshold parameter combination (including preset frequency range and preset threshold fine-tuning parameters). By comparing the predicted location with the preset points, the detection accuracy under the current candidate threshold parameter combination is calculated, which can be quantitatively evaluated using indicators such as overlap and the number of correct detections. After traversing all randomly generated candidate threshold parameter combinations, the candidate threshold parameter combination with the highest detection accuracy is determined as the threshold parameter combination for the grain pile inspection unit 1-2.

[0071] The grain pile pest identification method provided in this invention sets a frequency range and a threshold fine-tuning parameter set, randomly combines candidate parameters, and sequentially performs filtering, dynamic threshold setting, and predicted location determination. Then, based on preset points, it calculates the detection accuracy of each combination and selects the optimal threshold parameter combination. This method achieves automated optimization and personalized calibration of audio processing parameters, effectively eliminating the adaptability bias of fixed parameters in complex grain pile environments, and significantly improving the accuracy and robustness of pest identification.

[0072] Based on any of the above embodiments, the step of setting the dynamic signal threshold based on the target historical audio signal and the preset threshold fine-tuning parameter includes: The target historical audio signal is subjected to regularization processing to obtain a regularized historical audio signal; Based on the regularized historical audio signal, a numerical sequence is generated, and the upper quartile is determined according to the numerical sequence. The dynamic signal threshold is determined based on the upper quartile and the preset threshold fine-tuning parameters.

[0073] It should be noted that a numerical sequence refers to a set of data composed of regularized historical audio signals arranged in chronological order. Numerical sequences are based on regularized historical audio signals. The generated sequence contains a regularized amplitude for each element at a specific time t. Assuming a sampling duration of TS = 10 seconds at a certain vertical depth and a sampling frequency of 100 Hz, a total of 1000 sampling points are obtained. These sampling points, arranged chronologically, form a numerical sequence of length 1000. The upper quartile refers to the value at the 75th percentile (three-quarters of the way down) after sorting the numerical sequence from smallest to largest.

[0074] In practical implementation, the formula for calculating the dynamic signal threshold Td is: Td = Q3 + STH(x,y) Where Q3 is the upper quartile of the numerical sequence, reflecting the high-level distribution characteristics of the depth signal; STH(x,y) is a preset threshold fine-tuning parameter used to adaptively adjust the sensitivity.

[0075] It is understandable that the value of the preset threshold fine-tuning parameter must be less than the maximum value in the numerical sequence, that is: , This indicates the preset threshold fine-tuning parameter. This represents the maximum value in the numerical sequence; this ensures that the threshold adjustment range does not exceed the maximum intensity range of the historical signal, avoiding missed detections due to excessively high threshold settings.

[0076] For example: Figure 8As shown, we constructed some audio numerical sequences. Taking SNp(1,1,1) = 10 as an example, it means that the audio signal strength is 10 when the time is 1 unit on the temperature measuring cable (corresponding to the grain pile inspection unit) 1 meter away from the origin in the x direction and 1 meter away in the y direction.

[0077] Another assumption , =5, according to the dynamic signal threshold, then .

[0078] In the specific implementation, the target historical audio signal obtained after filtering is first regularized to eliminate the absolute difference in signal amplitude caused by differences in grain pile medium or equipment status at different depths and time periods, so that the data are comparable under the same dimension, and a regularized historical audio signal is obtained. Then, a numerical sequence is constructed based on the regularized signal, and the upper quartile (i.e., 3 / 4 quartile) of the sequence is calculated. This statistic can effectively reflect the high-level distribution characteristics of signal intensity and has stronger robustness and anti-interference ability compared with the mean or maximum value. Finally, the calculated upper quartile is combined with the preset threshold fine-tuning parameter to generate the final dynamic signal threshold.

[0079] The grain pile pest identification method provided in this invention eliminates amplitude differences by regularizing the target historical audio signal and extracting the upper quartile to characterize the high-level distribution features of the signal. This is then fused with a preset threshold fine-tuning parameter to generate a dynamic signal threshold. This method utilizes the adaptability of statistical features and the personalized calibration capability of the fine-tuning parameter, enabling the threshold to be dynamically adjusted according to the acoustic environment at different locations, effectively improving the accuracy and environmental adaptability of pest identification.

[0080] Based on any of the above embodiments, the step of performing regularization processing on the target historical audio signal to obtain a regularized historical audio signal includes: Extract the amplitude data of the target historical audio signal across all time periods to form the original amplitude sequence; Based on the original amplitude sequence, determine the lower quartile, minimum value, and maximum value; Based on the lower quartile, the minimum value, and the maximum value, the target historical audio signal is regularized to obtain a regularized historical audio signal.

[0081] It should be noted that the core of the regularization process for the target historical audio signal lies in eliminating the absolute differences in amplitude of audio signals at different locations and time periods through statistical feature extraction and normalization calculation, thus making the data comparable under the same dimension. Specifically, firstly, all amplitude data of the target historical audio signal are extracted throughout the entire time period to construct an original amplitude series, which completely preserves the intensity distribution information of the signal at different time points. Subsequently, the lower quartile, minimum value, and maximum value are calculated based on this series. The lower quartile reflects the low-order distribution characteristics of the signal intensity and is used to characterize the typical level of background noise. The minimum value represents the absolute lower limit of the signal, and the maximum value represents the peak level of the signal intensity. Finally, the target historical audio signal is regularized based on the above three statistical parameters. This is usually achieved by subtracting the minimum value from each data point and then dividing by the combination of the lower quartile and the maximum value, thus realizing the normalization transformation of the signal amplitude. Specifically, the specific calculation formula for regularizing the audio signal is as follows: In the formula, This represents the regularized historical audio signal; Q1(·) represents the 1 / 4 digit (lower quartile) of the original amplitude sequence; min(·) represents the minimum value of the original amplitude sequence; max(·) represents the maximum value of the original amplitude sequence. This indicates the target's historical audio signal.

[0082] The grain pile pest identification method provided in this invention extracts the full-time amplitude data of the target's historical audio signal and performs regularization processing based on the lower quartile, minimum value, and maximum value. This method utilizes the lower quartile to characterize the background noise level and combines the minimum and maximum values ​​to achieve signal amplitude normalization, effectively eliminating signal drift caused by differences in grain pile depth or equipment fluctuations, and suppressing background interference while preserving the relative intensity of pest characteristics.

[0083] Based on any of the above embodiments, determining the predicted pest location according to the dynamic signal threshold includes: The audio feature values ​​corresponding to different time periods in the regularized historical audio signal are compared one by one with the dynamic signal threshold. If there is at least one historical acquisition period whose corresponding audio feature value is greater than the dynamic signal threshold, then the historical audio signal is determined to meet the pest determination condition, and the historical height of the pest is determined based on the historical acquisition period and the corresponding vertical depth, and the predicted pest location is determined based on the historical height of the pest.

[0084] It should be noted that the process of determining the predicted pest location based on the dynamic signal threshold realizes the mapping from audio features to spatial location. Specifically, the audio feature values ​​corresponding to different time periods in the regularized historical audio signal are compared one by one with the dynamic signal threshold. By comparing point by point, abnormal audio events that may be caused by pest activity are identified. If the audio feature value corresponding to at least one historical acquisition time period is greater than the real-time dynamic signal threshold, it is determined that the audio signal of that time period meets the pest determination condition, indicating that there is a possibility of pest activity near the location of the inspection unit during that time period. On this basis, based on the historical acquisition time period that meets the condition and its corresponding vertical depth, the historical height of the pest in that time period is determined. Combined with the fixed horizontal coordinates of the inspection unit, the predicted pest location containing three-dimensional spatial information is finally obtained. This determination mechanism adaptively captures abnormal acoustic features through dynamic threshold and accurately locates the spatiotemporal location of the anomaly.

[0085] In specific implementations, such as Figure 9 As shown, the horizontal axis represents the vertical depth of the inspection unit on the temperature measuring cable, and the vertical axis represents the audio signal intensity at the corresponding vertical depth. When the regularized audio feature value exceeds the dynamic signal threshold, it is determined that there is suspected pest activity during that time period. Based on the position t of that time period in the acquisition sequence (e.g., the t-th acquisition time period), the preset fixed step size DM (i.e., vertical movement distance), and the acquisition duration TS, according to the formula... By calculating the vertical displacement and combining it with the starting position of the grain pile inspection unit at a preset vertical height (such as the top of the temperature measuring cable as the origin), the actual historical height of the pest can be obtained. This formula converts the time dimension into spatial depth, thereby obtaining the precise height of the pest on the vertical profile. The horizontal coordinate corresponding to the point that exceeds the threshold is the depth, realizing a precise mapping from audio features to vertical position.

[0086] For example: Assume TS is 10 seconds and DM is 100 centimeters. =[(1,1,3)], pz=30 (cm), thus obtaining =[(1,1,30)].

[0087] The grain pile pest identification method provided in this invention compares regularized audio feature values ​​with dynamic thresholds one by one. When a feature value exceeds the threshold, pest activity is determined, and the historical height of the pest is determined based on the target collection time period and corresponding vertical depth to generate a predicted location. This method achieves accurate conversion from historical audio signals to the spatial location of pests, effectively capturing abnormal acoustic events and accurately locating their vertical distribution.

[0088] Figure 10 This is the third flowchart of the grain pile pest identification method provided by the present invention. The threshold parameter combination includes a target preset frequency range and a target preset threshold fine-tuning parameter, such as... Figure 10 As shown, step 103 further includes steps 1031 to 1035: Step 1031: Based on the target preset frequency range, filter the real-time audio signal to obtain the target real-time audio signal; It should be noted that after entering the real-time pest detection stage, the real-time acquired audio signal is first filtered based on the target preset frequency range determined in the initialization and correction steps. This target preset frequency range is determined during the initialization process by traversing multiple candidate frequency intervals and selecting the best one, which can effectively match the acoustic propagation characteristics of the current inspection unit's location. The core of the filtering process is to use this frequency range to construct a bandpass filter to screen the frequency components in the real-time audio signal, filtering out mechanical operating noise, airflow interference, and other irrelevant background noise in the grain pile environment, while retaining the pest activity sound waves that may be contained within this characteristic frequency range. After filtering, the target real-time audio signal has a significantly improved signal-to-noise ratio, and the characteristic frequency components of pests are more prominent.

[0089] Step 1032: Perform regularization processing on the target real-time audio signal to obtain a regularized real-time audio signal; It should be noted that during real-time pest detection, the filtered real-time audio signal is regularized to eliminate absolute differences in signal amplitude caused by variations in grain pile depth, media inhomogeneity, or equipment status fluctuations. This ensures that audio data collected at different times and locations are comparable under the same dimension. The regularization process uses the same statistical parameters as the initialization and correction stage, typically determining the lower quartile, minimum, and maximum values ​​based on the amplitude distribution characteristics of the target's historical audio signal, and then normalizing the real-time signal. The regularized real-time audio signal obtained after regularization has its amplitude mapped to a uniform numerical range, preserving the relative intensity relationship of pest activity characteristic signals while effectively suppressing absolute amplitude fluctuations in background noise.

[0090] Step 1033: Based on the regularized real-time audio signal, generate a real-time numerical sequence, and determine the real-time upper quartiles according to the real-time numerical sequence; It should be noted that during real-time detection, statistical features are extracted from the regularized real-time audio signal. Specifically, a real-time numerical sequence is constructed and its upper quartile is calculated to characterize the high-level distribution characteristics of the signal intensity in the current time period. This upper quartile can effectively reflect the degree of acoustic anomalies that pest activities may cause, while avoiding interference from occasional noise on the threshold setting.

[0091] Step 1034: Determine the real-time dynamic signal threshold based on the real-time upper quartile and the target preset threshold fine-tuning parameters; It should be noted that during real-time detection, the determination of the dynamic signal threshold integrates the statistical characteristics of the current time period with the personalized parameters calibrated during the initialization phase. Specifically, the real-time upper quartile calculated based on the regularized real-time audio signal reflects the high-level distribution of signal intensity during the current acquisition period, which can effectively capture acoustic anomalies that may be caused by pest activity. This real-time upper quartile is combined with the target preset threshold fine-tuning parameters determined during the initialization correction phase to generate a real-time dynamic signal threshold suitable for the current moment.

[0092] Step 1035: Based on the real-time dynamic signal threshold, determine whether the real-time audio signal meets the pest determination criteria.

[0093] It should be noted that the audio feature values ​​corresponding to each acquisition time in the regularized real-time audio signal are compared one by one with the real-time dynamic signal threshold. This real-time dynamic threshold integrates the statistical characteristics of the current time period with pre-calibrated fine-tuning parameters, and can adaptively reflect the real-time acoustic background level inside the grain pile. If the audio feature value at at least one acquisition time is greater than the real-time dynamic threshold, the current real-time audio signal is determined to meet the pest detection criteria, indicating that there may be pest activity near the location of the inspection unit; otherwise, it is considered that there are no significant signs of pest activity during that time period.

[0094] The grain pile pest identification method provided in this invention achieves dynamic determination of pest activity through a real-time audio processing workflow: First, the real-time audio is filtered based on a target preset frequency range to effectively filter out environmental noise; then, the signal is regularized and the real-time upper quartiles are extracted to accurately characterize the high-level distribution features of the current signal; this statistic is fused with a pre-calibrated target threshold fine-tuning parameter to generate a real-time dynamic threshold that adapts to the current acoustic environment; finally, pest activity is determined by threshold comparison. This method achieves full-process coordination of filtering, regularization, statistical feature extraction, and dynamic threshold setting, significantly improving the environmental adaptability and judgment accuracy of real-time detection, and providing a reliable guarantee for the accurate identification of pests inside grain piles.

[0095] Based on any of the above embodiments, the step of generating pest event information when determining that the real-time audio signal meets the pest determination criteria includes: The audio feature values ​​corresponding to different acquisition time periods in the regularized real-time audio signal are compared one by one with the real-time dynamic signal threshold. If at least one target acquisition time period has an audio feature value greater than the real-time dynamic signal threshold, the real-time audio signal is determined to meet the pest determination condition. Based on the target acquisition time period and the corresponding vertical depth, the real-time height of the pest is determined, and pest event information is generated based on the real-time height of the pest.

[0096] In the specific implementation, the audio feature values ​​corresponding to different acquisition time periods in the regularized real-time audio signal are compared one by one with the real-time dynamic signal threshold. This point-by-point comparison method monitors for abnormal acoustic events that may be caused by pest activity in real time. If the audio feature value corresponding to at least one target acquisition time period is greater than the real-time dynamic signal threshold, the current real-time audio signal is determined to meet the pest detection criteria, indicating that pest activity exists near the location of the inspection unit during that time period. Based on this, and considering the target acquisition time period that meets the criteria, its position t in the acquisition sequence (e.g., the t-th acquisition time period), the preset fixed step size DM (i.e., vertical movement distance), and the acquisition duration TS, the data is processed according to the formula... By calculating the vertical displacement and combining it with the starting position of the grain pile inspection unit at a preset vertical height (such as the top of the temperature measuring cable as the origin), the actual real-time height of the pests can be obtained. This formula converts the time dimension into spatial depth, thereby obtaining the precise height of the pests on the vertical profile, determining the real-time height of the pests at the current moment, and combining it with the fixed horizontal coordinates of the inspection unit to generate pest event information containing information such as three-dimensional spatial position, collection time, and signal strength.

[0097] The grain pile pest identification method provided in this invention compares regularized audio feature values ​​with real-time dynamic thresholds one by one. When a feature value exceeds the threshold, pest activity is determined, and the real-time height of the pest is determined based on the target time period and corresponding vertical depth to generate event information. This method achieves accurate conversion from audio signals to spatiotemporal information of pests, and can capture abnormal acoustic events in real time and accurately locate their vertical positions.

[0098] Based on any of the above embodiments, the step of generating pest event information when determining that the real-time audio signal meets the pest determination criteria includes: When the real-time audio signal meets the pest determination criteria, initial pest event information is generated; The grain pile inspection unit 1-2 collects image data corresponding to the real-time audio signal and determines the image recognition result based on the image data. When the image recognition result and the initial pest event information indicate the presence of pests at the same vertical depth, the initial pest event information is confirmed as pest event information.

[0099] Understandably, generating pest event information through joint audio and video verification during real-time detection is a crucial mechanism for improving identification accuracy and reducing false alarm rates. Specifically, when the real-time audio signal meets the pest determination criteria, initial pest event information is first generated based on audio features, recording the three-dimensional location and occurrence time of the suspected pest activity. Simultaneously, the grain pile inspection unit 1-2 is controlled to collect image data corresponding to the real-time audio signal, and the images are analyzed using a pest image recognition algorithm to obtain image recognition results. Subsequently, the image recognition results are compared and verified with the initial pest event information. If both indicate the presence of pests at the same vertical depth, i.e., the audio features and visual features corroborate each other, the initial pest event information is confirmed as official pest event information and recorded and output.

[0100] The grain pile pest identification method provided in this invention generates pest event information through an audio-visual joint verification mechanism: when the audio meets the judgment conditions, initial event information is generated, and images are simultaneously acquired for identification. Confirmation is only made when pests are present at the same vertical depth indicated by the image and audio. This method combines the sensitivity of audio monitoring with the intuitiveness of visual recognition, effectively avoiding false alarms caused by environmental noise through double verification, significantly improving the reliability of pest detection, and providing an accurate basis for subsequent precise prevention and control.

[0101] Based on any of the above embodiments, it further includes: Acquire multiple pest event information confirmed by multiple grain pile inspection units 1-2 inside the grain pile, wherein each pest event information includes at least horizontal position coordinates and vertical depth coordinates; Based on a preset horizontal interval, the horizontal plane of the grain pile is divided into multiple grid units; Based on the vertical depth coordinates, pest event information at the same vertical depth is grouped into the same depth group data. Based on the horizontal position coordinates, determine the grid cell to which each pest event information belongs within each depth group of data; The horizontal coordinates of all pest events within the same depth group and belonging to the same grid cell are fused to obtain the corrected target horizontal coordinates.

[0102] In the specific implementation, firstly, pest event information confirmed by multiple grain pile inspection units 1-2 inside the grain pile is acquired. Each event information includes at least horizontal position coordinates and vertical depth coordinates, forming raw pest distribution data covering multiple spatial points inside the grain pile. Based on a preset horizontal interval, the horizontal plane of the grain pile is divided into multiple grid units, establishing a spatial reference system for subsequent cluster-based positioning correction. According to the vertical depth coordinates, pest event information at the same depth is grouped into the same depth group, realizing hierarchical dimensionality reduction processing from three-dimensional problems to two-dimensional planes. On this basis, the grid unit to which each pest event information in each depth group belongs is determined based on the horizontal position coordinates, completing preliminary spatial clustering. Finally, the horizontal position coordinates of all pest event information in the same depth group and belonging to the same grid unit are fused, for example, by taking the mean or weighted average of the coordinates, to obtain the corrected target horizontal position coordinates.

[0103] In the specific implementation, each pest event information includes at least horizontal position coordinates (px, py) and vertical depth coordinates pz, forming an original pest distribution dataset covering multiple spatial points within the grain pile. Based on a preset horizontal interval DT (i.e., the horizontal spacing of the temperature measuring cables), the horizontal plane of the grain pile is divided into multiple grid cells, with each grid cell corresponding to a cluster. According to the vertical depth coordinates, all pest events are grouped by depth, so that data at the same vertical depth are grouped into the same depth group, achieving dimensionality reduction of the three-dimensional localization problem to a two-dimensional plane.

[0104] For pest event information within the same depth group, clustering is performed according to the following formula: In the formula, DT is the horizontal spacing of the temperature measuring cable; This represents the set of all pest event information; IA(·) is an indicator function that indicates that when the Manhattan distance condition in parentheses is met, the pest event information is added to the cluster r(i,j,pz) corresponding to the current grid cell.

[0105] Understandably, this condition uses the center of the grid cell ((i+0.5)×DT, (j+0.5)×DT) as a reference, requiring that the Manhattan distance between the pest event and the center point does not exceed 0.5 (in DT, i.e., the actual distance does not exceed 0.5×DT), thereby assigning each pest event to the grid cell closest to it.

[0106] In the specific implementation, after completing the clustering of all grid cells, the horizontal coordinates of pest events in each non-empty cluster r(i,j,pz) are fused. The specific calculation formula is as follows: It should be noted that len(r(i,j,pz)) represents the number of pest events in the current cluster; the calculation result (px,py) is the corrected target horizontal position coordinate of the cluster, while the vertical depth pz remains unchanged.

[0107] For example: First, let's assume: =[(1,1,30),(1,2,30),(3,3,30),(4,4,30),(3,4,30),(4,3,30),(1,1,31)...], Select data with a uniform vertical height of pz=30, including: [(1,1,30),(1,2,30),(3,3,30),(4,4,30),(3,4,30),(4,3,30)] Next, the time information of pests at the same vertical height was clustered and calculated to obtain: r(0,0,30)=[(0,1,30)],r(1,1,30)=[(1,1,30),(1,2,30)]... Finally, the time information of pests in the same cluster was fused to calculate: =[(0,1,30), (1,1.5,30)...].

[0108] Understandably, based on the accurate three-dimensional spatial coordinates obtained after correction, grain depot managers can achieve targeted positioning and precise application of pesticides, directly placing pesticides or trapping devices into the grid units where pests gather. This transforms the traditional extensive control method of applying pesticides to the entire warehouse into localized and precise intervention. While significantly improving the control effect, this method effectively reduces the amount of pesticides used and the risk of grain contamination, lowers operating costs, and provides strong support for green grain storage and intelligent management.

[0109] In specific implementations, such as Figure 11 As shown, solid-lined circles represent temperature-sensing cables, and dashed-lined circles represent inspection units. Assuming the temperature-sensing cables are spaced 5 meters apart, the pest set data in the upper left corner is [(0,0),(0,5)], represented by q=0 / 1 in the figure. Four scenarios are illustrated: pests are found between two temperature-sensing cables (upper left example), between four temperature-sensing cables (lower left example), a pest is found on a single temperature-sensing cable (upper right example), and no pests are found on any temperature-sensing cable (lower right example), allowing for further horizontal positioning of the pests.

[0110] The grain pile pest identification method provided in this invention collects pest event information collaboratively by multiple inspection units, achieves dimensionality reduction of three-dimensional problems based on vertical depth hierarchical grouping, and then performs spatial clustering by dividing the grid units according to preset intervals, and fuses and corrects the horizontal position coordinates within the same grid. This method fully utilizes the advantages of array-based deployment, improving the pest location accuracy from "near a single temperature measuring cable" to grid-level precise coordinates, significantly refining the spatial resolution, and providing a reliable basis for accurate early warning and targeted control of grain pile pests.

[0111] Figure 12 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 12 As shown, the electronic device may include: a processor 1210, a communications interface 1220, a memory 1230, and a communication bus 1240, wherein the processor 1210, the communications interface 1220, and the memory 1230 communicate with each other through the communication bus 1240. The processor 1210 can call logical instructions in the memory 1230 to execute a grain pile pest identification method. The method includes: placing known pests at preset points near at least one grain pile inspection unit 1-2 inside the grain pile; collecting historical audio signals at different vertical depths during the vertical movement of the grain pile inspection unit 1-2; determining a threshold parameter combination based on the historical audio signals; collecting real-time audio signals at different vertical depths during the vertical movement of the grain pile inspection unit 1-2; determining whether the real-time audio signals meet the pest determination conditions based on the threshold parameter combination; and generating pest event information when the real-time audio signals meet the pest determination conditions.

[0112] Furthermore, the logical instructions in the aforementioned memory 1230 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0113] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the grain pile pest identification method provided by the above methods. The method includes: placing a known pest at a preset point near at least one grain pile inspection unit 1-2 inside the grain pile; collecting historical audio signals at different vertical depths during the vertical movement of the grain pile inspection unit 1-2; and determining a threshold parameter combination based on the historical audio signals; collecting real-time audio signals at different vertical depths during the vertical movement of the grain pile inspection unit 1-2; determining whether the real-time audio signals meet the pest determination conditions based on the threshold parameter combination; and generating pest event information when the real-time audio signals meet the pest determination conditions.

[0114] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the grain pile pest identification method provided by the above methods. The method includes: placing a known pest at a preset point near at least one grain pile inspection unit 1-2 inside the grain pile; collecting historical audio signals at different vertical depths during the vertical movement of the grain pile inspection unit 1-2; and determining a threshold parameter combination based on the historical audio signals; collecting real-time audio signals at different vertical depths during the vertical movement of the grain pile inspection unit 1-2; determining whether the real-time audio signals meet the pest determination conditions based on the threshold parameter combination; and generating pest event information when the real-time audio signals meet the pest determination conditions.

[0115] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0116] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A grain pile inspection structure, characterized in that, include: Multiple grain pile inspection units are respectively wrapped around multiple temperature measuring cables pre-embedded inside the grain pile, and each grain pile inspection unit moves up and down along the temperature measuring cables; The main controller is communicatively connected to each of the grain pile inspection units, and is used to issue movement commands or data collection commands to each of the grain pile inspection units, and to receive the detection data returned by each of the grain pile inspection units.

2. The grain pile inspection structure according to claim 1, characterized in that, The multiple temperature measuring cables are arranged in an array inside the grain pile, and each temperature measuring cable is wrapped with a grain pile inspection unit. The multiple grain pile inspection units form an inspection unit array inside the grain pile.

3. The grain pile inspection structure according to claim 1, characterized in that, The detection data includes audio data, image data, and location data; Each of the aforementioned grain pile inspection units includes: The body is wrapped around the temperature measuring cable; A camera, mounted on the wrapping body, is used to collect image data of the inside of the grain pile; An audio acquisition sensor is installed on the wrapping body to collect audio data inside the grain pile; A power unit, mounted on the wrapped body, is used to drive the grain pile inspection unit to move up and down along the temperature measuring cable; A grid photoelectric sensor is installed on the sleeve body to count the equidistant grids on the temperature measuring cable, determine the current vertical position based on the grid counting results, and generate position data. A communication module, disposed on the sleeve body, is communicatively connected to the main controller and is used to receive the movement command or acquisition command and send the audio data, image data and location data to the main controller.

4. A method for identifying pests in grain piles, characterized in that, The main controller applied in the grain pile inspection structure according to any one of claims 1-3 includes: After placing known pests at a predetermined point near at least one grain pile inspection unit inside the grain pile, historical audio signals at different vertical depths are collected as the grain pile inspection unit moves along the vertical direction, and a combination of threshold parameters is determined based on the historical audio signals. Real-time audio signals at different vertical depths are collected as the grain pile inspection unit moves along the vertical direction. Based on the combination of threshold parameters, determine whether the real-time audio signal meets the pest determination criteria; When the real-time audio signal is determined to meet the pest detection criteria, pest event information is generated.

5. The method for identifying grain pile pests according to claim 4, characterized in that, The acquisition of real-time audio signals at different vertical depths during the vertical movement of the grain pile inspection unit includes: The grain pile inspection unit is controlled to move vertically at a fixed step size. It remains stationary when it reaches a certain vertical depth and collects audio data for a preset duration at that vertical depth. The audio data collection at each vertical depth is completed sequentially to obtain the historical audio signal corresponding to each vertical depth.

6. The method for identifying grain pile pests according to claim 4, characterized in that, The step of determining the threshold parameter combination based on the historical audio signal includes: Define the frequency range set and the threshold fine-tuning parameter set. A preset frequency range is randomly selected from the set of frequency ranges, and a preset threshold fine-tuning parameter is randomly selected from the set of threshold fine-tuning parameters to form a candidate threshold parameter combination; Based on the preset frequency range, the historical audio signal is filtered to obtain the target historical audio signal; Based on the target historical audio signal and the preset threshold fine-tuning parameter, a dynamic signal threshold is set, and the predicted pest location is determined according to the dynamic signal threshold. Based on the preset locations and the predicted pest locations, the detection accuracy of the candidate threshold parameter combinations is calculated, and the candidate threshold parameter combination with the highest detection accuracy is determined as the threshold parameter combination.

7. The method for identifying grain pile pests according to claim 6, characterized in that, The step of setting a dynamic signal threshold based on the target historical audio signal and the preset threshold fine-tuning parameters includes: The target historical audio signal is subjected to regularization processing to obtain a regularized historical audio signal; Based on the regularized historical audio signal, a numerical sequence is generated, and the upper quartile is determined according to the numerical sequence. The dynamic signal threshold is determined based on the upper quartile and the preset threshold fine-tuning parameters.

8. The method for identifying grain pile pests according to claim 7, characterized in that, The step of performing regularization processing on the target historical audio signal to obtain a regularized historical audio signal includes: Extract the amplitude data of the target historical audio signal across all time periods to form the original amplitude sequence; Based on the original amplitude sequence, determine the lower quartile, minimum value, and maximum value; Based on the lower quartile, the minimum value, and the maximum value, the target historical audio signal is regularized to obtain a regularized historical audio signal.

9. The method for identifying grain pile pests according to claim 8, characterized in that, Determining the predicted pest location based on the dynamic signal threshold includes: The audio feature values ​​corresponding to different time periods in the regularized historical audio signal are compared one by one with the dynamic signal threshold. If there is at least one historical acquisition period whose corresponding audio feature value is greater than the dynamic signal threshold, then the historical audio signal is determined to meet the pest determination condition, and the historical height of the pest is determined based on the historical acquisition period and the corresponding vertical depth, and the predicted pest location is determined based on the historical height of the pest.

10. The method for identifying grain pile pests according to claim 4, characterized in that, The threshold parameter combination includes a target preset frequency range and a target preset threshold fine-tuning parameter; The step of determining whether the real-time audio signal meets the pest detection criteria based on the threshold parameter combination includes: Based on the target preset frequency range, the real-time audio signal is filtered to obtain the target real-time audio signal; The target real-time audio signal is subjected to regularization processing to obtain a regularized real-time audio signal; Based on the regularized real-time audio signal, a real-time numerical sequence is generated, and the real-time upper quartile is determined according to the real-time numerical sequence. Based on the real-time upper quartile and the target preset threshold fine-tuning parameters, the real-time dynamic signal threshold is determined; Based on the real-time dynamic signal threshold, determine whether the real-time audio signal meets the pest identification criteria.

11. The method for identifying grain pile pests according to claim 10, characterized in that, When the real-time audio signal is determined to meet the pest detection criteria, generating pest event information includes: The audio feature values ​​corresponding to different acquisition time periods in the regularized real-time audio signal are compared one by one with the real-time dynamic signal threshold. If at least one target acquisition time period has an audio feature value greater than the real-time dynamic signal threshold, the real-time audio signal is determined to meet the pest determination condition. Based on the target acquisition time period and the corresponding vertical depth, the real-time height of the pest is determined, and pest event information is generated based on the real-time height of the pest.

12. The method for identifying grain pile pests according to claim 4, characterized in that, When the real-time audio signal is determined to meet the pest detection criteria, generating pest event information includes: When the real-time audio signal meets the pest determination criteria, initial pest event information is generated; The grain pile inspection unit collects image data corresponding to the real-time audio signal and determines the image recognition result based on the image data. When the image recognition result and the initial pest event information indicate the presence of pests at the same vertical depth, the initial pest event information is confirmed as pest event information.

13. The method for identifying grain pile pests according to claim 4 or 12, characterized in that, Also includes: Acquire multiple pest event information confirmed by multiple grain pile inspection units inside the grain pile, wherein each pest event information includes at least horizontal position coordinates and vertical depth coordinates; Based on a preset horizontal interval, the horizontal plane of the grain pile is divided into multiple grid units; Based on the vertical depth coordinates, pest event information at the same vertical depth is grouped into the same depth group data. Based on the horizontal position coordinates, determine the grid cell to which each pest event information belongs within each depth group of data; The horizontal coordinates of all pest events within the same depth group and belonging to the same grid cell are fused to obtain the corrected target horizontal coordinates.

14. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the grain pile pest identification method as described in any one of claims 4 to 13.

Images