Target object targeting agent monitoring method and device based on AI visual identification

Abstract

The invention discloses a target object targeting agent monitoring method and device based on AI visual identification, belongs to the technical field of termite monitoring, and aims to solve the problems of high manual dependence degree, high misjudgment rate, low data transmission efficiency and the like in traditional target object monitoring. A slow-release targeting agent and a solar hybrid power supply system are arranged in the device, and low-power-consumption endurance is supported. According to the core scheme, vibration and infrared sensor signals are fused, and image acquisition is triggered through feature analysis; recognizing a target object through background difference, a multi-scale image and feature matching, and dynamically optimizing a detection threshold value; after video acquisition, activity intensity is calculated in combination with target tracking and gridding analysis, and transmission is optimized based on intensity and quota; light supplement is automatically adjusted according to ambient light, and image quality is ensured through pixel compensation and brightness adjustment. According to the scheme, a closed-loop system with multi-mode accurate triggering and intelligent identification is constructed, and the intelligence and efficiency of termite monitoring are improved.

Description

Technical Field

[0001] The present invention relates to the field of termite monitoring technology, and more specifically to a method and device for monitoring target objects with a targeted agent based on AI visual recognition. Background Art

[0002] Visual monitoring of termites in complex environments is an important technical direction in the field of intelligent sensing. The core challenge lies in overcoming the impact of lighting changes, background noise, and target dynamics on detection accuracy. Existing visual monitoring solutions generally have the following technical bottlenecks:

[0003] In the image acquisition and preprocessing stages, traditional equipment relies on fixed illumination compensation strategies. When faced with non-uniform illumination scenarios such as weak light and direct strong light, problems such as blurred termite outlines and insufficient contrast are prone to occur, resulting in the failure of subsequent feature extraction. Single-modal detection methods based on a single camera lack the ability to integrate and analyze auxiliary signals such as vibration and infrared, making it difficult to accurately separate targets in complex backgrounds. The missed detection rate and false alarm rate increase significantly with increasing environmental complexity. At the target recognition and behavior analysis level, existing algorithms mostly use a single-frame independent processing mode. This makes it impossible to use the spatiotemporal correlation information between consecutive frames to establish termite movement trajectories, making it difficult to assess the continuity and regularity of group activities. Feature extraction for dynamic termites relies on fixed parameter models, which have poor adaptability to termites of different shapes and scales. In particular, recognition accuracy plummets in scenarios where termites are partially occluded or their postures change. In the data processing and transmission link, a data optimization mechanism based on the characteristics of termite activity has not been established. The bandwidth usage is often too high due to redundant frame transmission, and key frames lose key details due to compression distortion. The real-time analysis module lacks adaptive adjustment capabilities, and the fixed threshold setting cannot balance the detection sensitivity and anti-interference ability in different monitoring scenarios. Frequent manual intervention and calibration are required, and the level of intelligence needs to be improved urgently.

[0004] The existing technology has significant deficiencies in multi-source information fusion strategies, adaptive feature extraction models, and intelligent transmission optimization mechanisms. Therefore, the present invention proposes a method and device for monitoring target-targeted agents based on AI visual recognition. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the purpose of the present invention is to provide a target targeting agent monitoring method and device based on AI visual recognition. Through real-time fusion of vibration and infrared signals for multimodal detection, target image recognition based on composite similarity, adaptive fill light adjustment and activity intensity-driven transmission optimization, the problems of environmental interference misjudgment, insufficient low-light recognition accuracy, data transmission redundancy and missed detection and false detection caused by fixed detection thresholds in traditional monitoring devices are solved, and the precise triggering, efficient recognition and energy-saving transmission of target activities in complex environments are achieved.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for monitoring a target agent based on AI visual recognition, comprising:

[0008] Step S1: Real-time collection and fusion of device vibration signals and infrared signals, extraction of vibration frequency characteristics, temperature gradient characteristics, and movement speed characteristics. By configuring a judgment threshold, it is determined whether the vibration signal and infrared signal are both effectively triggered. If so, an overlap analysis of the detection area is performed based on the vibration source phase difference positioning and the heat source triangulation coordinates. If there is an overlapping detection area, the camera is triggered to capture an image.

[0009] Step S2: Perform background difference processing on the captured image to extract the dynamic foreground area, segment the foreground target area, and determine whether there is a target object in the captured image through multi-scale image similarity matching combined with composite similarity calculation of shape features and texture features. If there is a target object, video capture is performed; and feedback adjustment of the judgment threshold is performed based on the target object recognition result;

[0010] Step S3: After starting video acquisition, the video frames are uploaded to the monitoring platform. The monitoring platform is used to extract target features from the video frames, generate speed sequence data based on target movement trajectory tracking, and obtain target activity intensity through grid distribution statistics and time window smoothing for transmission optimization.

[0011] Step S4: During image and video acquisition, the fill light brightness is automatically adjusted according to the ambient light. During image acquisition, pixel-level light intensity compensation is performed based on grayscale mean and contrast evaluation. During video acquisition, a smooth brightness gradient adjustment is used until the preset image quality threshold is reached.

[0012] Specifically, the steps of triggering the camera to capture images include:

[0013] The vibration sensor monitors the vibration signal of the device in real time, and the infrared sensor synchronously monitors the infrared signal of the heat source inside the device to form a multi-dimensional detection signal;

[0014] Target preprocessing is performed on vibration and infrared signals: the vibration signal is filtered through a bandpass filter to remove environmental noise and normalized; the infrared signal is distinguished from the background and dynamic heat source based on the Gaussian mixture model, and the temperature gradient and movement speed of the moving target are extracted;

[0015] The confidence level of the vibration signal is calculated based on the weighted fusion of the effective vibration number and frequency characteristics, and the confidence level of the infrared signal is calculated by mapping the ratio of temperature gradient to movement speed.

[0016] Configure the judgment thresholds, including vibration threshold and infrared threshold. When the vibration signal confidence is greater than the vibration threshold, it is determined that the vibration signal is effectively triggered. When the infrared signal confidence is greater than the infrared threshold, it is determined that the infrared signal is effectively triggered.

[0017] Specifically, the specific steps of triggering the camera to capture images also include:

[0018] When both the vibration signal and the infrared signal are effectively triggered, the detection area is divided according to the multimodal sensor layout. The vibration sensor determines the vibration source detection area through phase difference positioning, and the infrared sensor determines the heat source detection area through triangulation.

[0019] Determine whether the detection area effectively triggered by the vibration signal overlaps with the detection area effectively triggered by the infrared signal; if so, trigger the camera to capture the image; otherwise, no processing is performed;

[0020] Based on the recognition result of whether the target object exists in the image captured by the camera, feedback adjustment is performed on the judgment threshold, including:

[0021] If it is determined that there is no target object in the triggered image, the trigger judgment threshold is adjusted.

[0022] If it is determined that there is a target object in the image captured by the fixed image, and the vibration signal and the infrared signal are not effectively triggered, the feedback adjustment is triggered at the lower judgment threshold.

[0023] Specifically, the specific steps of determining whether there is a target object in the captured image include:

[0024] When the device is installed, the initial environment image is collected as the background image, the update cycle is set, the background update process is triggered, and the background image is updated;

[0025] After grayscale conversion and filtering for noise reduction, the collected image is differentiated from the background image pixel by pixel, and the absolute difference is calculated. The dynamic foreground area is selected by setting a dynamic threshold.

[0026] Adopting adaptive threshold segmentation algorithm to calculate target threshold of dynamic foreground area and segment the foreground target area;

[0027] The brightness of the foreground target area is adjusted, the size of the foreground target area is normalized, and then the similarity with the target database is calculated; the similarity between the foreground target area and the target database is composed of image similarity and feature similarity.

[0028] Specifically, the specific steps of determining whether the captured image contains a target object also include:

[0029] The image similarity of the target object in the captured image is determined by calculating the pixel-level similarity between the foreground target area and the image in the target database at multiple scales;

[0030] The calculation method for judging whether the target object feature similarity exists in the captured image is:

[0031] The shape and texture features of the foreground target area are extracted using a feature extraction algorithm. The shape feature similarity is obtained by calculating the Hu moment invariant Euclidean distance between the foreground target area and the target image in the target database.

[0032] The texture histogram is extracted using local binary patterns, and the texture similarity between the foreground target area and the target image in the target database is calculated using Bhattacharyya distance. Finally, the feature similarity is obtained through weighted summation.

[0033] Configure the image similarity threshold and feature similarity threshold. If the image similarity between the foreground target area and the image in the target database is greater than the image similarity threshold, or the feature similarity is greater than the feature similarity threshold, it is determined that there is a target in the dynamic foreground area of ​​the captured image.

[0034] Specifically, the specific steps of automatically adjusting the fill light brightness according to the ambient light include:

[0035] During the non-collection period, the ambient light intensity is measured, a light threshold is configured, and the ambient light intensity is compared with the light threshold for determining whether there is a target object in the collected image. If the ambient light intensity is greater than the light threshold, it is determined that there is sufficient natural light; otherwise, it is determined that there is insufficient natural light.

[0036] For triggering the camera to capture images or fixed images, if it is determined that there is sufficient natural light, the image will be captured directly using ambient light. If it is determined that there is insufficient natural light, the fill light adjustment process will be entered:

[0037] Set the basic fill light brightness and fill light duration interval, fill light at the basic fill light brightness and capture the image within the fill light duration interval when determining whether the target object exists in the captured image, and obtain the grayscale mean and contrast of the captured image;

[0038] Configure the grayscale threshold and contrast threshold. If the grayscale mean of the captured image is greater than the grayscale threshold and the contrast is greater than the contrast threshold, it is judged as a valid image; otherwise, it is judged as an invalid image.

[0039] If it is an invalid image, set the fill light brightness step and gradually increase the fill light brightness step based on the basic fill light brightness. Re-capture the image within the fill light interval, extract the dynamic area coordinates of the valid image and the invalid image through background difference, and replace the dynamic area pixel values ​​with the interpolation results of the two pixel values ​​to merge and generate a valid image.

[0040] Specifically, the specific steps of automatically adjusting the fill light brightness according to the ambient light also include:

[0041] For video capture, if it is determined that there is sufficient natural light, the video will be captured directly using ambient light. If it is determined that there is insufficient natural light, the dynamic fill light adjustment process will be entered:

[0042] Starting from the ambient light intensity, the fill light brightness is gradually and smoothly increased. The grayscale mean and contrast of the video frame are analyzed in real time. When the grayscale mean of the video frame is greater than the grayscale threshold and the contrast is greater than the contrast threshold, the dynamic fill light adjustment is stopped and the current fill light brightness is maintained to complete the video acquisition.

[0043] Specifically, the steps for obtaining the target object's activity intensity and performing transmission optimization include:

[0044] Enable the adaptive contrast enhancement algorithm for the video frames received by the monitoring platform, configure the detection confidence threshold, identify the target objects in the video frames and mark their bounding boxes, and output the target object coordinates and timestamp information;

[0045] Configure the trajectory matching threshold, bind the target IDs in consecutive video frames, and generate a trajectory sequence. Configure the time window and calculate the average speed of the target within the time window length based on the trajectory sequence.

[0046] Divide the video frame into grids, count the number of targets in each grid, and map the difference in number between grids into a clustering index;

[0047] Configure the weight ratio of individual number, movement speed, and aggregation, calculate the activity intensity index by weight, and smooth the activity intensity index within the configured smoothing window;

[0048] Configure low and high activity intensity thresholds. When the smoothed activity intensity index is less than the low activity intensity threshold, the video capture pause instruction is triggered.

[0049] Configure a single data quota. When the number of received video frames exceeds the single data quota and the activity intensity index exceeds the high activity intensity threshold, data compression mode is triggered:

[0050] Calculate the rate of change between video frames through pixel differences or feature vector differences, mark key frames and remove redundant frames, and upload key frames to the monitoring platform.

[0051] Specifically, the specific steps of feedback adjustment of the threshold for determining whether a target object exists in the captured image include:

[0052] When the trigger judgment threshold is adjusted, the confidence level of the detected vibration signal and infrared signal is recorded, and the false trigger count is increased;

[0053] Configure the false touch threshold. If the false trigger count is greater than the false touch threshold, it is determined to be a high-frequency false trigger and the feedback adjustment is frozen.

[0054] Configure the adjustment step, including vibration adjustment step and infrared adjustment step. Add the corresponding adjustment step based on the current vibration threshold and infrared threshold to increase the judgment threshold.

[0055] When the feedback adjustment is triggered under the judgment threshold, the image for which the feedback adjustment is triggered under the judgment threshold is marked as a missed detection sample, and the missed detection count is increased;

[0056] Configure the missed detection threshold. When the missed detection count is greater than the preset missed detection threshold, the detection sensitivity is determined to be insufficient. Based on the current vibration threshold and infrared threshold, the corresponding adjustment step is reduced to lower the judgment threshold.

[0057] A target-targeting agent monitoring device based on AI visual recognition, consisting of a trapping module, an intelligent power supply module, a visual recognition module, and an Internet of Things transmission module;

[0058] The trapping module includes a bionic nest structure shell and a targeting agent, which is used to attract the target object to enter and achieve group killing through the targeting agent;

[0059] The intelligent power supply module is used to provide power support through hybrid power supply and intelligent energy consumption regulation;

[0060] The visual recognition module is used to identify the target object and dynamically monitor its activity status;

[0061] The Internet of Things transmission module is used for remote monitoring and information feedback of the target object.

[0062] Beneficial effects of the present invention:

[0063] The present invention effectively solves the problem of false triggering caused by environmental noise interference in single sensors by extracting features from vibration and infrared signals, calculating confidence, and determining area overlap, significantly improving the accuracy of image acquisition triggering. A target recognition algorithm based on multi-scale image similarity and composite feature matching, combined with dynamic background updating and adaptive threshold segmentation, overcomes the bottleneck of target recognition under low or complex lighting conditions and reduces the missed detection rate. Adaptive ambient light fill adjustment, pixel-level light intensity compensation, and smooth brightness gradient adjustment ensure consistent image quality in different lighting environments and avoid feature loss due to insufficient brightness. The activity intensity-based transmission optimization strategy, including trajectory tracking, grid analysis, and keyframe screening, reduces redundant data transmission, network load, and system power consumption while ensuring the integrity of critical termite data. Furthermore, the threshold feedback adjustment mechanism dynamically adapts detection sensitivity to the environment through self-learning of false triggering and missed detection samples, reducing the false positive rate caused by traditional fixed thresholds. Overall, the present invention significantly improves the intelligent level and control efficiency of termite monitoring in complex environments, providing reliable technical support for early warning and precise extermination of termite infestations. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a flow chart of a method for monitoring a target agent based on AI visual recognition according to the present invention;

[0065] Figure 2 This is a flowchart of the specific steps of triggering the camera to capture images in the present invention;

[0066] Figure 3 This is a flowchart of the specific steps of the present invention for determining whether a target object exists in a captured image;

[0067] Figure 4 A flowchart of the specific steps for obtaining the target object's activity intensity and performing transmission optimization according to the present invention;

[0068] Figure 5 This is a structural schematic diagram of a target object targeting agent monitoring device based on AI visual recognition in the present invention. DETAILED DESCRIPTION

[0069] Example 1

[0070] See also Figure 1This embodiment introduces a method for monitoring target objects with a targeted agent based on AI visual recognition. The target objects of this invention are insects, including but not limited to social insects such as termites and ants. When termites are the target, a bionic nest-like shell with a built-in slow-release pheromone trap is used. This trap releases termite pheromones that attract the target ant species and attract them to feed. The pheromone trap is loaded with a low-toxic, biodegradable bait, which is delivered to the core of the nest through trophallaxis, achieving swarm extermination. The trapping radius covers a range of 5 to 10 meters, achieving both targeted trapping and environmentally friendly extermination.

[0071] In this embodiment, a bionic nest structure shell is used. The shell is injection-molded from a polymer composite material. The surface simulates the rough texture and arched structure of a natural termite nest, and has the characteristics of anti-crushing and anti-water immersion. A spiral channel is set on the inner wall of the shell. The width of the channel matches the target ant species, guiding the termites along the channel into the core trapping area, while using the light-shielding property of the structure to reduce their alertness; the built-in slow-release pheromone trapping rod is targeted at target ant species such as Reticulitermes and Coptotermes, and uses artificially synthesized aggregation pheromones such as farnesol derivatives and trace pheromones such as terpenoid compounds. The trapping is carried out by gas chromatography-mass spectrometry. Technology is used to optimize the ratio to ensure the attraction rate to specific ant species. The pheromones are wrapped in biodegradable microcapsules made of chitosan. The microcapsules gradually dissolve and release pheromones in the temperature range of termite activity through a temperature-triggered release mechanism. The pheromone trap sticks are loaded with environmentally friendly baits supplemented with natural carriers such as starch and cellulose. After feeding, the worker ants pass the agents to the larvae, queen ants and king ants through oral secretions, and the feeding behavior of termite communities is used to achieve the spread of the poison. Through bionic structure design, precise pheromone ratios and low-toxic slow-release technology, targeted trapping and nest-level killing of target ant species can be achieved.

[0072] A hybrid power supply system of solar panels and lithium batteries is adopted. It automatically charges when there is sufficient sunlight, and automatically enters sleep mode when not in operation, achieving low power consumption and long-term endurance. In this embodiment, a hybrid power supply system of solar panels and lithium batteries is adopted. The solar panels are treated with an anti-fouling coating and can automatically charge the lithium batteries under light conditions. The matching lithium batteries are adapted to a wide temperature working environment and have an integrated intelligent adjustment chip. They can automatically optimize the charging efficiency according to the light intensity and can maintain the continuous operation of the device in a dark environment. In the non-operating state, the target object activity signal is detected by the sensor, and the device automatically enters sleep mode when there is no activity, retaining only the basic monitoring function, significantly reducing the energy consumption of the entire machine and achieving low power consumption and long-term endurance.

[0073] The camera performs timed image acquisition, and the image is analyzed and identified through AI algorithms to determine whether there is a target object in the captured image. At the same time, vibration sensors and infrared sensors are integrated to trigger the camera to capture images, determine whether it is triggered by a target object to reduce unnecessary power consumption, and eliminate interference from non-target organisms. When it is determined that there is a target object, video acquisition is automatically triggered, and the video recording function is turned on to record the target object's feeding, movement and other behaviors, which are used to analyze the group activity and population changes. When acquiring images and videos, the fill light brightness is automatically adjusted according to the ambient light to avoid interfering with the target object's behavior. It automatically switches to infrared mode in low light or at night to ensure image clarity, and analyzes images in real time locally on the device. The video acquisition and communication link are woken up only when the target object is detected. When not in operation, the camera and processor enter sleep mode to achieve low power operation.

[0074] See also Figure 2 Preferably, the specific steps of triggering the camera to capture images include:

[0075] The vibration sensor monitors the vibration signal of the entire circumference of the device shell in real time, and the infrared sensor synchronously monitors the infrared signal of the heat source within the device to form a multi-dimensional detection signal;

[0076] Perform target pre-processing on vibration signals and infrared signals, including removing environmental noise through a bandpass filter and normalizing the vibration signals to highlight the regular vibration signals generated by the target's crawling and eating;

[0077] For infrared signals, the background heat source and dynamic heat source are distinguished based on the Gaussian mixture model, the temperature gradient and moving speed of the moving target are extracted, and jumping moving targets such as beetles are excluded.

[0078] The confidence of the vibration signal and infrared signal after preprocessing of the target object is calculated. The confidence of the vibration signal is calculated based on the weighted fusion of the effective vibration number and frequency characteristics, and the confidence of the infrared signal is calculated based on the proportional mapping of the temperature gradient and the moving speed. For example, the confidence of the vibration signal The calculation formula is:

[0079]

[0080] in, is the number of vibration signal acquisitions, The number of signals whose amplitude after bandpass filtering is greater than the preset amplitude threshold. The amplitude threshold is used to filter out environmental micro-vibration interference and retain the effective mechanical vibration signal generated by the target object's activity. It is the cosine similarity between the vibration signal frequency and the target's characteristic frequency. The target's characteristic frequency is the main frequency of the target's common species of moth-eating vibration. It is used to quantify the degree of match between the signal spectrum and the known target's activity frequency characteristics, eliminating non-target frequency interference. It is a preset upper limit of characteristic frequency, which is used to limit the cutoff frequency of the bandpass filter to prevent high-frequency noise from affecting the confidence calculation;

[0081] Infrared signal confidence The calculation formula is:

[0082]

[0083] in, is the temperature difference between the heat source and the environment, is the ambient temperature fluctuation range, is the measured heat source moving speed, is the typical crawling speed of the target;

[0084] Configure the judgment threshold, including vibration threshold and infrared threshold. When the vibration signal confidence is greater than the vibration threshold, it is judged as a valid vibration signal trigger. When the infrared signal confidence is greater than the infrared threshold, it is judged as a valid infrared signal trigger.

[0085] When both the vibration signal and the infrared signal are effectively triggered, the internal detection area of ​​the device is divided according to the multimodal sensor layout. The vibration sensor determines the vibration source detection area through phase difference positioning, and the infrared sensor determines the heat source detection area through triangulation.

[0086] Determine whether the detection area effectively triggered by the vibration signal overlaps with the detection area effectively triggered by the infrared signal; if so, trigger the camera to capture images to reduce invalid detections; otherwise, no processing is performed;

[0087] When the camera is triggered to capture an image, it captures the image inside the device and uses edge computing to identify and determine whether there is a target object in the captured image;

[0088] Based on the recognition result of whether the camera captures the target object, feedback is provided to adjust the image acquisition trigger condition, that is, feedback adjustment is provided to the judgment threshold, including:

[0089] If it is determined that there is no target in the triggered image, the trigger threshold is adjusted, the vibration signal confidence and infrared signal confidence of this detection are automatically recorded, and the false trigger count is increased, with its initial value being 0;

[0090] Configure a false touch threshold. If the false trigger count exceeds the threshold, it is considered a high-frequency false trigger. Feedback adjustments are temporarily frozen to prevent frequent adjustments in a short period of time from affecting stability.

[0091] Configure the adjustment step, including vibration adjustment step and infrared adjustment step. Add the corresponding adjustment step based on the current vibration threshold and infrared threshold respectively, and upload the adjusted judgment threshold and sensor parameters to the monitoring platform synchronously, including the confidence formula parameters, environmental data and historical false touch waveforms before and after adjustment, for manual review.

[0092] If it is determined that there is a target object in the image captured by the fixed image, and the vibration signal and the infrared signal are not effectively triggered, the feedback adjustment under the trigger judgment threshold is triggered, and the image is marked as a missed detection sample, and the missed detection count is increased, with its initial value being 0;

[0093] Configure the missed detection threshold. When the missed detection count is greater than the preset missed detection threshold, the detection sensitivity is determined to be insufficient. The thresholds are lowered based on the current vibration threshold and infrared threshold according to the adjustment step size, that is, the corresponding adjustment step size is reduced based on the current vibration threshold and infrared threshold. Relevant detection data, including missed detection images and recognition results, sensor original signal records, and comparison of threshold parameters before and after adjustment, will be sent to the monitoring platform for manual review.

[0094] Preferably, the specific steps of automatically adjusting the fill light brightness according to the ambient light include:

[0095] During the non-collection period, the system enters a low-power sleep mode, turns off the light-emitting components, and measures the ambient light intensity through a light-sensitive sensor.

[0096] Configure a light threshold. If the ambient light intensity is greater than the light threshold, such as light penetrating through soil cracks or diffuse light reflected on rainy days, it is considered sufficient natural light. Otherwise, it is considered insufficient natural light.

[0097] When triggering the camera to capture images or fixed images, if it is determined that there is sufficient natural light, the image is directly captured using ambient light; if it is determined that there is insufficient natural light, the fill light adjustment process is entered, namely:

[0098] Set the basic fill light brightness and configure the fill light duration interval, which should be shorter than the target's visual reaction time. Within the fill light duration interval, use the basic fill light brightness to fill the image and obtain the grayscale mean and contrast of the captured image.

[0099] Configure image quality thresholds, including grayscale thresholds and contrast thresholds. If the grayscale mean of the acquired image is greater than the grayscale threshold and the contrast is greater than the contrast threshold, the acquired image is retained and determined to be a valid image to further determine whether the acquired image contains a target object. Otherwise, it is determined to be an invalid image.

[0100] If the image is determined to be invalid, the fill light brightness step size is set, and the fill light brightness step size is gradually increased at the basic fill light brightness. During the fill light duration interval, image acquisition is performed. If the image is determined to be valid, the dynamic area coordinates in the invalid image are extracted by background difference, and the pixel values ​​of the dynamic area are replaced with the pixel value interpolation result of the valid image and the invalid image, and the valid image is generated by merging. For a single frame of invalid image, it is not directly deleted, but the information is merged with the valid images of the previous and next frames. The images with contrast less than or equal to the contrast threshold are merged and eliminated to reduce memory usage and retain the acquisition information.

[0101] When capturing video, if it is determined that there is sufficient natural light, the video is captured directly using ambient light; if it is determined that there is insufficient natural light, the dynamic fill light adjustment process is entered, namely:

[0102] Starting from ambient light intensity, the system gradually and smoothly increases the fill light brightness while simultaneously analyzing the grayscale mean and contrast of the video frame in real time. When the grayscale mean of the video frame exceeds the grayscale threshold and the contrast exceeds the contrast threshold, dynamic fill light adjustment stops and the current fill light brightness is maintained to complete video capture.

[0103] See also Figure 3 Preferably, the specific steps of determining whether a target object exists in the captured image include:

[0104] During device installation, an initial environmental image is captured as the background image. Static features such as the device's interior wall texture and structure are extracted to build a background pixel distribution database. An update cycle is set to trigger the background update process during periods when no target objects are active. This iterative update of the background image using an incremental Gaussian mixture model eliminates long-term environmental interference such as dust accumulation and slight lighting changes.

[0105] Perform grayscale conversion and median filtering on the active image. Difference the processed active image with the background image pixel by pixel, calculating the absolute difference. Set a dynamic threshold, such as the mean plus or minus twice the standard deviation, and retain pixels with absolute differences outside the dynamic threshold range. These pixels constitute the dynamic foreground area.

[0106] The dynamic foreground candidate area may contain residual background, such as reflections and semi-transparent stains. The adaptive threshold segmentation algorithm is used to calculate the target threshold of the dynamic foreground area, thereby segmenting the foreground target area and the residual background and outputting a binary mask.

[0107] Different lighting changes can cause foreground targets to be overexposed or underexposed, resulting in loss of details. For the foreground target area, adaptive histogram equalization is used to adjust the brightness while limiting the contrast enhancement range to avoid feature distortion caused by overexposure or underexposure.

[0108] The segmented foreground object region is normalized to ensure it matches the image size in the database, eliminating size variations caused by distance, such as changes in image size as the object approaches or moves away from the camera. The bounding box coordinates are retained for subsequent localization. Similarity is then calculated against the in-memory object database, which contains images of objects of the same size from different angles and shapes.

[0109] The similarity between the foreground target area and the target database consists of image similarity and feature similarity.

[0110] Pixel similarity at a single scale is easily affected by local noise. Image similarity is determined by calculating the pixel-level similarity between the foreground target area and the image in the target database at multiple scales. The highest value is selected as the image similarity to capture features from the overall outline to local details. Pixel-level similarity can use a multi-scale structural similarity index.

[0111] Feature similarity is calculated by using a feature extraction algorithm to extract the shape and texture features of the foreground target region. Shape similarity is then calculated by calculating the Hu moment-invariant Euclidean distance between the foreground target region and the target image in the target database. This method can stably describe key morphological features of the target, such as the beaded antennae and constricted waist and abdomen, as well as the elliptical outline of Coleoptera insects.

[0112] The texture histogram is extracted using local binary patterns. The Bhattacharyya distance is used to calculate the texture similarity between the foreground target area and the target image in the target database. Finally, the feature similarity is obtained through weighted summation. The subtle ripples on the target's surface, such as segment texture, are captured. Histogram comparison is used to quantify texture differences and exclude non-target organisms with smooth surfaces, such as ants.

[0113] Configure the image similarity threshold and feature similarity threshold, taking into account both detection sensitivity and specificity. If the image similarity between the foreground target area and the image in the target database is greater than the image similarity threshold, or the feature similarity is greater than the feature similarity threshold, then it is determined that the target exists in the dynamic foreground area of ​​the valid image; otherwise, it is determined that the target does not exist in the dynamic foreground area of ​​the valid image.

[0114] When video acquisition is started, the real-time video frames are uploaded to the monitoring platform; the monitoring platform extracts target object features from the received video frames, including the number of individuals, movement trajectories, group aggregation, etc., obtains the target object activity intensity, and optimizes transmission: if the target object activity intensity is detected to be lower than the preset threshold, a command to pause video acquisition is automatically sent to save power consumption; if the amount of transmitted data exceeds the daily quota, the data compression mode is triggered based on the target object activity intensity, and key frames are transmitted first; during non-transmission periods, the IoT transmission module enters sleep mode, retaining only the wake-up signal monitoring function.

[0115] See also Figure 4 Preferably, the specific steps of obtaining the target object activity intensity and performing transmission optimization include:

[0116] The target monitoring environment may have weak light or uneven lighting, which will cause the target outline to be blurred and affect the subsequent recognition accuracy. Therefore, the monitoring platform uses an adaptive contrast enhancement algorithm on the received video frames to improve the clarity of the target outline in low-light environments by adjusting the brightness gain parameter.

[0117] In order to balance detection accuracy and efficiency under the limited computing power of edge devices and reduce false detection of non-target objects, a detection confidence threshold is configured, and a lightweight target detection model is used to identify individual targets in video frames and annotate their bounding boxes, and output target coordinates and timestamp information.

[0118] Analyzing video frames individually cannot reveal the continuity of target behavior. To establish individual motion trajectories and analyze behavioral patterns, a multi-target tracking algorithm is enabled and trajectory matching thresholds, including spatial distance thresholds and time interval thresholds, are configured. Targets in consecutive video frames are then ID-bound to generate a motion trajectory sequence. The spatial distance threshold is used to measure the spatial difference between target positions in different video frames. When the distance between the position of a target in the current frame and the position in the previous frame falls within this spatial distance threshold, they are considered the same target. The time interval threshold takes into account the continuity of the target in the temporal dimension. It specifies the time interval range within which targets appear in consecutive video frames. For example, a time interval threshold of 0.5 seconds means that if a target appears in the current frame and its appearance in the previous frame is no more than 0.5 seconds away from the current frame, they are considered the same target.

[0119] Single-frame speed is easily affected by noise. Continuous multi-frame averaging can smooth out fluctuations and reflect the actual movement state. A time window is configured to calculate the target's movement speed in multiple consecutive frames. Based on the movement trajectory sequence, the average movement speed of the target within the time window length is calculated.

[0120] The degree of clustering of target objects is a key indicator for determining the hazard level. Direct counting is difficult to quantify the spatial distribution characteristics. The video frame is divided into grids, and the number of targets within each grid is counted.

[0121] The density variance algorithm is used to analyze the quantity differences between grids. Combined with the adjustable Gaussian kernel function parameters, the distribution differences are mapped into a clustering index in the [0,1] interval. The larger the value, the denser the clustering.

[0122] The activity intensity index is obtained by configuring the weight ratio of the number of individuals, movement speed, and aggregation degree;

[0123] Real-time calculations are easily affected by accidental factors, resulting in frequent fluctuations. A smoothing window is configured to set the time range for filtering processing. The time window filtering function is enabled to smooth the activity intensity index calculated in real time within the smoothing window length to reduce single-frame noise interference.

[0124] Automatically identify periods without targets or with minimal activity to save energy. Configure low and high activity intensity thresholds. When the smoothed activity intensity index is less than the low activity intensity threshold, the video capture pause command is triggered.

[0125] To prioritize the transmission of high-value data under limited bandwidth, a single data quota is configured. When the number of video frames received by the monitoring platform exceeds the single data quota and the smoothed activity intensity index exceeds the high activity intensity threshold, data compression mode is triggered, and key frames are transmitted first:

[0126] Configure the change rate threshold. When collecting video, calculate the change rate between the current video frame and the previous video frame through pixel difference method or feature vector difference. If it is greater than the change rate threshold, the current video frame is marked as a key frame. Otherwise, it is judged as a redundant frame and eliminated. Set the interval between adjacent key frames. For example, it can be configured to be at least 5 frames apart to avoid continuous transmission of overly similar frames. Upload the video frames marked as key frames to the monitoring platform to obtain the activity intensity of the target object.

[0127] Example 2

[0128] See also Figure 5This embodiment introduces a device for monitoring a target with a targeted agent based on AI visual recognition. The device consists of a trapping module, an intelligent power supply module, a visual recognition module, and an IoT transmission module. The trapping module is the core component of the device for achieving targeted trapping. Its core components include a targeting agent, which is a key element in attracting the target. In this embodiment, the targeting agent is a slow-release pheromone, which is placed on a trapping rod to form a slow-release pheromone trapping rod. The trapping module forms a close synergy with the targeting agent through a bionic nest-like shell. The shell simulates a natural habitat to reduce the target's alertness, while the trapping rod loaded with the targeting agent achieves precise release and efficient trapping of the pheromone, thereby attracting social insects such as termites and ants. The intelligent power supply module provides stable power to the entire device, integrating low-power technologies for long-term battery life and ensuring continuous operation in complex environments. The visual recognition module relies on multi-sensor fusion perception and AI algorithm analysis to accurately identify targets and dynamically monitor their activity, effectively eliminating interference from non-target organisms. The IoT transmission module handles data exchange between the device and the remote platform, enabling remote monitoring and information feedback of targets, enabling real-time uploading of monitoring data and precise issuance of control commands. These modules work together to achieve targeted monitoring and elimination of target social insects, combining the remarkable characteristics of strong targeting, environmental safety, and a high degree of intelligence.

[0129] The trapping module, the core trapping and killing component, comprises a biomimetic nest-like shell and a targeting agent. These two components work together to attract target ants and, with the help of the targeting agent, achieve swarm extermination. The biomimetic nest shell is injection-molded from a polymer composite material. Its surface simulates the rough texture and arched structure of the target ant species' natural nests, making it resistant to crushing and water immersion. The inner wall features a spiral channel tailored to the target ant species' size, guiding them into the core area while utilizing the structure's light-shielding properties to reduce their alertness. A slow-release pheromone-based targeting agent, a slow-release pheromone trapping rod, is embedded within the core of the biomimetic nest shell. This targeted agent, a combination of synthetic aggregation and tracking pheromones, is encapsulated in biodegradable chitosan microcapsules. A temperature-triggered mechanism gradually releases the bait into the target ant's active temperature range. The bait, based on natural materials such as starch and cellulose, utilizes the target ant's trophallaxis to disperse the toxicant within the swarm.

[0130] The intelligent power supply module is electrically connected to the decoy module, visual recognition module, and IoT transmission module. It provides power support through hybrid power supply and intelligent energy consumption regulation, achieving low power consumption and long-term endurance. It includes solar panels, lithium batteries, and an intelligent regulation chip. The solar panels are treated with an anti-fouling coating and can charge the lithium batteries under sunlight conditions. The lithium batteries are adapted to a wide operating temperature range and, in conjunction with the intelligent regulation chip, optimize charging efficiency based on light intensity. It also integrates a sleep-wake mechanism, automatically entering sleep mode when there are no target activity signals, retaining only basic monitoring functions. It automatically charges when there is sufficient sunlight, reducing energy consumption when not in operation, ensuring the device can continue to operate in dark environments.

[0131] The visual recognition module is electrically connected to the trapping module and the intelligent power supply module. Through image acquisition, sensor triggering, and AI algorithm analysis, it achieves accurate target identification, behavioral recording, and dynamic monitoring. It includes a camera, vibration sensor, infrared sensor, fill light component, and edge computing unit. The vibration sensor monitors the vibration signal of the device casing in real time, and the infrared sensor simultaneously monitors the infrared signal of the internal heat source. The fusion of the two triggers the camera to capture images, reducing unnecessary power consumption. The camera can perform timed image acquisition and triggered video acquisition. During acquisition, the fill light component automatically adjusts the brightness according to the ambient light, switching to infrared mode in low light or at night. The edge computing unit has a built-in AI algorithm that analyzes images locally in real time, identifies the target, and records its feeding, movement, and other behaviors to determine the activity level and population changes of the group.

[0132] The IoT transmission module is electrically connected to the visual recognition module and the intelligent power supply module, responsible for data transmission between the device and the external control system, enabling remote monitoring and information feedback. It receives target activity data from the visual recognition module, including images, videos, and quantitative analysis results, and uploads them to the monitoring platform. It also optimizes transmission strategies based on the intensity of target activity and data volume, prioritizing the transmission of key frames and entering sleep mode during non-transmission periods. Furthermore, it receives basic control parameters from the external control system and feeds back operating status information such as power consumption, recognition results, and fault indications to the control system, enabling remote management and intelligent scheduling.

[0133] Working principle and its effect:

[0134] The present invention uses a multimodal sensor to collect device vibration signals and infrared signals in real time, extracts characteristics such as vibration frequency, temperature gradient and target movement speed through preprocessing such as bandpass filtering and Gaussian mixture model, calculates signal confidence through weighted fusion and proportional mapping, combines phase difference positioning and triangulation to divide the detection area, and triggers image acquisition only when the vibration and infrared signals are effectively triggered and the detection areas overlap, effectively filtering out environmental noise interference and significantly reducing the false trigger rate.

[0135] The dynamic foreground is extracted from the captured image through background difference, and the target object is identified using a composite comparison algorithm that combines multi-scale pixel similarity with Hu rectangular features and local binary pattern texture features. At the same time, the background image is automatically updated during inactive periods, which solves the problem of target blur in low light or complex lighting and greatly improves the recognition accuracy.

[0136] In the fill light adjustment stage, the fill light brightness is dynamically adjusted according to the ambient light intensity. Pixel-level light intensity compensation is performed through grayscale mean and contrast evaluation during image acquisition. Smooth brightness gradient adjustment is used during video acquisition to ensure image quality consistency under different lighting conditions and avoid feature loss due to sudden brightness changes.

[0137] The monitoring platform tracks targets in video frames to generate movement trajectories. It then calculates an activity intensity index using grid-based distribution statistics and a time-window smoothing algorithm. When data volume exceeds the quota, it filters key frames for transmission based on the inter-frame change rate. This ensures valid information while significantly reducing redundant data transmission, lowering network load and power consumption. A threshold feedback adjustment mechanism, which self-learns from missed detections and false triggers, dynamically optimizes vibration and infrared signal thresholds, enabling detection sensitivity to adapt to environmental changes and effectively reducing false alarm and missed detection rates.

[0138] Overall, the present invention constructs a target object monitoring system with low misjudgment, high robustness and strong adaptability, providing an efficient technical path for early warning and precise extermination of ant pests.

[0139] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for monitoring a target agent based on AI visual recognition, characterized in that: include: Step S1: Real-time collection and fusion of device vibration signals and infrared signals, extraction of vibration frequency characteristics, temperature gradient characteristics, and movement speed characteristics. By configuring a judgment threshold, it is determined whether the vibration signal and infrared signal are both effectively triggered. If so, an overlap analysis of the detection area is performed based on the vibration source phase difference positioning and the heat source triangulation coordinates. If there is an overlapping detection area, the camera is triggered to capture an image. Step S2: Perform background difference processing on the captured image to extract the dynamic foreground area, segment the foreground target area, and determine whether there is a target object in the captured image through multi-scale image similarity matching combined with composite similarity calculation of shape features and texture features. If there is a target object, video capture is performed; and feedback adjustment of the judgment threshold is performed based on the target object recognition result; Step S3: After starting video acquisition, the video frames are uploaded to the monitoring platform. The monitoring platform is used to extract target features from the video frames, generate speed sequence data based on target movement trajectory tracking, and obtain target activity intensity through grid distribution statistics and time window smoothing for transmission optimization. Step S4: During image and video acquisition, the fill light brightness is automatically adjusted according to the ambient light. During image acquisition, pixel-level light intensity compensation is performed based on grayscale mean and contrast evaluation. During video acquisition, a smooth brightness gradient adjustment is used until the preset image quality threshold is reached.

2. The method for monitoring a target agent based on AI visual recognition according to claim 1, wherein: The specific steps of triggering the camera to collect images include: The vibration sensor monitors the vibration signal of the device in real time, and the infrared sensor synchronously monitors the infrared signal of the heat source inside the device to form a multi-dimensional detection signal; Target preprocessing is performed on vibration and infrared signals: the vibration signal is filtered through a bandpass filter to remove environmental noise and normalized; the infrared signal is distinguished from the background and dynamic heat source based on the Gaussian mixture model, and the temperature gradient and movement speed of the moving target are extracted; The confidence level of the vibration signal is calculated based on the weighted fusion of the effective vibration number and frequency characteristics, and the confidence level of the infrared signal is calculated by mapping the ratio of temperature gradient to movement speed. Configure the judgment thresholds, including vibration threshold and infrared threshold. When the vibration signal confidence is greater than the vibration threshold, it is determined that the vibration signal is effectively triggered. When the infrared signal confidence is greater than the infrared threshold, it is determined that the infrared signal is effectively triggered.

3. The method for monitoring a target-targeting agent based on AI visual recognition according to claim 2, wherein: The specific steps of triggering the camera to collect images also include: When both the vibration signal and the infrared signal are effectively triggered, the detection area is divided according to the multimodal sensor layout. The vibration sensor determines the vibration source detection area through phase difference positioning, and the infrared sensor determines the heat source detection area through triangulation. Determine whether the detection area effectively triggered by the vibration signal overlaps with the detection area effectively triggered by the infrared signal; if so, trigger the camera to capture the image; otherwise, no processing is performed; Based on the recognition result of whether the target object exists in the image captured by the camera, feedback adjustment is performed on the judgment threshold, including: If it is determined that there is no target object in the triggered image, the trigger judgment threshold is adjusted. If it is determined that there is a target object in the image captured by the fixed image, and the vibration signal and the infrared signal are not effectively triggered, the feedback adjustment is triggered at the lower judgment threshold.

4. The method for monitoring a target-targeted agent based on AI visual recognition according to claim 1, wherein: The specific steps of determining whether a target object exists in the captured image include: When the device is installed, the initial environment image is collected as the background image, the update cycle is set, the background update process is triggered, and the background image is updated; After grayscale conversion and filtering for noise reduction, the collected image is differentiated from the background image pixel by pixel, and the absolute difference is calculated. The dynamic foreground area is selected by setting a dynamic threshold. Adopting adaptive threshold segmentation algorithm to calculate target threshold of dynamic foreground area and segment the foreground target area; The brightness of the foreground target area is adjusted, the size of the foreground target area is normalized, and then the similarity with the target database is calculated; the similarity between the foreground target area and the target database is composed of image similarity and feature similarity.

5. The method for monitoring a target-targeted agent based on AI visual recognition according to claim 4, wherein: The specific step of determining whether the target object exists in the captured image also includes: The image similarity is determined by calculating the pixel-level similarity between the foreground target area and the image in the target database at multiple scales; The feature similarity is calculated as follows: The shape and texture features of the foreground target area are extracted using a feature extraction algorithm; the shape feature similarity is obtained by calculating the Hu moment invariant Euclidean distance between the foreground target area and the target image in the target database; The texture histogram is extracted using the local binary pattern, and the texture similarity between the foreground target area and the target image in the target database is calculated using the Bhattacharyya distance. Finally, the feature similarity is obtained through weighted summation. Configure the image similarity threshold and feature similarity threshold. If the image similarity between the foreground target area and the image in the target database is greater than the image similarity threshold, or the feature similarity is greater than the feature similarity threshold, it is determined that there is a target in the dynamic foreground area of ​​the captured image.

6. The method for monitoring a target-targeting agent based on AI visual recognition according to claim 1, wherein: The specific steps of automatically adjusting the fill light brightness according to the ambient light include: During the non-collection period, the ambient light intensity is measured, a light threshold is configured, and the ambient light intensity is compared with the light threshold. If the ambient light intensity is greater than the light threshold, it is determined that there is sufficient natural light; otherwise, it is determined that there is insufficient natural light; For triggering the camera to capture images or fixed images, if it is determined that there is sufficient natural light, the image will be captured directly using ambient light. If it is determined that there is insufficient natural light, the fill light adjustment process will be entered: Setting a basic fill light brightness and a fill light duration interval, performing fill light at the basic fill light brightness and capturing an image within the fill light duration interval, and obtaining a grayscale mean and contrast of the captured image; Configure the grayscale threshold and contrast threshold. If the grayscale mean of the captured image is greater than the grayscale threshold and the contrast is greater than the contrast threshold, it is judged as a valid image; otherwise, it is judged as an invalid image. If it is an invalid image, set the fill light brightness step and gradually increase the fill light brightness step based on the basic fill light brightness. Re-capture the image within the fill light interval, extract the dynamic area coordinates of the valid image and the invalid image through background difference, and replace the dynamic area pixel values ​​with the interpolation results of the two pixel values ​​to merge and generate a valid image.

7. The method for monitoring a target-targeted agent based on AI visual recognition according to claim 6, wherein: The specific step of automatically adjusting the fill light brightness according to the ambient light also includes: For video capture, if it is determined that there is sufficient natural light, the video will be captured directly using ambient light. If it is determined that there is insufficient natural light, the dynamic fill light adjustment process will be entered: Starting from the ambient light intensity, the fill light brightness is gradually and smoothly increased. The grayscale mean and contrast of the video frame are analyzed in real time. When the grayscale mean of the video frame is greater than the grayscale threshold and the contrast is greater than the contrast threshold, the dynamic fill light adjustment is stopped and the current fill light brightness is maintained to complete the video acquisition.

8. The method for monitoring a target-targeted agent based on AI visual recognition according to claim 1, wherein: The specific steps of obtaining the target object activity intensity and performing transmission optimization include: Enable the adaptive contrast enhancement algorithm for the video frames received by the monitoring platform, configure the detection confidence threshold, identify the individual targets in the video frames and mark their bounding boxes, and output the target coordinates and timestamp information; Configure the trajectory matching threshold, bind the target IDs in consecutive video frames, and generate a trajectory sequence. Configure the time window and calculate the average moving speed of the target within the time window length based on the trajectory sequence. Divide the video frame into grids, count the number of targets in each grid, and map the difference in number between grids into a clustering index; Configure the weight ratio of individual number, movement speed, and aggregation, calculate the activity intensity index by weight, and smooth the activity intensity index within the configured smoothing window; Configure low and high activity intensity thresholds. When the smoothed activity intensity index is less than the low activity intensity threshold, the video capture pause instruction is triggered. Configure a single data quota. When the number of received video frames exceeds the single data quota and the activity intensity index exceeds the high activity intensity threshold, data compression mode is triggered: Calculate the rate of change between video frames through pixel differences or feature vector differences, mark key frames and remove redundant frames, and upload key frames to the monitoring platform.

9. The method for monitoring a target-targeted agent based on AI visual recognition according to claim 3, wherein: The specific steps of performing feedback adjustment on the determination threshold include: When the trigger judgment threshold is adjusted, the confidence level of the detected vibration signal and infrared signal is recorded, and the false trigger count is increased; Configure the false touch threshold. If the false trigger count is greater than the false touch threshold, it is determined to be a high-frequency false trigger and the feedback adjustment is frozen. Configure the adjustment step, including vibration adjustment step and infrared adjustment step. Add the corresponding adjustment step based on the current vibration threshold and infrared threshold to increase the judgment threshold. When the feedback adjustment is triggered under the judgment threshold, the image for which the feedback adjustment is triggered under the judgment threshold is marked as a missed detection sample, and the missed detection count is increased; Configure the missed detection threshold. When the missed detection count is greater than the preset missed detection threshold, the detection sensitivity is determined to be insufficient. Based on the current vibration threshold and infrared threshold, the corresponding adjustment step is reduced to lower the judgment threshold.

10. A device for monitoring a target-targeting agent based on AI visual recognition, which is implemented based on a method for monitoring a target-targeting agent based on AI visual recognition according to any one of claims 1 to 9, characterized in that: It consists of a trap module, an intelligent power supply module, a visual recognition module, and an IoT transmission module; The trapping module includes a bionic nest structure shell and a targeting agent, which is used to attract the target object to enter and achieve group killing through the targeting agent; The intelligent power supply module is used to provide power support through hybrid power supply and intelligent energy consumption regulation; The visual recognition module is used to identify the target object and dynamically monitor its activity status; The Internet of Things transmission module is used for remote monitoring and information feedback of the target object.

Images