Method and system for intelligent shooting control and identification of forest and grass disease and pest trapping

Abstract

The present application provides a kind of forest grass disease and insect pest trapping synergistic intelligent shooting control identification method and system, it is related to image recognition technical field, including obtaining trapping data and constructing activity index curve, calculating activity intensity ratio, segmented mapping conversion is generated nonlinear compensation correction value in combination with environmental parameters, when correction value exceeds threshold value, trigger shooting, extract target motion trajectory and carry out compensation, obtain actual disease and insect pest behavior characteristics, identify type and statistics quantity.The present application can accurately identify disease and insect pest, improve shooting efficiency, realize precision control, reduce drug dosage and control cost.

Description

TECHNICAL FIELD

[0001] The present application relates to image recognition technology, and in particular to a forest pest trapping and intelligent shooting control recognition method and system. BACKGROUND

[0002] Forest pest control is an important part of ensuring the health of the ecological system and the sustainable development of forest resources. In forestry production, early monitoring and accurate identification of pests play a key role in prevention and control. Currently, forest pest monitoring is mainly carried out through the use of traps to capture pests, manual observation and recording, and image acquisition and analysis. As a commonly used monitoring tool, traps can attract specific insects and capture and record them, providing basic data for pest trend analysis. With the development of information technology, intelligent image recognition systems have been gradually applied to forest pest monitoring, enabling real-time monitoring and rapid identification of pests through automatic shooting and image processing technology.

[0003] Existing monitoring technologies lack in-depth analysis and prediction capabilities for insect activity patterns, and mostly use fixed time interval shooting methods, which cannot dynamically adjust the monitoring strategy according to pest activity intensity, resulting in insufficient monitoring data during the active period of pests or a large amount of redundant data during the inactive period, affecting monitoring efficiency and resource utilization. Traditional monitoring methods do not fully consider the complex influence of environmental factors on pest behavior, especially the nonlinear influence of temperature, humidity and other environmental parameters on pest activity, making it difficult to accurately reflect the actual pest occurrence situation with the collected data, reducing the accuracy and scientificity of the monitoring. Moreover, existing image recognition systems have low accuracy in processing dynamic targets, especially in complex backgrounds or lighting conditions, making it difficult to effectively capture and analyze the behavior characteristics and motion trajectories of insects, and unable to accurately distinguish different pest species with similar morphologies, limiting the effectiveness and effectiveness of prevention and control decisions. SUMMARY

[0004] The present application provides a forest pest trapping and intelligent shooting control recognition method and system, which can solve the problems in the prior art.

[0005] In a first aspect of the present application, a forest pest trapping and intelligent shooting control recognition method is provided, comprising:

[0006] Obtaining target insect information and trapping location information;

[0007] Constructing an activity index curve based on the target insect information, and calculating the activity intensity ratio of adjacent time periods in the activity index curve;

[0008] The environment parameters corresponding to the trapping position information are acquired, the activity intensity ratio is segmented and mapped and converted according to the environment parameters, a non-linear compensation correction value is generated, when the non-linear compensation correction value exceeds a preset correction threshold, the shooting device is controlled to perform a shooting operation, and a pest image is acquired;

[0009] A target motion trajectory is extracted from the pest image, the target motion trajectory is segmented and compensated according to the non-linear compensation correction value, and an actual pest behavior feature is obtained;

[0010] The pest type is identified and the pest quantity is counted according to the actual pest behavior feature, and an intelligent control strategy is generated.

[0011] The activity index curve of the target insect is constructed based on the target insect information, which includes:

[0012] The collision signal of the target insect with the wall of the trap and the residence duration information of the target insect are acquired according to the target insect information, the collision signal is filtered by using a multi-time scale adaptive threshold value to remove environmental interference signals;

[0013] The activity amount of the target insect per unit time is calculated according to the amplitude characteristics of the collision signal and the residence duration information;

[0014] The activity amount of the target insect per unit time is compared with the activity amount in a historical time window, and an activity continuity coefficient is generated;

[0015] The activity index of the target insect is calculated based on the activity amount per unit time and the activity continuity coefficient, and the activity index is recorded in time sequence to construct the activity index curve of the target insect.

[0016] The environment parameters corresponding to the trapping position information are acquired, the activity intensity ratio is segmented and mapped and converted according to the environment parameters, a non-linear compensation correction value is generated, when the non-linear compensation correction value exceeds a preset correction threshold, the shooting device is controlled to perform a shooting operation, and a pest image is acquired, which includes:

[0017] The environment parameters corresponding to the trapping position information are acquired, the activity intensity ratio is segmented and mapped and converted according to the environment parameters, a non-linear compensation correction value is generated, when the non-linear compensation correction value exceeds a preset correction threshold, the shooting device is controlled to perform a shooting operation, and a pest image is acquired, which includes:

[0018] The activity intensity ratio is segmented and weighted according to the boundary characteristics in the insect aggregation area information, and a regional activity characteristic value is generated;

[0019] The change direction of the temperature and humidity gradient field and the light intensity field is predicted to predict the insect migration path, the regional activity characteristic value is spatially mapped and converted based on the insect migration path, and a non-linear compensation correction value is generated;

[0020] Collect the real-time change direction of the temperature and humidity gradient field and the light intensity field, and calculate the deviation degree of the real-time change direction from the insect migration path;

[0021] When the nonlinear compensation correction value exceeds the preset correction threshold and the deviation degree is greater than zero, correct the insect migration path to obtain a corrected migration path;

[0022] According to the corrected migration path and the nonlinear compensation correction value, predict the insect group behavior characteristics and determine the optimal shooting opportunity, control the shooting device to perform the shooting operation, and obtain the pest image.

[0023] According to the corrected migration path and the nonlinear compensation correction value, predict the insect group behavior characteristics and determine the optimal shooting opportunity, control the shooting device to perform the shooting operation, and obtain the pest image.

[0024] According to the displacement change of the corrected migration path, establish an insect group activity intensity curve, take the peak point and the valley point of the activity intensity curve as the insect group behavior characteristic points, and extract the time interval between the behavior characteristic points to determine the insect group behavior period;

[0025] Based on the behavior period, divide the nonlinear compensation correction value into a plurality of compensation intervals, calculate the change gradient of the nonlinear compensation correction value in each compensation interval, and determine the behavior characteristics of the insect group according to the change gradient;

[0026] Establish a gradient tracking window in the compensation interval, and when the amplitude of the change gradient exceeds a preset gradient threshold, adaptively expand the range of the gradient tracking window to the adjacent compensation interval to form a behavior prediction interval;

[0027] Segment fitting is performed on the activity intensity curve in the behavior prediction interval, the conversion time of the insect group behavior is predicted according to the slope change of the fitting curve, and the conversion time is determined as the optimal shooting time.

[0028] From the pest image, extract the target motion trajectory, and segmentally compensate the target motion trajectory according to the nonlinear compensation correction value to obtain the actual pest behavior characteristics, including:

[0029] From the pest image sequence, extract the target position coordinates to construct the target motion trajectory, calculate the transition probability of the target state at adjacent time points, predict the expected position of the target according to the transition probability, associate the deviation between the expected position and the actual position with the nonlinear compensation correction value, and generate a state deviation sequence;

[0030] Based on the state deviation sequence, determine the compensation trigger time, calculate the motion trend score of the target at each compensation trigger time, segment the target motion trajectory according to the motion trend score, and obtain the to-be-compensated trajectory segment;

[0031] The state deviation in the to-be-compensated trajectory segment is accumulated and counted, and the accumulated counting result is used as a reward function, and the optimal compensation parameter is determined through iterative optimization, and the to-be-compensated trajectory segment is compensated according to the optimal compensation parameter.

[0032] The compensated trajectory segment is adaptively recombined and boundary smoothed according to the gradient change of the optimal compensation parameter, and the motion feature of the smoothed trajectory is extracted to obtain the actual pest behavior feature.

[0033] The compensation trigger time is determined based on the state deviation sequence, the motion trend score of the target is calculated at each compensation trigger time, the target motion trajectory is segmented according to the motion trend score, and the to-be-compensated trajectory segment is obtained.

[0034] The state deviation sequence is multi-scale decomposed to obtain deviation components of different frequency bands, the probability distribution entropy values of the deviation components of each frequency band are calculated, the wave trough position is identified based on the time sequence change of the probability distribution entropy values, and the initial compensation trigger time is determined.

[0035] At the initial compensation trigger time, a target function including a state deviation minimum term, a motion constraint term and an environmental disturbance compensation term is constructed, and the gradient projection method is used to iteratively optimize the target function to obtain the compensation trigger time.

[0036] At the compensation trigger time, the amplitude spectrum and the phase spectrum of the state deviation sequence are extracted to construct a feature matrix, the feature matrix is decomposed to obtain a feature vector, and the motion trend score is calculated based on the energy distribution of the feature vector.

[0037] The motion mode mutation point is identified according to the motion trend score, the effective segmentation point is screened in combination with the density of the compensation trigger time, and the to-be-compensated trajectory segment is obtained.

[0038] The pest type is identified and the pest quantity is counted according to the actual pest behavior feature, and a control scheme is generated.

[0039] Based on the actual pest behavior feature, a pest identification threshold is generated.

[0040] The behavior parameters of the insect body target are collected and extracted.

[0041] The behavior parameters of the insect body target are matched with the actual pest behavior feature, and the pest type is identified according to the pest identification threshold.

[0042] Based on the actual pest behavior feature, a target tracking area is determined, the insect body target is tracked and counted in the target tracking area, the pest quantity is obtained, and the distribution density of the pest quantity in the target tracking area is calculated.

[0043] According to the type of the disease and insect pest, a corresponding damage degree weight is determined, the distribution density is weighted with the damage degree weight to obtain a control index, and an intelligent control strategy is generated from a control measure library based on the control index.

[0044] In a second aspect of the embodiment of the present application, a forest and grass disease and insect pest trapping and cooperative intelligent shooting control and identification system is provided, comprising:

[0045] A first unit is configured to acquire target insect information and trapping location information.

[0046] A second unit is configured to construct an activity index curve based on the target insect information, and calculate an activity intensity ratio of adjacent time periods in the activity index curve.

[0047] A third unit is configured to acquire environmental parameters corresponding to the trapping location information, perform segmented mapping conversion on the activity intensity ratio according to the environmental parameters, and generate a nonlinear compensation correction value.

[0048] A fourth unit is configured to extract a target motion trajectory from the disease and insect pest image, perform segmented compensation on the target motion trajectory according to the nonlinear compensation correction value, and obtain an actual disease and insect pest behavior feature.

[0049] A fifth unit is configured to identify a disease and insect pest type and count a disease and insect pest quantity according to the actual disease and insect pest behavior feature, and generate an intelligent control strategy.

[0050] In a third aspect of the embodiment of the present application, an electronic device is provided, comprising:

[0051] A processor;

[0052] A memory for storing processor-executable instructions;

[0053] The processor is configured to invoke the instructions stored in the memory to execute the method described above.

[0054] In a fourth aspect of the embodiment of the present application, a computer-readable storage medium is provided, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the method described above.

[0055] In the embodiment, by acquiring the trapping data of the forest pest trap, constructing the activity index curve and correcting the environmental parameters, the activity law of the pest is accurately grasped, the shooting efficiency and recognition accuracy are improved, and the device resource consumption is reduced. Through the control of the nonlinear compensation correction value, the method can intelligently judge the appropriate shooting opportunity, avoid invalid shooting, and accurately extract the behavior characteristics of the pest, so that the pest identification is more comprehensive and reliable. The invention combines the trapping data with the environmental parameters to construct a complete forest pest monitoring and control system, which not only improves the accuracy and efficiency of pest identification, but also automatically generates a control scheme according to the identification results, providing technical support for scientific prevention and control of forest pests, and having significant ecological and economic benefits. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A flowchart of the forest pest trapping and cooperative intelligent shooting control and recognition method of the embodiment of the present application is shown in

[0057] Figure 2 A flowchart of the pest behavior characteristic extraction method of the embodiment of the present application is shown in DETAILED DESCRIPTION

[0058] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme of the embodiment of the present application will be described clearly and completely below in combination with the drawings of the embodiment of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0059] The technical scheme of the present application will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in some embodiments.

[0060] Figure 1 A flowchart of the forest pest trapping and cooperative intelligent shooting control and recognition method of the embodiment of the present application is shown in Figure 1 The method comprises:

[0061] acquiring target insect information and trapping location information;

[0062] constructing an activity index curve based on the target insect information, and calculating the activity intensity ratio of adjacent time periods in the activity index curve;

[0063] Obtaining the environment parameters corresponding to the trapping position information, segmenting and mapping the activity intensity ratio according to the environment parameters to generate a non-linear compensation correction value, when the non-linear compensation correction value exceeds a preset correction threshold, controlling the shooting device to perform a shooting operation to obtain a pest image;

[0064] Extracting a target motion trajectory from the pest image, segmenting and compensating the target motion trajectory according to the non-linear compensation correction value to obtain an actual pest behavior feature;

[0065] Identifying the pest type and counting the number of pests according to the actual pest behavior feature to generate an intelligent control strategy.

[0066] In an optional embodiment, constructing the activity index curve based on the target insect information includes:

[0067] Obtaining a collision signal and a residence duration information of the target insect and the trap wall surface according to the target insect information, the collision signal uses a multi-time scale adaptive threshold filtering method to remove environmental interference signals;

[0068] According to the amplitude characteristics of the collision signal and the residence duration information, the activity amount of the target insect per unit time is calculated;

[0069] Comparing the activity amount of the target insect per unit time with the activity amount in the historical time window to generate an activity continuity coefficient;

[0070] Based on the activity amount per unit time and the activity continuity coefficient, the activity index of the target insect is calculated, and the activity index is recorded in time sequence to construct the activity index curve of the target insect.

[0071] In the specific implementation process, first, data collection is needed from the collision signal and the residence duration information of the target insect and the trap wall surface. A high-sensitivity piezoelectric sensor is installed on the inner wall of the trap, which will generate corresponding electrical signals when the insect collides with the wall. At the same time, the position change of the insect in the trap is monitored by an infrared sensor array, and the residence duration is recorded. In practical applications, taking the rice planthopper as an example, when it moves in the trap, it usually generates a collision signal of 5-15 mV, while the environmental interference signal is generally between 0.5-3 mV.

[0072] To remove environmental interference signals, the method uses a multi-time scale adaptive threshold filtering technique. In specific implementation, first set three time windows: short window (5 seconds), medium window (30 seconds) and long window (5 minutes). Statistical analysis is performed on the signal in each time window, and the mean and standard deviation of the signal strength are calculated. The short window is used to capture the sudden interference, the medium window is suitable for periodic interference, and the long window reflects the baseline change of the environment. According to the signal characteristics under different time scales, the corresponding threshold is calculated, specifically: the mean value of the window signal plus 1.5 to 3 times (adjust according to the target insect species) of the standard deviation. Taking rice planthopper as an example, the measured data show that the short window threshold is set to the mean value plus 2.5 times the standard deviation, which is the best effect at this time, about 95% of the environmental noise can be filtered out while retaining 98% of the effective collision signal. After adaptive threshold filtering of the three time windows, the filtering results are fused to obtain the final collision signal sequence.

[0073] After obtaining the effective collision signal, the signal amplitude feature and the residence time information are further extracted. For signal amplitude, the peak size, duration and energy density of each collision are recorded. For example, a healthy adult rice planthopper usually produces a collision signal of 8-12 mV, with a duration of about 15-25 milliseconds; while an old or weak individual produces a signal of 5-8 mV, with a duration of about 10-15 milliseconds. In terms of residence time, the infrared sensor array is used to track the residence time of insects in different areas of the trap. Healthy individuals have an average residence time of about 25-40 seconds in the food area, and about 5-15 seconds in the non-food area.

[0074] When calculating the activity amount of the target insect per unit time, the collision signal characteristics and residence time information are combined. The activity amount calculation considers the following factors: collision frequency, average amplitude of collision signal, total energy of collision signal, and residence time ratio. For each minute period, record the collision frequency N, calculate the average amplitude A, get the total energy E, and calculate the time ratio R of the active area and the static area. The activity amount index per unit time can be obtained by integrating these parameters. For example, in practical application, when the rice planthopper has a collision frequency of 8-12 times per minute, an average amplitude of 9 mV, and an activity-static time ratio of about 3:1, its activity amount index per unit time reaches a high level, indicating that the insect is in an active state.

[0075] To evaluate the continuity of activity, the method compares the amount of activity in the current time unit (e.g., 1 minute) with the amount of activity in the historical time window (e.g., the past 10 minutes). By comparing the trend of the current activity amount with the historical activity amount, the activity continuity coefficient is calculated. For example, if the current period activity amount is more than 1.2 times the average historical activity amount, and the upward trend is maintained for 3 consecutive periods, the activity continuity coefficient is higher (e.g., 0.85-0.95); if the activity amount fluctuates greatly or shows a downward trend, the continuity coefficient is lower (e.g., 0.3-0.5).

[0076] Based on the activity amount per unit time and the activity continuity coefficient, the activity index of the target insect is calculated. The activity index is obtained by combining the current activity amount and the activity continuity coefficient, and is normalized to make the final result fall between 0-100. For example, when the activity amount index is high (e.g., 12 collisions per minute) and the activity continuity coefficient is also high (e.g., 0.9), the calculated activity index may be between 85-95; while when the activity amount is low (e.g., 3 collisions per minute) and the continuity is poor (e.g., 0.4), the activity index may be only between 20-30.

[0077] The calculated activity index is recorded in time sequence, i.e., the activity index curve of the target insect is constructed. The curve reflects the change of the activity state of the target insect at different time points. By analyzing the activity curve, the activity pattern of the insect, the diurnal cycle change and the response characteristics to environmental factors can be identified. Such data provides precise time window selection basis for agricultural pest control.

[0078] In an optional implementation, the environment parameters corresponding to the trapping position information are obtained, the activity intensity ratio is segmented and mapped and converted according to the environment parameters, and a non-linear compensation correction value is generated. When the non-linear compensation correction value exceeds a preset correction threshold, the shooting device is controlled to perform a shooting operation to obtain the pest image, which includes:

[0079] The environment parameters corresponding to the trapping position information are obtained, and the environment parameters are constructed into a temperature and humidity gradient field and an illumination intensity field. The insect aggregation area is identified according to the temperature and humidity gradient field and the illumination intensity field, and the insect aggregation area information is generated.

[0080] The activity intensity ratio is segmented and weighted according to the boundary features in the insect aggregation area information, and a regional activity feature value is generated.

[0081] The insect migration path is predicted according to the change direction of the temperature and humidity gradient field and the illumination intensity field, the regional activity feature value is spatially mapped and converted based on the insect migration path, and a non-linear compensation correction value is generated.

[0082] Collect the real-time change direction of the temperature and humidity gradient field and the light intensity field, and calculate the deviation degree of the real-time change direction from the insect migration path;

[0083] When the nonlinear compensation correction value exceeds the preset correction threshold and the deviation degree is greater than zero, the insect migration path is corrected to obtain a corrected migration path;

[0084] According to the corrected migration path and the nonlinear compensation correction value, the insect group behavior characteristics are predicted and the optimal shooting opportunity is determined, and the shooting device is controlled to perform a shooting operation to obtain a pest image.

[0085] Specifically, first, the sensor network arranged in the farmland area collects environmental parameters of the trapping position, including temperature, humidity, light intensity and other data. Five-point sampling method is used to collect environmental data in the center and within a 5-meter range around each monitoring point to form a data matrix in the micro area. The temperature and humidity data are constructed into a temperature and humidity gradient field according to a grid density of 0.5 meters x 0.5 meters, and a light intensity field is constructed with a sampling density of 1 meter x 1 meter. Data can be collected every 15 minutes to ensure the real-time and accuracy of the gradient field.

[0086] The constructed temperature and humidity gradient field and light intensity field are visualized by heat maps, wherein the temperature field range is usually 15-35℃, the humidity field range is 40%-90%, and the light intensity field range is 0-100000 lux. The area with temperature of 25-30℃, humidity of 65%-75% and light intensity of 20000-40000 lux is identified as a potential insect aggregation area. Combined with historical data analysis, the system further screens out the area with a temperature and humidity change gradient less than 0.2℃ / meter and 2% / meter and a gentle light intensity change as an insect high-probability aggregation area, thereby generating insect aggregation area information.

[0087] After obtaining the insect aggregation area information, the activity intensity ratio is segmented and weighted according to the boundary characteristics of the area. Specifically, the aggregation area is divided into a core area, a transition area and an edge area, which are respectively given weight coefficients of 1.5, 1.2 and 0.8. The insect activity trajectory data identified by the image processing algorithm is compared with the historical benchmark activity data to calculate the activity intensity ratio. For example, when the number of insect activity trajectories in the core area is 1.8 times the benchmark value, the activity intensity ratio is 2.7 after applying a weight of 1.5; when the number of insect activity trajectories in the transition area is 1.5 times the benchmark value, the activity intensity ratio is 1.8 after applying a weight of 1.2; and when the number of insect activity trajectories in the edge area is 0.9 times the benchmark value, the activity intensity ratio is 0.72 after applying a weight of 0.8. Through this weighting method, more accurate regional activity characteristic values are generated.

[0088] According to the change direction of temperature and humidity gradient field and light intensity field, the migration path of insects is predicted. By analyzing the gradient direction of environmental parameters, the preferred migration direction of insects is identified, such as migrating to areas with suitable temperature, moderate humidity and suitable light. A grid method is used to divide the monitoring area into 10 cm x 10 cm grid cells, and each grid cell contains an environmental fitness score. The prediction of insect migration path is achieved by finding the sequence of grid cells with high to low environmental fitness score, forming the predicted migration route map.

[0089] The regional activity feature value is spatially mapped and converted based on the predicted insect migration path to generate a non-linear compensation correction value. The spatial mapping conversion uses a distance decreasing algorithm, that is, as the distance from the insect aggregation center point increases, the activity feature value decreases in a non-linear manner. For example, when the core area activity feature value is 2.7, the compensation correction value on the predicted migration path 1 meter away from the core area may be 2.3, 3 meters away 1.8, and 5 meters away 1.2. This non-linear mapping ensures that the system can accurately capture the spatial distribution characteristics of insect activity.

[0090] Real-time change data of temperature and humidity gradient field and light intensity field is continuously collected, and environmental parameters are updated every 5 minutes. By calculating the included angle between the real-time environmental change direction and the predicted insect migration path, the system determines the degree of deviation. For example, when the included angle is greater than 30 degrees, it indicates that there is a significant deviation between the environmental change direction and the predicted migration path; when the included angle is less than 15 degrees, it is considered that the environmental change direction basically conforms to the predicted path.

[0091] When the calculated non-linear compensation correction value exceeds the preset correction threshold of 1.5 and the deviation degree is greater than zero (i.e. there is a deviation), the migration path correction mechanism is started. The correction process recalculates the environmental fitness score based on real-time environmental change data and updates the migration path prediction. The correction algorithm will retain the stable environmental section of the original path, while adjusting the predicted direction of the obvious change section, and finally generate the corrected migration path.

[0092] Based on the corrected migration path and non-linear compensation correction value, the behavior characteristics of insect groups are predicted and the optimal shooting opportunity is determined. When it is predicted that the insect group activity will peak at a specific location (the compensation correction value exceeds 2.0), the shooting signal is triggered. The shooting device adjusts the focal length, viewing angle and exposure parameters according to the predicted location information to ensure that the pest image is captured at the best time. A 3-second continuous shooting mode is adopted, capturing 5 high-definition images per second to obtain a complete record of pest activity status.

[0093] Through the implementation of the above technical means, the insect activity law can be accurately grasped, high-quality pest images can be obtained at the optimal time, and the accuracy and efficiency of intelligent monitoring of agricultural pests can be effectively improved, providing reliable data support for subsequent pest control.

[0094] In an alternative embodiment, predicting the insect swarm behavior characteristics and determining the optimal shooting opportunity according to the corrected migration path and the nonlinear compensation correction value comprises:

[0095] According to the displacement change of the corrected migration path, an insect swarm activity intensity curve is established, the peak points and the trough points of the activity intensity curve are taken as insect swarm behavior characteristic points, and the time interval between the behavior characteristic points is extracted to determine the insect swarm behavior period;

[0096] Based on the behavior period, the nonlinear compensation correction value is divided into multiple compensation intervals, the change gradient of the nonlinear compensation correction value in each compensation interval is calculated, and the behavior characteristics of the insect swarm are determined according to the change gradient;

[0097] A gradient tracking window is established in the compensation interval, when the amplitude of the change gradient exceeds a preset gradient threshold, the range of the gradient tracking window is adaptively expanded to the adjacent compensation interval to form a behavior prediction interval;

[0098] The activity intensity curve in the behavior prediction interval is segmented and fitted, the conversion time of the insect swarm behavior is predicted according to the slope change of the fitted curve, and the conversion time is determined as the optimal shooting time.

[0099] In this embodiment, first, the continuous monitoring data of the insect swarm in a specific area is obtained, including position coordinates, moving speed, direction change and other information. After preprocessing the original data recorded by the high-precision image sensor, the corrected migration path and the corresponding nonlinear compensation correction value are generated.

[0100] For example, in a 10m x 10m observation area, a high-speed camera device with 60 frames per second is used to continuously collect activity data for 4 hours. After image recognition and coordinate conversion, the corrected migration path data set is obtained, containing position information of about 864,000 time points. Based on this migration path, the displacement change between adjacent time points is calculated, and an activity intensity curve is constructed. Specifically, the average displacement change is calculated every 30 seconds to generate an activity intensity curve with 480 data points. In this curve, multiple peaks and troughs can be clearly identified, such as a peak value of 18.3 cm / s at time point 85 (42.5 minutes) and a trough value of 4.2 cm / s at time point 123 (61.5 minutes). By analyzing all the peak and trough points, the average behavior period of the bee swarm is determined to be 38 minutes.

[0101] According to the determined behavior cycle of 38 minutes, the non-linear compensation correction value is divided into multiple compensation intervals, each interval has a length of 38 minutes. Among them, the 4-hour observation period is divided into about 6.3 complete compensation intervals. For each compensation interval, the change gradient of the non-linear compensation correction value inside it is calculated. The change gradient is obtained by dividing the compensation correction value difference of adjacent time points by the time interval. In the third compensation interval (76-114 minutes), the non-linear compensation correction value rises from 0.82 to 1.43, and the average change gradient is calculated to be 0.016 / minute. By comparing the gradient characteristics in different compensation intervals, the behavior characteristics of the insect population can be determined. For example, when the change gradient is positive and greater than 0.01 / minute, it indicates that the foraging activity is enhanced; when the change gradient is negative and less than -0.01 / minute, it indicates that the nest returning activity is intensified.

[0102] In the compensation interval where the high change gradient is identified, a gradient tracking window is established for fine observation. The initial window width is set to 10 minutes, and when the absolute value of the change gradient exceeds the preset threshold of 0.02 / minute, the window is automatically expanded by 5 minutes to the adjacent area. For example, in the 130-140 minute segment of the fourth compensation interval (114-152 minutes), the change gradient is detected to be 0.025 / minute, which exceeds the threshold, and the tracking window is expanded to 125-145 minutes, forming a new behavior prediction interval.

[0103] The activity intensity curve in the determined behavior prediction interval is processed by piecewise fitting. The data of 125-145 minutes is divided into three segments: 125-132 minutes, 132-138 minutes and 138-145 minutes, and a linear function is used for fitting respectively. In the first segment, the slope is 0.31 cm / s / minute, in the second segment, the slope is 1.27 cm / s / minute, and in the third segment, the slope is -0.42 cm / s / minute. By analyzing the change of the slope of adjacent segments, the time points of 132 minutes where the slope changes sharply from low to high and the time points of 138 minutes where the slope changes sharply from high to low are determined as the behavior transition times. These transition times indicate major changes in the behavior pattern of the insect population, such as from cruising flight to foraging in clusters, or from dispersed activity to coordinated movement.

[0104] After determining the behavior transition time, the optimal shooting time is selected according to the transition characteristics. For a bee population, when a rapid transition of activity intensity from low to high (positive slope sharply increasing) is detected, the time window of 2 minutes before and 3 minutes after this time is marked as the key shooting interval. For example, at 132 minutes, such a transition is detected, and 130-135 minutes is automatically set as the optimal shooting time window, and a high-frequency capture instruction is triggered on the shooting device to record the population activity at a rate of 2 frames / second during this period.

[0105] In the embodiment, the activity intensity curve of the insect population is established, the behavior characteristic points are extracted to determine the behavior cycle of the population, and the activity law of the insect population is effectively described. Based on the behavior cycle, the compensation interval is divided and the change gradient is calculated, which can accurately reflect the dynamic change characteristics of the insect population behavior. The adaptive extended gradient tracking window is adopted, which can dynamically capture the continuous change process of the insect population behavior. By segmenting and fitting the activity intensity curve in the behavior prediction interval, the conversion time of the insect population behavior can be predicted in time, so as to grasp the best shooting opportunity. The overall technical scheme realizes the accurate prediction of the insect population behavior and the optimization determination of the shooting opportunity, improves the accuracy and timeliness of the insect population behavior monitoring. At the same time, the scheme has strong adaptability and can adapt to the behavior characteristics of different insect populations.

[0106] As shown in Figure 2 , the behavior characteristic extraction method of the pest is shown.

[0107] In an optional implementation, the target motion trajectory is extracted from the pest image, the target motion trajectory is segmented and compensated according to the nonlinear compensation correction value, and the actual pest behavior characteristic includes:

[0108] The target position coordinates are extracted from the pest image sequence to construct the target motion trajectory, the transition probability of the target state at adjacent time points is calculated, the expected position of the target is predicted according to the transition probability, the deviation between the expected position and the actual position is associated with the nonlinear compensation correction value, and a state deviation sequence is generated;

[0109] The compensation trigger time is determined based on the state deviation sequence, the motion trend score of the target is calculated at each compensation trigger time, the target motion trajectory is segmented according to the motion trend score, and a to-be-compensated trajectory segment is obtained;

[0110] The state deviation in the to-be-compensated trajectory segment is accumulated and counted, the accumulated counting result is taken as a reward function, the optimal compensation parameter is determined by an iterative optimization method, and the to-be-compensated trajectory segment is compensated according to the optimal compensation parameter;

[0111] The compensated trajectory segment is adaptively recombined and boundary smoothed according to the gradient change of the optimal compensation parameter, the motion characteristics of the smoothed trajectory are extracted, and the actual pest behavior characteristic is obtained.

[0112] Exemplarily, firstly, the collected forest and grass disease and pest image sequence is processed to extract the target position coordinates to construct the motion trajectory. In the processing, the background difference and morphological filtering are combined to extract the centroid position coordinates of the disease and pest target in each frame of image. The coordinate points in the continuous multiple frames of image are connected in time sequence to form the initial motion trajectory. For the target state of adjacent time, the transition probability is calculated, which is based on the historical motion characteristics of the target and contains the speed and acceleration information. The transition probability calculation uses the conditional probability density function, which is obtained by statistical analysis of the historical data. Based on the calculated transition probability, the expected position of the target at the next time is predicted, and the deviation vector of the expected position from the actual observation position is associated with the nonlinear compensation correction value. The nonlinear compensation correction value is calculated by a mapping function, which considers the influence of the deviation size and direction. The deviation vector sequence in the continuous time window is generated to form the state deviation sequence.

[0113] The method for determining the compensation trigger time based on the state deviation sequence is to set a deviation threshold value, and when the modulus values of the continuous several deviations in the state deviation sequence all exceed the threshold value, the compensation trigger time is determined. The motion trend score of the target is calculated at each compensation trigger time, which reflects the consistency and predictability of the target motion. The motion trend score is obtained by comprehensive evaluation of the motion direction change rate and speed stability of the target in a certain time window. The included angle of the adjacent displacement vectors in the calculation time window is calculated, and if the included angle is small, it means that the motion direction is stable, and the score increases. The speed change rate is calculated, and if the change rate is small, it means that the speed is stable, and the score increases. According to the motion trend score, the target motion trajectory is divided into multiple trajectory segments using the piecewise threshold method, and the trajectory segment with a score lower than the threshold value is marked as a trajectory segment to be compensated.

[0114] For the trajectory segment to be compensated, the state deviations inside it are accumulated and counted, and the counting method includes calculating the modulus value of the deviation vector, the direction consistency index, and the time correlation of the deviation. These statistical results are combined to form a reward function, which evaluates the pros and cons of the compensation effect. The optimal compensation parameters are determined by iterative optimization, specifically using a policy search algorithm. The algorithm starts from an initial compensation parameter set, and through multiple iterations, the parameter value is adjusted according to the feedback of the reward function, and finally converges to the optimal parameter set. The compensation parameters include the position correction factor, the speed adjustment factor, and the acceleration suppression factor. According to the optimal compensation parameters, each point coordinate in the trajectory segment to be compensated is transformed, and the transformation formula includes three parts of position offset, speed adjustment, and acceleration constraint. The transformed trajectory segment is obtained.

[0115] After the trajectory segment compensation is completed, the compensated trajectory segment is adaptively reorganized according to the gradient change of the optimal compensation parameter. The gradient change reflects the change trend of the compensation parameter in the trajectory segment, and the region with large gradient change needs to be further subdivided, and the region with small gradient change can be combined. The reorganization process is realized by a dynamic programming algorithm, which finds the optimal segmentation point according to the set cost function. Boundary smoothing is performed at the segmentation point, and a cubic spline interpolation method is used to ensure the continuity and smoothness between adjacent trajectory segments. The complete trajectory after boundary smoothing extracts motion features, including average speed, speed change rate, motion direction change frequency, and residence time distribution, which together constitute the actual pest behavior feature set.

[0116] The forest pest trapping and cooperative intelligent shooting control recognition method provided by the embodiment effectively overcomes the influence of the instability of image acquisition in the complex outdoor environment on the target trajectory extraction, and improves the accuracy of pest behavior feature recognition. The method can adapt to different light conditions and background changes, intelligently identifies and compensates for abnormal points in the trajectory, so that the extracted behavior features are more consistent with the actual activity rules of pests. At the same time, through trajectory segmentation processing and boundary smoothing technology, the continuity and consistency of the behavior features are ensured, which provides a reliable basis for subsequent pest species identification, hazard level assessment and prevention measures.

[0117] In an optional embodiment, a compensation trigger time is determined based on a state deviation sequence, a motion trend score of a target is calculated at each compensation trigger time, and the target motion trajectory is segmented based on the motion trend score to obtain a trajectory segment to be compensated, including:

[0118] The state deviation sequence is subjected to multi-scale decomposition to obtain deviation components of different frequency bands, the probability distribution entropy values of the deviation components of each frequency band are calculated, the wave trough positions are identified based on the time sequence change of the probability distribution entropy values, and the initial compensation trigger time is determined;

[0119] At the initial compensation trigger time, a target function including a state deviation minimum term, a motion constraint term and an environmental interference compensation term is constructed, the gradient projection method is used to iteratively optimize the target function, and the compensation trigger time is obtained;

[0120] At the compensation trigger time, the amplitude spectrum and the phase spectrum of the state deviation sequence are extracted to construct a feature matrix, the feature matrix is decomposed to obtain a feature vector, and a motion trend score is calculated based on the energy distribution of the feature vector;

[0121] The motion mode mutation points are identified according to the motion trend score, the effective segmentation points are screened in combination with the density of the compensation trigger time, and the trajectory segment to be compensated is obtained.

[0122] Specifically, the state deviation sequence is first decomposed into deviation components of different frequency bands. Specifically, the state deviation sequence is decomposed into high-frequency, medium-frequency and low-frequency deviation components by wavelet transform. For example, for a state deviation sequence with a sampling frequency of 200 Hz, it can be decomposed into a low-frequency component of 0-5 Hz, a medium-frequency component of 5-50 Hz and a high-frequency component of 50-100 Hz.

[0123] For each frequency band deviation component, its probability distribution entropy value is calculated. Specifically, a probability histogram is constructed for each frequency band deviation component, the deviation value range is divided into several equal intervals, the number of samples in each interval is counted and normalized to a probability value. Taking the low-frequency component as an example, the value range of [-10, 10] can be divided into 100 intervals, the sample proportion in each interval is counted to form a probability distribution. Then the entropy value of the probability distribution is calculated to represent the uncertainty of the deviation distribution. The entropy value is analyzed in time series to identify the valley position of the entropy value curve. The valley position corresponds to the time when the state deviation distribution is relatively concentrated and the certainty is high. These moments are suitable as initial compensation trigger moments. For example, when the entropy value is less than 0.3, it can be marked as a valley position as an initial compensation trigger moment.

[0124] At the initial compensation trigger moment, a target function containing three parts is constructed: a state deviation minimum term, a motion constraint term and an environmental disturbance compensation term. The state deviation minimum term is used to make the compensated state as close as possible to the desired state, the motion constraint term ensures that the compensation action meets the physical feasibility, and the environmental disturbance compensation term reduces the influence of external environmental factors. Gradient projection method is used to iteratively optimize the target function. In each iteration, the solution is updated along the negative gradient direction of the target function, and the updated solution is projected into the feasible region defined by the constraint conditions. The iteration process continues until the convergence condition is met, such as the difference between the solutions of adjacent two iterations is less than the preset threshold 0.001, or the maximum iteration number 200 is reached. The result obtained by optimization is the compensation trigger moment.

[0125] At the determined compensation trigger moment, the amplitude spectrum and the phase spectrum of the state deviation sequence are extracted to construct a feature matrix. Specifically, the state deviation sequence is subjected to Fourier transform to obtain a frequency domain representation, and the amplitude spectrum and the phase spectrum are extracted. Taking the amplitude spectrum as an example, the amplitudes of the top 10 frequency points with the highest power proportion are selected as features to form an amplitude feature vector. Similarly, the phase values of the corresponding frequency points are selected to form a phase feature vector. The amplitude feature vector and the phase feature vector are combined to form a feature matrix. The feature matrix is subjected to singular value decomposition to obtain a main feature vector. A motion trend score is calculated based on the energy distribution of the feature vector, which reflects the change trend of the target motion mode. Specifically, the projection energy of the feature vector under different motion modes is calculated, and the motion trend score between 0 and 1 is obtained after normalization. For example, a score close to 0 indicates that the target is in a stable state, and a score close to 1 indicates that the target is in a rapidly changing state.

[0126] A motion mode mutation point is identified according to the motion trend score. When the change rate of the motion trend score exceeds a preset threshold of 0.3, it is marked as a motion mode mutation point. Effective segmentation points are screened in combination with the density of the compensation trigger moments. Specifically, within the time window of the compensation trigger moment, if the number of compensation triggers between two mutation points exceeds 5 times, the two mutation points are retained as effective segmentation points. The target motion trajectory is segmented into multiple trajectory segments according to the effective segmentation points, and the trajectory segments with dense compensation trigger moments are identified as trajectory segments to be compensated.

[0127] In this embodiment, the key moments of system state change can be effectively identified through multi-scale decomposition and entropy analysis of the state deviation sequence. The accurate determination of the compensation trigger moment is achieved by using a target function containing multiple constraints. The motion trend score can be accurately calculated through feature matrix decomposition and energy distribution analysis, so as to identify the mutation characteristics of the target motion mode. The effective segmentation points are screened in combination with the density of the compensation trigger moments, which avoids redundant segmentation and improves the accuracy and rationality of trajectory segmentation. The overall scheme realizes adaptive segmentation of the target motion trajectory and intelligent determination of the compensation opportunity, improves the accuracy and real-time performance of motion trajectory compensation, and has strong anti-interference ability and environmental adaptability.

[0128] In an alternative embodiment, the pest type is identified according to the actual pest behavior characteristics, and the number of pests is counted to generate a control scheme, including:

[0129] Based on the actual pest behavior characteristics, a pest identification threshold is generated;

[0130] Collecting target information of the insect body, extracting behavior parameters of the insect body target;

[0131] Matching the behavior parameters of the insect body target with the actual pest behavior characteristics, identifying the pest type according to the pest identification threshold;

[0132] Determine the target tracking area based on the actual pest behavior characteristics, track and count the pest targets in the target tracking area to obtain the pest quantity, and calculate the distribution density of the pest quantity in the target tracking area;

[0133] Determine the corresponding damage degree weight according to the pest type, weight the distribution density and the damage degree weight to obtain the control index, and generate an intelligent control strategy from the control measure library based on the control index.

[0134] Among them, the actual pest behavior characteristics usually include multi-dimensional data such as movement speed, movement trajectory, activity frequency, and feeding characteristics. These characteristic data are processed, a clustering analysis method is used to establish a characteristic space, and the behavior characteristics of pests of the same type are gathered to form a characteristic cluster. The boundary distance of the characteristic cluster is used as the initial recognition threshold, and different threshold values are set for different types of pests. For example, for pine caterpillars, the recognition threshold of their typical behavior characteristics includes a crawling speed threshold of 6-10 cm / min and a direction change frequency threshold of 1-3 times per minute. For locusts, set a jump height threshold of 10-15 cm and a travel speed threshold of 20-30 cm / min. These thresholds are obtained through a large number of sample data training and are dynamically adjusted according to environmental conditions to ensure recognition accuracy.

[0135] The collection of pest target information is completed by high-definition camera equipment. The collected images are preprocessed and then enter the feature extraction link. The behavior parameters of the pest targets include several key steps: image segmentation, target tracking, and feature calculation. Image segmentation uses background difference method combined with morphological processing to accurately separate the pest targets. Target tracking uses a tracking algorithm based on correlation filtering to establish the correspondence of targets in consecutive image sequences. Calculate the behavior parameters of the tracked targets, including real-time speed, acceleration, movement direction, and residence time. The calculation of behavior parameters uses a sliding window method to reduce the influence of environmental noise. The calculated parameters form a behavior feature vector, which contains a complete behavior description of the pest.

[0136] The matching process of the behavior parameters of the insect target with the actual pest behavior characteristics can adopt a pattern matching algorithm. The target behavior parameter vector is compared with the preset behavior characteristic templates of various pests in terms of similarity, and the similarity calculation method adopts a weighted Euclidean distance. For key dimensions in the behavior characteristics, such as specific movement patterns, a higher weight is given. For example, for the monochamus alternatus in coniferous forest areas, the weight of its unique zigzag crawling pattern is set to 2.0, and the weight of the general speed parameter is set to 0.8. The calculated similarity is compared with a pest identification threshold, and when the similarity exceeds the threshold, it is determined to be the corresponding type of pest. For the case where the similarity is close but does not exceed the threshold, a secondary determination mechanism is introduced, and the morphological characteristics of the target are analyzed as auxiliary identification basis to improve the identification accuracy.

[0137] The determination of the target tracking area is based on the activity range information in the actual pest behavior characteristics. According to the activity habits of different pests, tracking areas of different shapes and sizes are set. For ground-activity pests such as mole crickets, rectangular tracking areas are set; for tree-climbing pests such as monochamus alternatus, strip-shaped tracking areas extending along the tree trunk are set. Virtual counting lines are set in the tracking area, and when the insect target passes through the counting line, a counting operation is triggered. To avoid repeated counting, each target is given a unique identification code, and whether it is a new target is determined by matching the identification code. In the tracking process, a prediction-update mechanism is used to handle the short-time occlusion problem of the target. When the target temporarily disappears, its possible position is predicted, and the matching target is searched in the predicted area to realize continuous tracking. After the pest quantity statistics are completed, the target tracking area is divided into grid cells, the number of pests in each cell is calculated, and a distribution density heat map is generated to visually display the pest distribution.

[0138] The corresponding relationship between the pest type and the damage degree weight is stored in the preset weight database. The weight value is comprehensively evaluated according to factors such as the damage mode, damage degree, and reproduction speed of the pest to the plant. For example, the monochamus alternatus directly damages the tree trunk, resulting in tree death, so the weight value is set to 0.9; while general leaf-eating pests such as the pine caterpillar have relatively light damage, so the weight value is set to 0.7. The pest distribution density and the damage degree weight are weighted and calculated, and the calculation method is the product of the distribution density and the corresponding weight, to obtain a control index. The control index is divided into different levels, such as low risk, medium risk, high risk, and emergency, and each level corresponds to a different control strategy. The control measure library stores control schemes for different pests and different damage degrees, including physical control, biological control, chemical control, and other methods. The most suitable control scheme is selected from the control measure library according to the control index, and adjusted according to environmental factors, such as preferentially selecting a biological control scheme in an ecologically sensitive area, and using a comprehensive control scheme in a general area.

[0139] For example, when the distribution density of pine caterpillar is detected as 3.5 per square meter, and the damage degree weight is 0.7, the calculated control index is 2.45, which belongs to the medium risk level. From the control measure library, the biological control scheme is selected, and the natural enemy trichogramma is recommended to release to control the number of pine caterpillars, and the sex lure trap is set to assist in monitoring. The control scheme includes specific operation guidance and time arrangement to ensure the control effect.

[0140] The embodiment realizes intelligent monitoring and control of forest and grass diseases and pests by accurate identification and quantity statistics of behavior characteristics of the diseases and pests. The behavior characteristic identification technology breaks through the limitations of traditional morphological identification, and can accurately distinguish different disease and pest species with similar appearances in a complex background. The target tracking and counting technology solves the problems of repeated counting and missing counting in traditional manual investigation, and greatly improves the accuracy and efficiency of quantity statistics. The introduction of the control index makes the control measures more accurate, avoiding resource waste and environmental pollution caused by excessive control. The overall technical scheme realizes the intelligentization of the whole process from monitoring to control, not only reduces the labor cost and improves the work efficiency, but more importantly, improves the scientificity and ecological friendliness of the forest and grass disease and pest control, and provides strong technical support for the sustainable management of forest and grass resources.

[0141] In a second aspect, the embodiment of the present application provides a forest and grass disease and pest trapping and cooperative intelligent shooting control identification system, which comprises:

[0142] A first unit is configured to acquire target insect information and trapping location information;

[0143] A second unit is configured to construct an activity index curve based on the target insect information, and calculate an activity intensity ratio of adjacent time periods in the activity index curve;

[0144] A third unit is configured to acquire environmental parameters corresponding to the trapping location information, perform segmented mapping conversion on the activity intensity ratio according to the environmental parameters, generate a nonlinear compensation correction value, and control a shooting device to perform a shooting operation to acquire a disease and pest image when the nonlinear compensation correction value exceeds a preset correction threshold;

[0145] A fourth unit is configured to extract a target motion trajectory from the disease and pest image, and perform segmented compensation on the target motion trajectory according to the nonlinear compensation correction value to obtain actual disease and pest behavior characteristics;

[0146] A fifth unit is configured to identify a disease and pest type and count a disease and pest quantity according to the actual disease and pest behavior characteristics, and generate an intelligent control strategy.

[0147] In a third aspect, the embodiment of the present application provides an electronic device, which comprises:

[0148] A processor;

[0149] a memory for storing processor-executable instructions;

[0150] The processor is configured to invoke the instructions stored in the memory to perform the method described above.

[0151] In a fourth aspect, the present application provides a computer readable storage medium having stored thereon computer program instructions, which when executed by a processor implement the method described above.

[0152] The present application can be a method, an apparatus, a system, and / or a computer program product. The computer program product can include a computer readable storage medium having computer readable program instructions stored therein, which when executed by a processor, perform various aspects of the present application.

[0153] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for coordinated intelligent shooting, control, and identification of forest and grassland pests and diseases, characterized in that, include: Obtain information about the target insect and its location; An activity index curve is constructed based on the target insect information, and the ratio of activity intensity between adjacent time periods in the activity index curve is calculated. The environmental parameters corresponding to the trapping location information are obtained. The activity intensity ratio is segmented and mapped according to the environmental parameters to generate a nonlinear compensation correction value. When the nonlinear compensation correction value exceeds the preset correction threshold, the shooting device is controlled to perform a shooting operation to obtain images of pests and diseases. Extract the target motion trajectory from the disease and pest images, and compensate the target motion trajectory in segments according to the nonlinear compensation correction value to obtain the actual disease and pest behavior characteristics; Identify pest and disease types and count pest and disease numbers based on actual pest and disease behavior characteristics, and generate intelligent prevention and control strategies. The activity index curve constructed based on target insect information includes: Based on the target insect information, the collision signal and dwell time information between the target insect and the trap wall are obtained. The collision signal is filtered by a multi-time-scale adaptive threshold filtering method to remove environmental interference signals. Based on the amplitude characteristics and dwell time information of the collision signal, the activity level of the target insect per unit time is calculated; The activity level of the target insect within a unit of time is compared with the activity level within a historical time window to generate an activity continuity coefficient. Based on the activity level and activity continuity coefficient per unit time, the activity index of the target insect is calculated, and the activity index is recorded according to the time series to construct the activity index curve of the target insect.

2. The method according to claim 1, characterized in that, The system acquires environmental parameters corresponding to the trapping location information, performs a segmented mapping transformation on the activity intensity ratio based on the environmental parameters, generates a nonlinear compensation correction value, and controls the imaging device to perform an imaging operation when the nonlinear compensation correction value exceeds a preset correction threshold to acquire images of pests and diseases, including: Obtain environmental parameters corresponding to the trapping location information, construct the environmental parameters into a temperature and humidity gradient field and a light intensity field, identify insect gathering areas based on the temperature and humidity gradient field and light intensity field, and generate insect gathering area information; Based on the boundary features in the insect aggregation area information, the activity intensity ratio is divided into segments and weighted to generate regional activity feature values; Insect migration paths are predicted based on the changing directions of temperature and humidity gradient fields and light intensity fields. Regional activity characteristic values ​​are spatially mapped and transformed based on insect migration paths to generate nonlinear compensation correction values. The real-time change directions of the temperature and humidity gradient field and the light intensity field are collected, and the degree of deviation between the real-time change direction and the insect migration path is calculated. When the nonlinear compensation correction value exceeds the preset correction threshold and the deviation is greater than zero, the insect migration path is corrected to obtain the corrected migration path. Based on the corrected migration path and nonlinear compensation correction value, the behavior characteristics of insect groups are predicted and the optimal shooting time is determined. The shooting device is then controlled to perform the shooting operation to acquire images of pests and diseases.

3. The method according to claim 2, characterized in that, Based on the corrected migration path and nonlinear compensation correction value, predicting insect group behavior characteristics and determining the optimal shooting time includes: Based on the displacement changes of the corrected migration path, an insect group activity intensity curve is established. The peaks and troughs of the activity intensity curve are used as insect group behavior feature points. The time interval between the behavior feature points is extracted and determined as the insect group behavior cycle. Based on the behavioral cycle, the nonlinear compensation correction value is divided into multiple compensation intervals. The change gradient of the nonlinear compensation correction value in each compensation interval is calculated, and the behavioral characteristics of the insect group are determined based on the change gradient. A gradient tracking window is established within the compensation interval. When the magnitude of the changing gradient exceeds a preset gradient threshold, the range of the gradient tracking window is adaptively expanded to the adjacent compensation interval to form a behavior prediction interval. The activity intensity curve within the behavior prediction interval is piecewise fitted, and the transition time of insect group behavior is predicted based on the slope change of the fitted curve. The transition time is then determined as the optimal shooting time.

4. The method according to claim 1, characterized in that, The target motion trajectory is extracted from the pest and disease image, and the target motion trajectory is segmented and compensated according to the nonlinear compensation correction value to obtain the actual pest and disease behavior characteristics, including: The target location coordinates are extracted from the image sequence of pests and diseases to construct the target movement trajectory. The transition probability of the target state at adjacent time moments is calculated. The expected position of the target is predicted based on the transition probability. The deviation between the expected position and the actual position is associated with the nonlinear compensation correction value to generate a state deviation sequence. The compensation trigger time is determined based on the state deviation sequence. At each compensation trigger time, the target's motion trend score is calculated. The target's motion trajectory is segmented according to the motion trend score to obtain the trajectory segment to be compensated. The state deviation within the trajectory segment to be compensated is accumulated and statistically analyzed. The accumulated statistical results are used as the reward function. The optimal compensation parameters are determined through iterative optimization. The trajectory segment to be compensated is then compensated based on the optimal compensation parameters. Based on the gradient change of the optimal compensation parameters, the compensated trajectory segments are adaptively recombined and the boundaries are smoothed. The motion features of the smoothed trajectory are then extracted to obtain the actual pest and disease behavior features.

5. The method according to claim 4, characterized in that, The compensation trigger time is determined based on the state deviation sequence. At each compensation trigger time, the target's motion trend score is calculated. The target's motion trajectory is segmented according to the motion trend score to obtain the trajectory segment to be compensated, including: Multi-scale decomposition of the state deviation sequence is performed to obtain deviation components in different frequency bands. The probability distribution entropy value of each frequency band deviation component is calculated. Based on the temporal change of the probability distribution entropy value, the trough position is identified, and the initial compensation trigger time is determined. At the initial compensation trigger time, an objective function is constructed that includes a minimum state deviation term, a motion constraint term, and an environmental disturbance compensation term. The objective function is iteratively optimized using the gradient projection method to obtain the compensation trigger time. At the compensation trigger moment, the amplitude spectrum and phase spectrum of the state deviation sequence are extracted to construct a feature matrix. The feature matrix is ​​decomposed to obtain feature vectors. The motion trend score is calculated based on the energy distribution of the feature vectors. Based on the motion trend score, identify the abrupt change points in the motion pattern, and combine the density of compensation trigger times to filter effective segment points, thus obtaining the trajectory segment to be compensated.

6. The method according to claim 1, characterized in that, Based on the actual behavioral characteristics of pests and diseases, identify the types of pests and diseases and count their numbers, then generate control plans including: Based on the actual behavioral characteristics of pests and diseases, generate pest and disease identification thresholds; Collect insect target information and extract insect target behavioral parameters; The behavioral parameters of the insect target are matched with the actual pest behavior characteristics, and the pest type is identified according to the pest identification threshold. The target tracking area is determined based on the actual behavior characteristics of pests and diseases. The insect targets are tracked and counted within the target tracking area to obtain the number of pests and diseases. The distribution density of the number of pests and diseases within the target tracking area is then calculated. The corresponding hazard severity weight is determined according to the type of pest or disease. The control index is obtained by weighting the distribution density and the hazard severity weight. Intelligent control strategies are generated from the control measures library based on the control index.

7. A forestry and grassland pest and disease trapping and coordinated intelligent shooting control and recognition system, used to implement the method of any one of claims 1-6, characterized in that, include: The first unit is used to obtain information about the target insects and the trapping location; The second unit is used to construct an activity index curve based on the target insect information and to calculate the ratio of activity intensity between adjacent time periods in the activity index curve. The third unit is used to obtain environmental parameters corresponding to the trapping location information, perform segmented mapping transformation on the activity intensity ratio according to the environmental parameters, generate nonlinear compensation correction value, and control the shooting device to perform shooting operation to obtain images of pests and diseases when the nonlinear compensation correction value exceeds the preset correction threshold. The fourth unit is used to extract the target motion trajectory from the pest and disease image, and to perform segmented compensation on the target motion trajectory according to the nonlinear compensation correction value to obtain the actual pest and disease behavior characteristics. The fifth unit is used to identify pest types and count pest numbers based on actual pest behavior characteristics, and to generate intelligent prevention and control strategies.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images