A pest intelligent monitoring optimal interval period discrimination method

Abstract

The present application relates to the technical field of intelligent pest monitoring, and more specifically, to a method for determining an optimal interval for intelligent pest monitoring, which obtains continuous trapping count data of target pests and constructs an original monitoring sequence, forms a control response event section set containing a trigger start time, a section peak time and a section release time according to a control threshold, generates a candidate monitoring interval set based on a preset minimum sampling interval, an allowed maximum monitoring interval and a control response allowed time limit, and performs event protection type resampling on each candidate interval, calculates information loss degree, trend fidelity, resource consumption degree, control event coverage, control response delay penalty degree and phase transition retention degree, and filters and determines an optimal monitoring interval under the constraint of edge resource budget. The method can reduce device energy consumption, communication and storage resource consumption while ensuring control event coverage and timely response.

Description

Technical Field

[0001] This invention relates to the field of intelligent pest monitoring technology, and more specifically, to a method for determining the optimal interval period for intelligent pest monitoring. Background Technology

[0002] Pest population monitoring is fundamental to determining the control window and timing of pesticide application. Existing intelligent field pest monitoring equipment typically uses color attraction, sex pheromone trapping, or light attraction to acquire pest images or counting signals, and generates continuous trapping and counting data through cameras, image recognition modules, automatic counting devices, and communication modules.

[0003] Under long-term field operation conditions, intelligent pest monitoring equipment is limited by power supply capacity, communication status, storage capacity, and maintenance frequency. If short monitoring intervals are used for a long period, the number of times camera wake-up, image recognition, counting and summarizing, data uploading, and local storage are executed increases, resulting in greater consumption of equipment resources. During periods of low pest density or stable populations, this type of high-frequency data contributes little to the judgment of prevention and control.

[0004] If a long, fixed monitoring interval is used, it is easy to miss the start time when the pest population reaches the control threshold, the duration of the period above the threshold, the peak time of the period, or the time when the period ends, especially during the pest population's rise, peak, or resurgence phases. For pests such as the tea green leafhopper and the tea geometrid moth, control decisions depend not only on whether the pest population has reached the control threshold but also on the duration of the process after reaching the threshold. Excessively long monitoring intervals can lead to incomplete recording of control response event segments, affecting the determination of the control window.

[0005] Existing intelligent pest monitoring methods mostly rely on fixed cycles or manual experience to set monitoring intervals, making it difficult to simultaneously adapt to different pest trapping methods, population development stages, control thresholds, and equipment resource status. While some data processing solutions involve adjusting sampling frequency or data granularity, they primarily focus on data volume, overall trend similarity, or resource consumption, lacking constraints on the start time of control threshold triggering, peak time of a segment, segment release time, and control response delay.

[0006] Therefore, we propose a method for determining the optimal interval for intelligent pest monitoring to solve the above problems. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method for determining the optimal interval period for intelligent pest monitoring, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for determining the optimal interval period for intelligent pest monitoring, comprising the following steps: S1, Construction of a set of control response event segments: using an intelligent pest monitoring device installed in the target monitoring area, continuous trapping and counting data of target pests are collected according to a preset minimum sampling interval to form an original monitoring sequence. Original monitoring sequence This includes the number of pests trapped at each sampling time; and the number of pests trapped in the original monitoring sequence based on the control threshold and allowable time limit for the control response for the target pest. The process involves: S1) Constructing a set of original control response event segments, each segment including at least one original control response event segment, and each segment including trigger start time, segment peak time, segment release time, maximum number of traps in the segment, and segment event number; S2) Generating a set of candidate monitoring intervals: Based on the preset minimum sampling interval, the maximum allowable monitoring interval for the target pest, the allowable time limit for control response, and the set of original control response event segments, a set of candidate monitoring intervals T is generated; where each candidate monitoring interval t in the set of candidate monitoring intervals T is not less than the preset minimum sampling interval and not greater than the maximum allowable monitoring interval, and the candidate monitoring interval t satisfies the event coverage constraint for each original control response event segment in the set of original control response event segments; S3) Event-protected resampling: For each candidate monitoring interval t in the set of candidate monitoring intervals T, the original monitoring sequence is resampled... The candidate monitoring interval t is divided into multiple time-continuous resampling windows. For each resampling window, a window fluctuation discrimination vector is calculated, consisting of the window growth rate, window coefficient of variation, window peak-to-valley difference, and proximity of the control threshold. A fluctuation enhancement marker for the corresponding resampling window is generated based on this discriminant vector. When the corresponding resampling window overlaps with any original control response event segment in time, the trigger start time, segment peak time, or segment release time of that original control response event segment within the resampling window are preferentially retained, along with the number of pests trapped and the sampling time. When the corresponding resampling window does not overlap with any original control response event segment in time, and the fluctuation enhancement marker characterizes the... When the resampling window is in a state of increased volatility, the maximum number of pests trapped within the resampling window and its corresponding sampling time are retained. When the corresponding resampling window does not overlap with any original control response event segment in time, and the increased volatility marker indicates that the resampling window is in a stable state, the number of pests trapped within the resampling window is uniformly aggregated to generate the representative number of trapped pests and the representative sampling time for the resampling window. This yields the resampling monitoring sequence X(t) corresponding to each candidate monitoring interval t. S4, Time axis alignment and event protection quality index calculation: The resampling monitoring sequence X(t) corresponding to each candidate monitoring interval t is mapped back to the original monitoring sequence according to its representative sampling time. The timeline forms a sequence with the original monitoring sequence. Time-aligned interval restoration sequence Based on the original monitoring sequence Resampled monitoring sequence X(t), interval restored sequence And the original prevention and control response event segment set, respectively, calculate the information loss L(t), trend fidelity F(t), resource consumption R(t), prevention and control event coverage C(t), prevention and control response delay penalty P(t), and stage transition retention G(t) corresponding to the candidate monitoring interval t; where, the information loss L(t) is used to characterize the resampled monitoring sequence X(t) relative to the original monitoring sequence The degree of information entropy loss; trend fidelity F(t) is used to characterize the reconstructed sequence at intervals. For the original monitoring sequence The degree to which the population change trend, peak position, and fluctuation pattern of pests are maintained; resource consumption R(t) is used to characterize at least two of the following resource consumption parameters of intelligent pest monitoring equipment under the candidate monitoring interval t: number of data acquisitions, data upload volume, storage usage, and device wake-up energy consumption; control event coverage C(t) is used to characterize the proportion of the resampled monitoring sequence X(t) that successfully covers the original control response event segment; control response delay penalty P(t) is used to characterize the degree of delayed response of the resampled monitoring sequence X(t) relative to the original control response event segment; stage transition retention G(t) is used to characterize the interval-restored sequence The degree of retention of the transition points between the rising, peak, stable, trough, or secondary peak phases of the target pest population; S5, selection of qualified candidate intervals and optimization of resource budget constraints: obtain the edge resource budget status of the intelligent pest monitoring equipment, which includes at least two of the following: remaining power, available storage capacity, and communication status; determine the weight coefficients corresponding to information loss degree L(t), trend fidelity F(t), resource consumption R(t), control event coverage C(t), control response delay penalty P(t), and phase transition retention degree G(t) based on the target pest's trapping method, population stage, control threshold, monitoring application requirements, and edge resource budget status; first, based on the preset trend fidelity threshold, preset event coverage threshold, preset response delay allowable value, and preset... Setting a phase transition maintenance threshold, candidate monitoring intervals with trends fidelity F(t) lower than the preset trend fidelity threshold, prevention event coverage C(t) lower than the preset event coverage threshold, prevention response delay penalty P(t) higher than the preset allowable response delay value, or phase transition maintenance G(t) lower than the preset phase transition maintenance threshold are removed from the candidate monitoring interval set T, resulting in a qualified candidate monitoring interval set. If the qualified candidate monitoring interval set is not empty, then, under the condition of satisfying the resource constraints corresponding to the edge resource budget state, each candidate monitoring interval t in the qualified candidate monitoring interval set is comprehensively evaluated to obtain a comprehensive monitoring quality value Q(t), and the candidate monitoring interval with the largest comprehensive monitoring quality value Q(t) is selected from the qualified candidate monitoring interval set as the optimal monitoring interval. If the set of qualified candidate monitoring intervals is empty, the smaller of the preset minimum sampling interval or the previous round's effective optimal monitoring interval will be determined as the safe monitoring interval, and this safe monitoring interval will be used as the optimal monitoring interval for this round. Simultaneously, a rollback reason marker is generated; S6, device-side execution and confidence hysteresis dynamic update: the optimal monitoring interval period is... The output is sent to the intelligent insect pest monitoring device, enabling the device to monitor according to the optimal monitoring interval. Adjust at least two of the following for subsequent pest monitoring: camera wake-up cycle, image recognition module operation cycle, automatic counting cycle, data upload cycle, and local cache clearing cycle; when the newly added trap count data reaches the preset update data volume or preset update cycle, the newly added trap count data will be incorporated into the original monitoring sequence. Based on the updated original monitoring sequences The population growth rate, peak-valley variation characteristics, fluctuation intensity, and stage identification confidence level of the target pest are calculated to identify the current population development stage of the target pest. When the target pest is in the rising, peak, or secondary peak stage, and the stage identification confidence level reaches a preset confidence threshold, the upper limit of the candidate monitoring interval set T is reduced, and the weights corresponding to the control event coverage rate C(t), control response delay penalty P(t), and trend fidelity F(t) are increased. When the target pest is in a stable or trough stage, and the stage identification confidence level reaches a preset confidence threshold, the candidate monitoring interval set T is expanded. The upper limit of the monitoring interval is set, and the weight corresponding to the resource consumption R(t) is increased; when the confidence level of the stage identification is lower than the preset confidence threshold, the current execution monitoring interval is maintained or the current execution monitoring interval is rolled back to a shorter safe monitoring interval; then steps S2 to S5 are re-executed; after obtaining a new optimal monitoring interval, if the population occurrence stages identified in two consecutive update cycles are consistent, or the change of the new optimal monitoring interval relative to the current execution monitoring interval reaches the preset switching amplitude threshold, then the new optimal monitoring interval is updated to the device execution monitoring interval, otherwise the current execution monitoring interval is maintained.

[0009] In a preferred embodiment, the intelligent insect monitoring device includes a color-attracting intelligent monitoring device, a sex-attracting intelligent monitoring device, or a light-attracting intelligent monitoring device. The color-attracting intelligent monitoring device includes a sticky insect board, a high-definition camera, and an image recognition module. The sticky insect board is used to attract target pests exhibiting color attraction, the high-definition camera is used to capture images of the sticky insect board, and the image recognition module is used to identify the target pests in the sticky insect board images and output the number of pests captured at the corresponding sampling time. The sex-attracting intelligent monitoring device includes a sex pheromone lure, a trap, and an automatic counting device. The sex pheromone lure and trap are used to attract target pests exhibiting sex pheromone attraction, and the automatic counting device outputs the number of pests captured at the corresponding sampling time. The light-attracting intelligent monitoring device is used to attract target pests based on a light source and outputs the number of pests captured at the corresponding sampling time. In step S1, the initial monitoring sequence is formed. This also includes the original monitoring sequence. Preprocessing is performed, including: sorting the number of pests trapped in chronological order according to the sampling time; removing invalid sampling points formed during equipment offline, image recognition failure, automatic counting anomalies, or trap replacement; marking missing sampling points or performing interpolation based on adjacent valid sampling points; marking single-point outliers exceeding a preset anomaly multiple; and based on the preprocessed original monitoring sequence. Constructing the original prevention and control response event segment set. The construction of the original prevention and control response event segment set includes: when the original monitoring sequence... When the number of pests trapped changes from below the control threshold corresponding to the target pest to reaching or exceeding the control threshold, the sampling time is determined as the trigger start time; when the number of pests trapped changes from reaching or exceeding the control threshold to below the control threshold and remains below the control threshold for a preset release duration, the corresponding sampling time is determined as the segment release time; between the trigger start time and the segment release time, the sampling time with the largest number of pests trapped is determined as the segment peak time.

[0010] In a preferred embodiment, in step S3, the proximity of the control threshold is determined based on the ratio between the maximum number of pests trapped in the corresponding resampling window and the control threshold of the target pest. When the proximity of the control threshold reaches a preset threshold proximity coefficient, it is determined that there are sampling points close to the control threshold within the corresponding resampling window. When the corresponding resampling window overlaps with multiple key sampling times within the same original control response event segment, the corresponding number of pests trapped and the sampling time are retained according to the priority of trigger start time, segment peak time, and segment release time. In step S3, the window growth rate is determined based on the change between the number of pests trapped at the end and the number of pests trapped at the beginning within the corresponding resampling window. The window coefficient of variation is determined based on the ratio between the standard deviation and the mean of the number of pests trapped within the corresponding resampling window. The window peak-to-valley difference is determined based on the difference between the maximum number of pests trapped and the minimum number of pests trapped within the corresponding resampling window. When at least two of the window growth rate, window coefficient of variation, window peak-to-valley difference, and control threshold proximity satisfy the corresponding preset threshold, a fluctuation enhancement marker representing that the resampling window is in a fluctuation enhancement state is generated.

[0011] In a preferred embodiment, in step S4, the prevention and control event coverage rate C(t) is determined according to the ratio between the number of original prevention and control response event segments successfully covered by the resampled monitoring sequence X(t) and the total number of original prevention and control response event segments; the prevention and control response delay penalty P(t) is determined according to the delay time of the resampled response time matching the original prevention and control response event segment in the resampled monitoring sequence X(t) relative to the trigger start time of the original prevention and control response event segment, and a first penalty coefficient is used when the resampled response time is later than the trigger start time of the original prevention and control response event segment, and a second penalty coefficient is used when the resampled response time is earlier than or equal to the trigger start time of the original prevention and control response event segment, wherein the first penalty coefficient is greater than the second penalty coefficient. In step S4, the information loss degree L(t) is determined according to the original monitoring sequence The normalized difference between the information entropy of the original monitoring sequence X(t) and the information entropy of the resampled monitoring sequence is determined; the trend fidelity F(t) is determined according to the original monitoring sequence. Restored sequence with interval The weighted result of the Pearson correlation coefficient and the normalized dynamic time-normalized distance between the two was used to determine the stage transition retention G(t) according to the interval-based sequence reconstruction. Identified population stage transition points and original monitoring sequences The matching ratio between the identified population stage transition points is determined; the resource consumption R(t) is determined by weighting at least two of the following: sampling frequency ratio, data upload volume ratio, storage usage ratio, and device wake-up energy consumption ratio. In steps S5 and S6, after generating the rollback reason marker, the monitoring quality index corresponding to the rollback reason marker is checked first during the next dynamic update; when the monitoring quality index corresponding to the rollback reason marker continuously reaches the corresponding preset recovery threshold, the safe rollback state is exited, and the optimal monitoring interval selection based on the qualified candidate monitoring interval set is restored; when the monitoring quality index corresponding to the rollback reason marker does not reach the corresponding preset recovery threshold, the safe monitoring interval is maintained.

[0012] The technical effects and advantages of this invention are as follows: 1. This invention uses the prevention and control response event segments in the original monitoring sequence as the criteria for determining the interval. Each original prevention and control response event segment records the trigger start time, segment peak time, segment release time, maximum number of traps in the segment, and segment event number. This ensures that the generation, resampling, and quality evaluation of subsequent candidate monitoring intervals all revolve around the prevention and control response event segments. The set of candidate monitoring intervals is generated by constraints imposed by a preset minimum sampling interval, a maximum allowable monitoring interval, an allowable time limit for prevention and control response, and the set of original prevention and control response event segments. Each candidate monitoring interval must meet the event coverage requirement of the original prevention and control response event segments, avoiding the inclusion of long intervals that clearly cannot cover the prevention and control response events in the optimization scope.

[0013] 2. This invention employs event-protected resampling rules to process the resampling windows corresponding to each candidate monitoring interval. For resampling windows that overlap with the original control response event segment, the corresponding pest trapping quantity and sampling time are retained in the order of trigger start time, segment peak time, and segment release time. For resampling windows that do not overlap with the original control response event segment but are in a state of increased fluctuation, the maximum pest trapping quantity within the window and its corresponding sampling time are retained. For resampling windows in a stable state, stable aggregation is performed to generate representative trapping quantity and representative sampling time. Through the above rules, the resampling monitoring sequence retains the key sampling times of the control response event segment and the main fluctuation points of non-event segments, providing a data foundation for subsequent calculations of control event coverage, control response delay penalty, and trend fidelity. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the optimal interval period determination method for intelligent pest monitoring based on the protection of prevention and control response event segments provided in an embodiment of the present invention; Figure 2 This is a graph showing the changes in monitoring quality indicators of the tea green leafhopper under different candidate monitoring intervals in an embodiment of the present invention. Figure 3 This is a graph showing the change in the overall monitoring quality of the tea green leafhopper with the monitoring interval under different monitoring application requirements in an embodiment of the present invention. Figure 4 This is a graph showing the changes in monitoring quality indicators of the tea geometrid moth under different candidate monitoring intervals in an embodiment of the present invention. Figure 5 This is a graph showing the change in the overall monitoring quality of the gray tea geometrid moth under different monitoring application requirements as a function of the monitoring interval in an embodiment of the present invention. Figure 6 This is a comparison chart of the pest monitoring effects before and after applying this method in an embodiment of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Reference Figures 1-6 A method for determining the optimal interval for intelligent pest monitoring includes the following steps: S1. Construction of the prevention and control response event segment set, which includes the following steps:

[0017] Intelligent pest monitoring equipment is deployed within the target monitoring area, which is the field area where pest population dynamics monitoring and control response assessment are to be conducted. The intelligent pest monitoring equipment is selected based on the attraction characteristics of the target pests, and includes at least one of color-attracting intelligent monitoring equipment, sex-attracting intelligent monitoring equipment, or light-attracting intelligent monitoring equipment. When the target pest exhibits color attraction, color-attracting intelligent monitoring equipment is used. This equipment includes a sticky trap, a high-definition camera, and an image recognition module. The sticky trap is placed within the target monitoring area to attract color-attracting target pests. The high-definition camera is positioned facing the effective attraction surface of the sticky trap and acquires images of the trap at preset minimum sampling intervals. The image recognition module is connected to the high-definition camera to receive the sticky trap images, identify and count the target pests in the images, and output the number of pests captured at the corresponding sampling time. When the target pest exhibits sex pheromone attraction, sex-attracting intelligent monitoring equipment is used. The sex-attracting intelligent monitoring device includes a sex pheromone lure, a trap, and an automatic counting device. The sex pheromone lure is placed inside or in conjunction with the trap to attract target pests attracted by sex pheromones. The automatic counting device is configured corresponding to the pest entry channel, insect landing channel, or counting area of ​​the trap, generating a counting signal when a target pest enters the trap or passes through the counting area, and outputting the number of pests captured at the corresponding sampling time. When the target pest is suitable for light-based attraction, a light-attracting intelligent monitoring device is used. The light-attracting intelligent monitoring device attracts target pests based on an insect-attracting light source and outputs the number of pests captured at the corresponding sampling time through an image acquisition component, a sensor counting component, or an automatic counting component configured in conjunction with the insect-attracting light source.

[0018] After deployment, the intelligent pest monitoring equipment continuously collects the trapping count data of target pests at preset minimum sampling intervals. Each sampling session records at least the sampling time and the corresponding number of pests trapped at that time. When multiple intelligent pest monitoring devices are involved, each sampling session also records the corresponding device identifier to distinguish the sampling data based on the device's origin. The sampling times and corresponding pest trapping counts obtained through continuous collection form the original monitoring sequence. Original monitoring sequence It includes multiple sampling data points arranged in chronological order, and each sampling data point includes at least the sampling time and the number of pests trapped.

[0019] In forming the original monitoring sequence Subsequently, the original monitoring sequence Preprocessing is performed. Preprocessing is executed in the following order: "time series arrangement—invalid point removal—missing point handling—outlier marking." First, the original monitoring sequence is processed according to the sampling time. The sampling data points in the sequence are sorted in time sequence to form the original monitoring sequence. It can reflect the chronological relationship between the number of target pests trapped and the changes over time. In cases where multiple sampling records exist at the same sampling time, valid sampling records are retained according to device identification or preset data priority, or multiple valid sampling records from the same source at the same sampling time are merged into the pest trapping count for that corresponding sampling time. Then, the original monitoring sequence is analyzed. Invalid sampling points are removed. Invalid sampling points include those generated during equipment offline periods, those resulting from image recognition failures, those caused by automatic counting anomalies, and those generated during trap replacement. For color-attracting intelligent monitoring devices, when the sticky insect board image is missing, the image is blurry, the image recognition module does not output valid recognition results, or the sticky insect board replacement causes discontinuous trap count comparisons, the corresponding sampling point is marked as invalid. For sex-attracting intelligent monitoring devices, when the automatic counting device does not output a counting signal, the counting signal is continuously abnormal, or the counting aperture changes due to trap maintenance or sex pheromone lure replacement, the corresponding sampling point is marked as invalid. For light-attracting intelligent monitoring devices, when the light source is abnormal, the acquisition component does not output valid data, or the counting results are significantly missing, the corresponding sampling point is marked as invalid. Invalid sampling points are not included in the determination of the trigger start time, peak time, and release time of subsequent prevention and control response event segments. Next, missing sampling points are processed. When the time interval between two adjacent valid sampling points is greater than the preset minimum sampling interval, the sampling time in which no valid trap count was obtained is marked as a missing sampling point. For missing sampling points, only the missing marker can be retained, so that it is not included in the calculation of the effective trap quantity during subsequent event segment construction and resampling; alternatively, interpolation can be performed based on adjacent effective sampling points to obtain the interpolated trap quantity for the corresponding missing sampling point. The interpolated sampling points retain the interpolation marker to distinguish them from the actual effective sampling points obtained. Subsequently, the original monitoring sequence is processed. Single-point outliers are marked as anomalous. When the number of pests trapped at a sampling point changes more than a preset abnormal multiple relative to adjacent valid sampling points, and there are no records of equipment maintenance, trap replacement, control operations, or pest outbreaks to support this, the sampling point is marked as an anomalous sampling point. Data marked as anomalous sampling points is not directly deleted, but is judged in conjunction with adjacent valid sampling points when constructing control response event segments. When an anomalous sampling point is located at a suspected control threshold trigger position, the trend of changes in consecutive valid sampling points before and after it is prioritized to determine whether it constitutes a control response event segment.

[0020] After preprocessing, based on the preprocessed original monitoring sequences The control threshold corresponding to the target pest and the allowable time limit for control response are used to construct the original control response event segment set. The control threshold is used to determine whether the number of target pests trapped has reached the level that requires a control response. The allowable time limit for control response is used to limit the time boundary for subsequent candidate monitoring intervals and the determination of control response delay, and serves as the association constraint parameter after the original control response event segment is constructed.

[0021] When constructing the original prevention and control response event segment set, the preprocessed original monitoring sequences are read sequentially along the time order. The effective sampling points in the dataset were analyzed, and the relationship between the number of pests trapped at each sampling point and the control threshold corresponding to the target pest was compared. When the original monitoring sequence... If the number of pests trapped at a valid sampling point preceding a certain sampling point is lower than the control threshold for the target pest, and the number of pests trapped at that sampling point reaches or exceeds the control threshold, the sampling time corresponding to that sampling point is determined as the trigger start time of an original control response event segment. The trigger start time characterizes the starting point from which the target pest population transitions from a non-control response state to a control response state. After determining the trigger start time, subsequent valid sampling points are read sequentially from that trigger start time, and the time range during which the number of pests trapped continuously reaches or exceeds the control threshold is included in the same original control response event segment. When the number of pests trapped changes from reaching or exceeding the control threshold to falling below the control threshold, and the state of falling below the control threshold continues for a preset release duration, the corresponding sampling time that meets this condition is determined as the segment release time of that original control response event segment. The segment release time characterizes the time boundary of the target pest population exiting the control response state. Between the trigger start time and the segment release time, the number of pests trapped at each valid sampling point within this time range is compared. The sampling time corresponding to the sampling point with the largest number of pests trapped is determined as the segment peak time of the original control response event segment, and the maximum number of pests trapped is determined as the segment maximum trapping quantity. When multiple sampling points have the same maximum number of pests trapped, the sampling point that appears earliest in time is selected as the segment peak time to ensure that the peak points within the same original control response event segment have a unique correspondence.

[0022] A complete original control response event segment includes at least the trigger start time, segment peak time, segment release time, maximum number of pests trapped in the segment, and segment event number. The segment event number is generated sequentially according to the trigger start time of each original control response event segment and is used for matching and referencing different control response event segments in subsequent protective resampling, control event coverage calculation, control response delay penalty calculation, and stage transition retention calculation. When the target pest repeatedly reaches or exceeds the control threshold within a monitoring cycle, multiple original control response event segments are constructed sequentially according to the above rules, and these segments are combined into a set of original control response event segments based on their segment event numbers. Adjacent original control response event segments should be distinguished by the target pest trapping quantity being below the control threshold and continuously reaching the preset release duration. If the duration below the control threshold does not reach the preset release duration, consecutive or nearly consecutive threshold exceedances are considered as the same original control response event segment to avoid unreasonable segmentation of control response event segments due to short-term fluctuations. Through the above processing, S1 outputs the preprocessed original monitoring sequence. And the original set of prevention and control response event segments. Preprocessed original monitoring sequences. Used for subsequent candidate monitoring interval set constraint generation, event-protected resampling, and time axis alignment; the original control response event segment set is used to determine whether the candidate monitoring interval meets the event coverage constraint, and is used to preferentially retain the number of pests trapped and the sampling time corresponding to the trigger start time, segment peak time, and segment release time in event-protected resampling. S2. Generation of candidate monitoring interval set constraints, which includes the following steps: The raw monitoring sequence after completing S1 and obtaining preprocessing. Following the original prevention and control response event segment set, the preprocessed original monitoring sequences were used. The set of candidate monitoring intervals T is generated by taking the preset minimum sampling interval, the maximum allowable monitoring interval for the target pest, the allowable time limit for the control response, and the original set of control response event segments as inputs. When generating the candidate monitoring interval set T, the basic value boundaries of the candidate monitoring intervals are first determined. The preset minimum sampling interval is used as the lower limit of the candidate monitoring intervals, and the maximum allowable monitoring interval for the target pest is used as the upper limit of the candidate monitoring intervals. Simultaneously, the allowable time limit for the control response is used as a response constraint for the candidate monitoring intervals, ensuring that the candidate monitoring intervals do not exceed the time boundary that satisfies the timeliness requirement of the control response. Therefore, the candidate monitoring interval t should satisfy the following conditions: t is not less than the preset minimum sampling interval, t is not greater than the maximum allowable monitoring interval, and t does not exceed the upper limit of the monitoring interval corresponding to the allowable time limit for the control response. In one specific implementation, when the preset minimum sampling interval is 1 day and the maximum allowable monitoring interval is 30 days, an initial candidate monitoring interval set can be formed by using integer multiples of the preset minimum sampling interval as the basis for candidate values; alternatively, 1 day, 2 days, 3 days, 4 days, 5 days, 6 days, 7 days, 10 days, 14 days, 21 days, and 30 days can be selected as initial candidate monitoring intervals according to the needs of field monitoring operations. For other minimum sampling intervals, such as 1 hour, 6 hours, or 12 hours, integer multiples of the corresponding preset minimum sampling interval are generated as candidate values, and truncated according to the maximum allowable monitoring interval and the allowable time limit for prevention and control response to obtain the initial candidate monitoring interval set. After obtaining the initial candidate monitoring interval set, for each candidate monitoring interval t, based on the preprocessed original monitoring sequence... The original monitoring sequence is divided according to the start and end sampling times and the candidate monitoring interval t. The system is divided into multiple time-continuous resampling windows. The duration of each resampling window matches the corresponding candidate monitoring interval t. Adjacent resampling windows are set consecutively in time, without any time gaps between them, to ensure that each resampling window covers the original monitoring sequence. The effective monitoring time range. For the original monitoring sequence. The remaining time period that is less than a complete candidate monitoring interval t can be used as the last resampling window. If the remaining time period is shorter than the preset minimum sampling interval or does not contain any valid sampling points, the remaining time period is merged with the previous resampling window. This ensures that each candidate monitoring interval t can form a sequence consistent with the original monitoring sequence. The corresponding resampling window sequence provides the window basis for subsequent event-protected resampling in S3.

[0023] After completing the resampling window division, each original prevention and control response event segment is time-correlated with the resampling window sequence under the corresponding candidate monitoring interval t. For any original prevention and control response event segment, the resampling window into which its trigger start time, segment peak time, and segment release time fall are determined, or the resampling window that overlaps with the original prevention and control response event segment in time is determined. If a resampling window overlaps with an original prevention and control response event segment in time, then the resampling window is used as the event association window for event-protected resampling in subsequent S3; if a resampling window does not overlap with any original prevention and control response event segment in time, then the resampling window enters fluctuation enhancement processing or stable aggregation processing in subsequent S3 according to the window fluctuation discrimination vector.

[0024] For each candidate monitoring interval t, it is further determined whether it meets the event coverage constraint for each original prevention and control response event segment in the original prevention and control response event segment set. For any candidate monitoring interval t, if at least one resampling window in its corresponding resampling window sequence overlaps with the occurrence time of a certain original prevention and control response event segment, and the resampling window can retain at least one key sampling moment among the trigger start time, segment peak time, or segment release time of the original prevention and control response event segment in S3, then the candidate monitoring interval t is determined to meet the event coverage requirement for the original prevention and control response event segment. If the trigger start time, segment peak time, and segment release time of a certain original prevention and control response event segment do not fall within the resampling window that can be retained for the candidate monitoring interval t, or if the delay of its earliest resampling key sampling moment relative to the trigger start time exceeds the allowable time limit for prevention and control response, then the candidate monitoring interval t is determined to not meet the event coverage requirement for the original prevention and control response event segment, and the candidate monitoring interval t is removed or marked as a low-priority candidate monitoring interval.

[0025] The event coverage constraints include: for each original prevention and control response event segment in the set of original prevention and control response event segments, there should be at least one resampling window in the resampling window sequence corresponding to the candidate monitoring interval t that overlaps with the occurrence time of the original prevention and control response event segment; and the resampling window should be able to retain the number of pests trapped and the sampling time corresponding to at least one of the key sampling times among the trigger start time, segment peak time or segment release time in subsequent event-protected resampling. When the trigger start time, peak time, and end time of the same original prevention and control response event segment fall into different resampling windows, the resampling window sequence corresponding to the candidate monitoring interval t can provide a retention position for each key sampling moment. When two or more key sampling moments of the same original prevention and control response event segment fall into the same resampling window, according to the retention priority of event-protected resampling in subsequent S3, the priority key sampling moment among the trigger start time, peak time, or end time is taken as the event retention object of the resampling window, and the subordinate relationship between other key sampling moments and the resampling window is recorded so that it can be called when calculating the prevention and control event coverage C(t), prevention and control response delay penalty P(t), and stage transition retention G(t) in subsequent calculations.

[0026] When the resampling window sequence corresponding to a candidate monitoring interval t cannot cover a certain original prevention and control response event segment, or cannot bring any of the key sampling moments of the trigger start time, segment peak time, or segment release time in the original prevention and control response event segment into the retainable window, the candidate monitoring interval t is removed from the initial candidate monitoring interval set. Candidate monitoring intervals t that still cover all original prevention and control response event segments and meet the allowable time limit requirements for prevention and control response are retained in the candidate monitoring interval set T. For each original prevention and control response event segment in the original prevention and control response event segment set, the time difference between the trigger start time of the event segment and the earliest retainable event sampling moment under the candidate monitoring interval t can be used to determine whether the candidate monitoring interval t meets the allowable time limit for prevention and control response. When the time difference is not greater than the allowable time limit for prevention and control response, the candidate monitoring interval t is determined to meet the response coverage requirements of the original prevention and control response event segment; when the time difference is greater than the allowable time limit for prevention and control response, the candidate monitoring interval t is determined to not meet the response coverage requirements of the original prevention and control response event segment, and the candidate monitoring interval t is removed or marked as a low-priority candidate monitoring interval.

[0027] During the generation of the candidate monitoring interval set T, a correspondence between the candidate monitoring interval t and the resampling window and the original prevention and control response event segment can be established simultaneously. This correspondence includes at least: the resampling window number corresponding to the candidate monitoring interval t, the start and end times of each resampling window, the event sequence number of the original prevention and control response event segment that overlaps with each resampling window, and the key sampling time that can be preferentially retained within the corresponding resampling window. This correspondence is used in S3 to determine the retention objects for event-protective resampling and in S4 to calculate the prevention and control event coverage rate C(t), the prevention and control response delay penalty P(t), and the stage transition retention rate G(t). If, after filtering by the above basic value boundary constraints, prevention and control response allowable time limit constraints, and event coverage constraints, the candidate monitoring interval set T is not empty, then the filtered candidate monitoring interval set T is output to S3 for subsequent event-protective resampling. If the candidate monitoring interval set T is empty after screening, the preset minimum sampling interval will be used as the minimum candidate monitoring interval to form the candidate monitoring interval set T. The reason for the empty candidate set will be recorded as the event coverage constraint not being met or the response time limit constraint not being met, so as to serve as the basis for determining the safety monitoring interval and generating the rollback reason mark in subsequent S5. S3, Event-protected resampling, specifically includes the following steps: After completing S2 and obtaining the candidate monitoring interval set T, the resampling window division results corresponding to each candidate monitoring interval t, and the correspondence between the resampling window and the original prevention and control response event segment, event-protected resampling processing is performed for each candidate monitoring interval t in the candidate monitoring interval set T.

[0028] For any candidate monitoring interval t, according to the resampling window division result determined in S2, the preprocessed original monitoring sequence is... The system is divided into multiple time-continuous resampling windows. Each resampling window has a corresponding start and end time, and these windows are arranged sequentially in chronological order. Each resampling window includes one or more sampling data points falling within its time range. Each sampling data point includes the sampling time and the number of pests trapped at that time. For each resampling window, the valid sampling data points within that window are first read and arranged from earliest to latest sampling time. If no valid sampling data points exist within a resampling window, it is marked as an empty window. Empty windows do not directly generate a representation of the trapped number; instead, based on the processing results of the adjacent valid resampling windows, they are either marked as missing or merged when forming the subsequent resampling monitoring sequence X(t). If valid sampling data points exist within a resampling window, the window fluctuation discrimination process is initiated. This process includes calculating the window growth rate, window coefficient of variation, window peak-to-valley difference, and the proximity of the control threshold. The window growth rate is determined based on the change between the number of pests trapped at the end of the resampling window and the number of pests trapped at the beginning. In one embodiment, the number of pests trapped at the earliest valid sampling point within the resampling window is taken as the initial number of pests trapped, and the number of pests trapped at the latest valid sampling point is taken as the final number of pests trapped. The window growth rate is then determined based on the difference or normalized difference between the final number of pests trapped and the initial number of pests trapped. When the initial number of pests trapped is 0, a preset minimum value or a difference method is used to determine the window growth rate to avoid the window growth rate being uncalculated due to a denominator of 0.

[0029] The window coefficient of variation is determined based on the ratio between the standard deviation and the mean of the number of pests trapped within the corresponding resampling window. Specifically, the mean and standard deviation of the number of pests trapped at each valid sampling point within the resampling window are calculated, and the window coefficient of variation is obtained based on the ratio between the standard deviation and the mean. When the mean is 0, the window coefficient of variation is set to 0 or determined according to the ratio after a preset minimum value correction. The window coefficient of variation is used in the subsequent generation of fluctuation enhancement markers. The window peak-to-valley difference is determined based on the difference between the maximum and minimum number of pests trapped within the corresponding resampling window. Specifically, the maximum and minimum number of pests trapped are found within the resampling window, and the difference between them is taken as the window peak-to-valley difference. When multiple maximum pests trapped exist within the same window, the sampling point with the earliest maximum value is taken as the sampling point corresponding to the maximum value of the window.

[0030] The proximity of the control threshold is determined based on the ratio between the maximum number of pests trapped within the corresponding resampling window and the control threshold for the target pest. Specifically, the maximum number of pests trapped within the resampling window is compared with the control threshold corresponding to the target pest to obtain the proximity of the control threshold. When the proximity of the control threshold reaches a preset threshold proximity coefficient, it is determined that there are sampling points within the resampling window that are close to the control threshold; when the number of pests trapped at any sampling point within the resampling window reaches or exceeds the control threshold, it is directly determined that there are sampling points within the resampling window that have reached the control threshold. After completing the above calculations, the window fluctuation discrimination vector of the resampling window is jointly constructed by the window growth rate, window coefficient of variation, window peak-to-valley difference, and control threshold proximity. Each item in the window fluctuation discrimination vector is compared with the corresponding preset threshold; the window growth rate threshold, window coefficient of variation threshold, window peak-to-valley difference threshold, and preset threshold proximity coefficient can be pre-calibrated based on historical monitoring data of the target pest, local control thresholds, and the error range of the monitoring equipment. Specifically, first, continuous trapping and counting data for one or more occurrence seasons of the target pest are selected. The growth rate, coefficient of variation, peak-to-trough difference, and the ratio of the maximum number of pests trapped to the control threshold are calculated for each monitoring window. Then, the statistical quantile value that can distinguish between stable windows and windows with increased fluctuations is used as the corresponding threshold. The preset threshold proximity coefficient can be set to a value between 0.8 and 1.0. When the ratio of the maximum number of pests trapped to the control threshold in the resampling window reaches this preset threshold proximity coefficient, it is determined that there are sampling points in that window that are close to the control threshold.

[0031] When at least two of the window growth rate, window coefficient of variation, window peak-to-valley difference, and proximity to the prevention and control threshold meet their respective preset thresholds, a fluctuation enhancement marker is generated to indicate that the resampled window is in a state of increased fluctuation. When fewer than two of the items meet their respective preset thresholds, a stability marker is generated to indicate that the resampled window is in a stable state. For resampled windows that already have sampling points that have reached or are close to the prevention and control threshold, even if the other fluctuation indicators do not meet their respective preset thresholds, the window is included in the event protection priority processing scope. After obtaining the window fluctuation discrimination vector and fluctuation enhancement marker, event protection resampling is performed on the resampled window. Event protection resampling is performed in the following order: priority for overlapping prevention and control response event segments, second priority for sampling points close to the threshold, retention of the maximum value for fluctuation enhancement windows, and performance of stability aggregation for stable windows.

[0032] When a resampling window overlaps with any original control response event segment in time, the resampling window is designated as an event-related window. For event-related windows, it is first determined whether the resampling window covers the trigger start time, segment peak time, or segment termination time of the original control response event segment. If the resampling window covers any of the aforementioned key sampling times, the pest trapping quantity and sampling time at the corresponding key sampling time are preferentially retained as the representative trapping quantity and representative sampling time of the resampling window, or as the event protection record of the resampling window.

[0033] When a resampling window overlaps with multiple key sampling times within the same original control response event segment, the priority of retaining items in the resampling window is determined according to the priority of trigger start time, segment peak time, and segment release time. Specifically, when two or three of the trigger start time, segment peak time, and segment release time simultaneously fall within the same resampling window, the pest trapping quantity and sampling time corresponding to the trigger start time are retained first; when there is no trigger start time but there are segment peak time and segment release time within the resampling window, the pest trapping quantity and sampling time corresponding to the segment peak time are retained first; when there is only the segment release time within the resampling window, the pest trapping quantity and sampling time corresponding to the segment release time are retained. When a resampling window overlaps with multiple original control response event segments, the original control response event segments are prioritized for processing according to the segment event number of the original control response event segments in ascending order; under the same segment event number, the retention items are further determined according to the priority of trigger start time, segment peak time, and segment release time. For other key sampling moments that are not identified as the sampling points represented by this resampling window, establish a subordinate association record with this resampling window for matching and calling when calculating the prevention and control event coverage C(t), prevention and control response delay penalty P(t), and phase transition retention G(t) in the subsequent calculation.

[0034] When a resampling window does not overlap with any original control response event segment in time, but there are sampling points within the window that reach or are close to the control threshold, the resampling window is treated as a threshold proximity window. For a threshold proximity window, if there are sampling points that reach the control threshold, the insect trapping quantity and sampling time corresponding to the earliest sampling point that reaches the control threshold are retained first; if there are no sampling points that reach the control threshold but there are sampling points that are close to the control threshold, the insect trapping quantity and sampling time corresponding to the sampling point with the highest control threshold proximity are retained; when there are multiple sampling points with the same control threshold proximity, the earliest sampling point is retained. When a resampling window does not overlap with any original control response event segment in time, and there are no sampling points that reach or are close to the control threshold, but the fluctuation enhancement marker indicates that the resampling window is in a fluctuation enhancement state, the maximum insect trapping quantity and its corresponding sampling time within the resampling window are retained as the representative trapping quantity and representative sampling time of the resampling window. This processing method is used to retain local peaks or significant fluctuations in insect populations within non-control response event segments, providing support for subsequent calculations of trend fidelity F(t) and stage transition retention G(t). When the corresponding resampling window does not overlap with any original control response event segment in time, and there are no sampling points reaching or approaching the control threshold, and the fluctuation enhancement marker indicates that the resampling window is in a stable state, a stationary aggregation is performed on the number of pests trapped within the resampling window to generate a representative trapped number and a representative sampling time for the resampling window. In one specific implementation, the stationary aggregation can use one of the mean, median, or cumulative value; when subsequent evaluation focuses more on the stable change in insect population density, the mean or median is used as the representative trapped number; when subsequent evaluation focuses more on the cumulative trapped number within the candidate monitoring interval, the cumulative value is used as the representative trapped number. The representative sampling time can be the start time, center time, or end time of the resampling window; while maintaining alignment with the subsequent time axis, the center time of the resampling window is preferably used as the representative sampling time.

[0035] After completing the above processing for each resampling window under the candidate monitoring interval t, the representative trap quantity and representative sampling time generated by each resampling window are arranged according to the order of representative sampling times to obtain the resampling monitoring sequence X(t) corresponding to the candidate monitoring interval t. For resampling windows containing event protection records or subordinate association records, the record and the corresponding representative sampling point are written into the auxiliary field of the resampling monitoring sequence X(t) for subsequent processing in S4, including prevention event coverage C(t), prevention response delay penalty P(t), stage transition retention G(t), and time axis alignment. When generating the resampling monitoring sequence X(t), at least one window processing result is output for each resampling window. For valid windows, the output includes the representative trap quantity, representative sampling time, window number, candidate monitoring interval t, fluctuation enhancement flag, event association flag, and the corresponding original prevention response event segment event sequence number; for empty windows, the output includes the missing flag, window number, and candidate monitoring interval t. If an empty window is located between two valid windows, it will be processed according to the missing marker when the time axis is aligned in S4. If empty windows appear consecutively and cause the corresponding candidate monitoring interval t to fail to meet the event coverage requirements, the candidate monitoring interval t will be removed in S5 or the safety rollback logic will be triggered.

[0036] S4, Time Axis Alignment and Event Protection Quality Index Calculation, specifically includes the following steps: After completing S3 and obtaining the resampled monitoring sequence X(t) corresponding to each candidate monitoring interval t, for each candidate monitoring interval t in the candidate monitoring interval set T, time axis alignment and event protection quality index calculation are performed respectively. The resampled monitoring sequence X(t) includes the representative sampling time, representative trap quantity, window number, fluctuation enhancement marker, event association marker, corresponding original prevention and control response event segment event sequence number, and subordinate association record output by each resampled window.

[0037] First, the resampled monitoring sequence X(t) corresponding to the candidate monitoring interval t is mapped back to the original monitoring sequence according to its representative sampling time. The timeline forms a sequence with the original monitoring sequence. Time-aligned interval restoration sequence Specifically, using the original monitoring sequence The sampling times in the original timeline are used as the standard time axis. According to the resampling window to which each sampling time belongs, the representative trapping quantity of that resampling window is assigned to the corresponding position on the standard time axis. When an event protection record exists within a resampling window, the pest trapping quantity corresponding to the trigger start time, segment peak time, or segment release time in the event protection record is preferentially mapped to the same sampling time in the original time axis. For other original sampling times within the same resampling window that do not belong to the critical moments of event protection, the representative trapping quantity of that resampling window is used to fill the gap, so that the interval sequence can be restored. Compared with the original monitoring sequence It has a consistent time axis length and sampling time order. For resampling windows marked as empty windows in S3, missing markers are retained when the time axis is aligned; when a missing marker is located between two valid resampling windows, imputation can be performed based on the representative trapping quantity of adjacent valid resampling windows; when missing markers appear consecutively and affect the coverage judgment of the prevention and control response event segment, the candidate monitoring interval t is recorded as insufficient event coverage or insufficient trend maintenance in subsequent index calculations. The interval is then restored into a sequence. Subsequently, based on the original monitoring sequence Resampled monitoring sequence X(t), interval restored sequence And the original prevention and control response event segment set, respectively, calculate the information loss degree L(t), trend fidelity F(t), resource consumption degree R(t), prevention and control event coverage C(t), prevention and control response delay penalty degree P(t), and stage transition retention degree G(t) corresponding to the candidate monitoring interval t. Among them, the information loss degree L(t) is calculated according to the original monitoring sequence. The normalized difference between the information entropy of the original monitoring sequence and the information entropy of the resampled monitoring sequence X(t) is determined. Specifically, based on the same trap quantity grouping rule, the original monitoring sequences are statistically analyzed separately. The frequency of occurrence of each trap quantity interval in the resampled monitoring sequence X(t) is calculated, and then the original monitoring sequence is calculated separately. Information entropy The information entropy H(t) of the resampled monitoring sequence X(t) is calculated, and the information loss L(t) is calculated as follows: L(t) = | -H(t)| / max( , );in, This is a preset minimum value used to avoid a denominator of zero. The larger L(t) is, the greater the value of L(t) after resampling at the candidate monitoring interval t relative to the original monitoring sequence. The greater the information loss, the more significant the loss. This information loss degree L(t) is used in subsequent S5 to determine whether the candidate monitoring interval t will cause excessive loss of original insect information.

[0038] The calculation results show that: when t=2 days, L=0.023; when t=4 days, L=0.059; when t=7 days, L=0.099; when t=14 days, L=0.162; and when t=30 days, L=0.251. This indicates that the longer the interval, the greater the information loss.

[0039] Trend fidelity F(t) according to the original monitoring sequence Restored sequence with interval The weighted result of the Pearson correlation coefficient and the normalized dynamic time-normalized distance between them was determined. Specifically, in the formation of the original monitoring sequence... Interval restoration sequences with the same sampling time order Then, the original monitoring sequence was calculated. Restored sequence with interval Pearson correlation coefficient between and will Normalization Simultaneously, the original monitoring sequence was calculated. Restored sequence with interval The dynamic time-warped distance between them is normalized to The trend fidelity F(t) is then calculated as follows: F(t) = α· +(1-α)·(1- ); where α is the balance coefficient between trend correlation and consistency of fluctuation pattern, and the value of α ranges from 0 to 1. The larger F(t) is, the stronger the interval reconstruction sequence under the candidate monitoring interval t. The better the original monitoring sequence is preserved The population change trend, peak position, and fluctuation pattern are analyzed. The trend fidelity F(t) is used for trend fidelity threshold screening in subsequent S5.

[0040] The calculation results show that: when t=2 days, F=0.982; when t=4 days, F=0.950; when t=7 days, F=0.899; when t=14 days, F=0.824; and when t=30 days, F=0.736. This indicates that the longer the interval, the lower the trend fidelity.

[0041] The resource consumption rate R(t) is determined by weighting at least two of the following: the proportion of data collection frequency, the proportion of data upload volume, the proportion of storage usage, and the proportion of device wake-up energy consumption of the intelligent insect pest monitoring device under the candidate monitoring interval t. Specifically, using the resource consumption when monitoring is performed according to the preset minimum sampling interval as a benchmark, the proportion of sampling frequency, the proportion of data upload volume, the proportion of storage usage, and the proportion of device wake-up energy consumption corresponding to the candidate monitoring interval t are calculated respectively. For different types of intelligent insect pest monitoring devices, at least two of these can be selected for calculation based on the actual recordable resource items. The resource consumption rate R(t) can be determined as follows: R(t) = ·Rs(t)+ ·Ru(t)+ ·Rm(t)+ ·Re(t); where Rs(t) is the percentage of sampling times, Ru(t) is the percentage of data uploaded, Rm(t) is the percentage of storage usage, and Re(t) is the percentage of device wake-up energy consumption; , , and For the corresponding weight coefficients, the sum of the weight coefficients involved in the calculation is 1. , , and Determined based on the resource constraints of the intelligent insect monitoring equipment. When the equipment primarily operates through camera data acquisition and image recognition, [the following should be considered]: [Improvement / Enhancement / Temperature Conservation]. and When communication is restricted, improve When available storage capacity is limited, improve When the device has low remaining power or high power consumption during device wake-up, increase... Resource items not involved in resource consumption calculations have a weight of 0, while the sum of the weight coefficients involved in the calculation is 1. In the absence of significant resource bottlenecks, , , and We used values ​​of 0.25, 0.25, 0.20, and 0.30 respectively. A larger R(t) indicates higher equipment resource consumption during the candidate monitoring interval t; a smaller R(t) indicates lower resource consumption during the candidate monitoring interval t. The calculation results show that: when t=2 days, R=0.500; when t=4 days, R=0.250; when t=7 days, R=0.143; when t=14 days, R=0.071; and when t=30 days, R=0.033. This indicates that the longer the interval, the lower the resource consumption.

[0042] The prevention and control event coverage rate C(t) is determined as the ratio between the number of original prevention and control response event segments successfully covered by the resampled monitoring sequence X(t) and the total number of original prevention and control response event segments. Specifically, for each original prevention and control response event segment in the set of original prevention and control response event segments, its segment event number, trigger start time, segment peak time, and segment release time are read, and the corresponding event segment is checked in the event association markers and subordinate association records of the resampled monitoring sequence X(t) to see if it has been retained or covered. When there is an event protection record in the resampled monitoring sequence X(t) corresponding to a certain original prevention and control response event segment, and the event protection record covers at least one of the key sampling moments of the original prevention and control response event segment's trigger start time, segment peak time, or segment release time, the original prevention and control response event segment is determined to have been successfully covered by the candidate monitoring interval t. If none of the key sampling moments in the original prevention and control response event segment are included in the event protection record or subordinate association record of the resampled monitoring sequence X(t), the original prevention and control response event segment is determined to have not been successfully covered.

[0043] After the statistics are completed, the event coverage rate C(t) is calculated as follows: C(t) = / ;in, This represents the number of original prevention and control response event segments successfully covered under the candidate monitoring interval t. This represents the total number of event segments in the original prevention and control response event segment set. When the original prevention and control response event segment set is empty, C(t) takes a preset full coverage value, and insufficient event coverage is not used as the basis for removing the candidate monitoring interval t in subsequent S5.

[0044] The penalty for delay in the prevention and control response is determined according to the delay time of the resampling response time in the resampling monitoring sequence X(t) that matches the original prevention and control response event segment relative to the trigger start time of the original prevention and control response event segment. Specifically, for each successfully covered original prevention and control response event segment, the representative sampling time or event protection sampling time in the resampling monitoring sequence X(t) that matches the original prevention and control response event segment is determined as the resampling response time; preferably, when the trigger start time of the event segment is directly retained, the trigger start time is used as the resampling response time; when the trigger start time is not directly retained but the segment peak time or segment release time is retained, the earliest retained critical sampling time is used as the resampling response time.

[0045] For the j-th original prevention and control response event segment, let its trigger start time be... The resampling response time corresponding to the candidate monitoring interval t is j. If j, then the response penalty value for that event segment is calculated based on the time difference between the two. When ,j later than When j, the delayed trigger penalty is calculated according to the first penalty coefficient; when j is earlier than or equal to When j, the penalty for an early or synchronous response is calculated according to the second penalty coefficient, where the first penalty coefficient is greater than the second penalty coefficient. This rule is used to give delayed responses a higher constraint strength in subsequent screening. For original prevention and response event segments that are not successfully covered, the response penalty value of that event segment is recorded as the maximum penalty value.

[0046] The penalty for delayed prevention and control response, P(t), is calculated by averaging or weighted averaging the penalty values ​​of each original prevention and control response event segment. The weights can be determined based on the maximum number of traps, duration, or event number of the original prevention and control response event segment. A larger P(t) indicates a more significant delay in the prevention and control response relative to the original prevention and control response event segment at the candidate monitoring interval t. P(t) is used in subsequent S5 for screening based on preset allowable response delay values. The phase transition retention rate, G(t), is calculated by restoring the sequence according to the interval. Identified population stage transition points and original monitoring sequences The matching ratio between the identified population stage transition points is determined. Specifically, this is first based on the original monitoring sequence. Calculate the stage transition points of the target pest, including the time points from the rising phase to the peak phase, from the peak phase to the plateau phase, from the plateau phase to the trough phase, and from the trough phase to the second peak phase; then reconstruct the sequence based on the interval period. The corresponding stage transition points are calculated using the same stage identification rules. When identifying population stage transition points, the original monitoring sequence... The system is divided into pre-defined phase identification windows, and the average number of traps, window growth rate, peak-to-valley difference, and fluctuation intensity within each phase identification window are calculated. When the average number of traps in two consecutive phase identification windows continuously increases from a low level, and the window growth rate reaches a pre-defined growth threshold, the corresponding time point is identified as the phase transition point into the rising phase. When the number of traps reaches or approaches the prevention and control threshold, and forms a segment peak or prevention and control response event segment within an adjacent window, the corresponding time point is identified as the phase transition point into the peak phase. When the window growth rate decreases and the number of traps tends to stabilize near or below the prevention and control threshold, the corresponding time point is identified as the phase transition point into the stable phase. When the number of traps remains below the pre-defined low-density threshold, the corresponding time point is identified as the phase transition point into the trough phase. When continuous growth reappears after the trough phase and approaches or reaches the prevention and control threshold, the corresponding time point is identified as the phase transition point into the second peak phase.

[0047] For the original monitoring sequence At each stage transition point, the sequence is restored during the interval. Among the identified stage transition points, it is checked whether there exists a corresponding stage transition point whose time difference does not exceed the preset stage transition matching tolerance. If it exists, the stage transition point is determined to be maintained by the candidate monitoring interval t; if it does not exist, the stage transition point is determined not to be maintained. The stage transition maintenance degree G(t) can be determined as follows: G(t) = / ;in, This represents the number of stage transition points successfully matched during the candidate monitoring interval t. Original monitoring sequence The total number of stage transition points identified in the original monitoring sequence. When no stage transition point is identified, G(t) is set to a preset full retention value or is not used as a basis for elimination in subsequent qualified candidate screening. When calculating the above indicators, information loss L(t), trend fidelity F(t), resource consumption R(t), prevention and control event coverage C(t), prevention and control response delay penalty P(t), and stage transition retention G(t) are all correlated with the candidate monitoring interval t and stored using the candidate monitoring interval t as an index. Each indicator record corresponding to a candidate monitoring interval t includes at least: candidate monitoring interval t, information loss L(t), trend fidelity F(t), resource consumption R(t), prevention and control event coverage C(t), prevention and control response delay penalty P(t), stage transition retention G(t), the event number of the successfully covered original prevention and control response event segment, the event number of the uncovered original prevention and control response event segment, the resampling response time, and the stage transition point matching result.

[0048] S5. Selection of Qualified Candidate Intervals and Optimization of Resource Budget Constraints, specifically including the following steps: After completing S4 and obtaining the information loss L(t), trend fidelity F(t), resource consumption R(t), prevention event coverage C(t), prevention response delay penalty P(t), and stage transition retention G(t) for each candidate monitoring interval t in the candidate monitoring interval set T, first obtain the edge resource budget status of the intelligent pest monitoring equipment, then select qualified candidate monitoring intervals based on the event protection quality index and the edge resource budget status, and determine the optimal monitoring interval for this round from the qualified candidate monitoring intervals. The edge resource budget status is collected or reported by the intelligent insect pest monitoring device before the current round of interval optimization. The edge resource budget status includes at least two of the following: remaining power, available storage capacity, and communication status. For intelligent insect pest monitoring devices with battery or solar power modules, their current remaining power or remaining power level is read; for intelligent insect pest monitoring devices with local cache or local storage modules, their available storage capacity is read; for intelligent insect pest monitoring devices that need to upload insect pest data via wireless network, their communication status is read, which may include whether the communication link is available, communication signal strength, data upload success rate, or communication latency. For resource items that cannot be obtained, they are marked as unavailable and do not participate in the current round of resource budget constraint calculation; for resource items that have been obtained, their corresponding resource status level is determined according to preset classification rules. The edge resource budget status is divided into resource-sufficient, resource-limited, and resource-scarce states. If the remaining battery power is not lower than the preset battery power threshold, the available storage capacity is not lower than the preset storage threshold, and the communication status meets the preset communication requirements, the resource status is determined to be sufficient. If any of the remaining battery power, available storage capacity, or communication status is close to the corresponding preset lower limit but can still support subsequent monitoring, the resource status is determined to be limited. If the remaining battery power is lower than the preset lower limit, the available storage capacity is lower than the preset lower limit, or the communication status continuously fails to meet the upload requirements, the resource status is determined to be strained. This resource status level is used to subsequently determine the constraint strength and weighting coefficient of the resource consumption degree R(t).

[0049] Subsequently, based on the target pest's trapping method, population development stage, control threshold, monitoring application requirements, and marginal resource budget status, weighting coefficients were determined for information loss L(t), trend fidelity F(t), resource consumption R(t), control event coverage C(t), control response delay penalty P(t), and stage transition retention G(t), respectively. These weighting coefficients are used to calculate the comprehensive monitoring quality value Q(t) within the set of qualified candidate monitoring intervals. The sum of all weighting coefficients is 1. The weighting coefficients corresponding to information loss L(t), trend fidelity F(t), resource consumption R(t), control event coverage C(t), control response delay penalty P(t), and stage transition retention G(t) are denoted as follows: , , , , and And satisfy + + + + + =1. The base values ​​for each weight coefficient are first determined based on the monitoring application requirements, and then adjusted according to the population development stage and edge resource budget status. When the monitoring application requirement is triggered by a prevention and control threshold, priority is given to constraining the coverage of the prevention and control response event segment and the response delay. The base weights can be set as follows: =0.10、 =0.20、 =0.10、 =0.25、 =0.25、 =0.10; When the monitoring application requires population dynamics research, priority should be given to maintaining insect infestation information, change trends, and stage transitions. The basic weight can be set as follows: =0.20、 =0.25、 =0.10、 =0.10、 =0.10、 =0.25; When the monitoring application requires long-term low-density monitoring, priority should be given to controlling equipment resource consumption, and the basic weight can be set as follows: =0.10、 =0.15、 =0.35、 =0.15、 =0.15、 =0.10. After the basic weight is determined, if the target pest is in the population growth phase, peak phase, or re-growth phase, then the weight will be adjusted accordingly. , and Increase by 0.05, and from , and The same total amount will be deducted proportionally; if the target pest is in a stable, declining, or trough period, then... Increase by 0.10, and from , and A deduction of 0.10 will be applied proportionally. If the marginal resource budget status is resource-constrained, then the budget will be adjusted based on the phased revision. Increase by 0.05; if the marginal resource budget status is resource-scarce, then... Increase by 0.10 to 0.15, and from , and The amount shall be deducted proportionally, but shall not be reduced. and To maintain coverage of prevention and control events and constraints on response latency, the value should be kept below 0.10. This setting can be configured when the monitoring application requires a prevention and control threshold trigger, the target pest is at its peak, and the edge resource budget is in a resource-constrained state. =0.05、 =0.23、 =0.10、 =0.27、 =0.27、 =0.08; When the monitoring application requires long-term low-density monitoring, the target pest is in a low period, and the marginal resource budget status is resource-scarce, it can be set to... =0.05、 =0.10、 =0.50、 =0.15、 =0.15、 =0.05. After each correction, all weight coefficients are normalized so that their sum is 1.

[0050] When determining the weighting coefficients, the basic weights are first determined based on the monitoring application requirements, and then adjusted according to the population occurrence stage and the marginal resource budget status. When the monitoring application requirement is to trigger the prevention and control threshold, the weights corresponding to the prevention and control event coverage rate C(t), the prevention and control response delay penalty P(t), and the trend fidelity F(t) are increased to ensure that the candidate monitoring intervals prioritize the coverage of prevention and control response event segments and the timeliness of response. When the monitoring application requirement is to study population dynamics, the weights corresponding to the information loss L(t), the trend fidelity F(t), and the stage transition retention G(t) are increased to ensure that the candidate monitoring intervals prioritize the retention of pest information, trend patterns, and stage transition points. When the monitoring application requirement is to conduct long-term low-density monitoring, the weight corresponding to the resource consumption R(t) is increased, while ensuring that the prevention and control event coverage rate C(t) and the prevention and control response delay penalty P(t) are not lower than the subsequent screening constraints.

[0051] When the target pest is in its population growth phase, peak phase, or secondary peak phase, the weights corresponding to the control event coverage rate C(t), control response delay penalty P(t), and trend fidelity F(t) are increased, while the impact of simple resource conservation on the overall evaluation is reduced. When the target pest is in a stable or low-lying phase, the weight corresponding to the resource consumption rate R(t) is increased, provided that the constraints on control event coverage and response delay are met. If the marginal resource budget status is resource-scarce, the weight corresponding to the resource consumption rate R(t) is increased, and stricter resource constraints are set for the resource consumption rate R(t). If the marginal resource budget status is resource-sufficient, the weights are determined according to the monitoring application needs and the stage of population occurrence, without further relaxing the control event coverage constraints due to resource status. After determining the weight coefficients, each candidate monitoring interval t in the candidate monitoring interval set T is screened for eligibility. The eligibility screening is performed before the calculation of the overall monitoring quality value Q(t) and is used to exclude candidate monitoring intervals that, although showing significant resource conservation, cannot meet the protection requirements of the control response. Specifically, for each candidate monitoring interval t, its trend fidelity F(t), prevention and control event coverage C(t), prevention and control response delay penalty P(t), and phase transition retention G(t) are read sequentially and compared with preset trend fidelity threshold, preset event coverage threshold, preset response delay allowable value, and preset phase transition retention threshold, respectively.

[0052] When the trend fidelity F(t) corresponding to the candidate monitoring interval t is lower than the preset trend fidelity threshold, it is determined that the candidate monitoring interval t cannot maintain the original monitoring sequence. The candidate monitoring interval t is removed based on the pest population change trend. If the coverage rate C(t) of the control event corresponding to the candidate monitoring interval t is lower than the preset event coverage threshold, the candidate monitoring interval t is determined to be unable to meet the coverage requirements of the original control response event segment set and is removed. If the control response delay penalty P(t) corresponding to the candidate monitoring interval t is higher than the preset allowable response delay value, the candidate monitoring interval t is determined to have a control response delay risk and is removed. If the stage transition retention rate G(t) corresponding to the candidate monitoring interval t is lower than the preset stage transition retention threshold, the candidate monitoring interval t is determined to be unable to maintain the original monitoring sequence. The candidate monitoring interval t is removed from the population stage transition point.

[0053] After the above screening, the candidate monitoring intervals that were not eliminated constitute the set of qualified candidate monitoring intervals. Each candidate monitoring interval t in this set of qualified candidate monitoring intervals meets the minimum requirements for trend fidelity, prevention and control event coverage, prevention and control response delay, and phase transition maintenance, and is then included in the subsequent resource budget constraint and comprehensive evaluation processing.

[0054] If the set of qualified candidate monitoring intervals is not empty, resource constraints are further assessed based on the edge resource budget status. For each qualified candidate monitoring interval t, the expected number of samples, expected data upload volume, expected storage usage, and expected device wake-up energy consumption corresponding to that interval are compared with the maximum allowed number of samples, maximum data upload volume, maximum storage usage, and maximum wake-up energy consumption under the current edge resource budget status. If any expected resource consumption exceeds the corresponding allowed upper limit, the candidate monitoring interval t is removed from the resource-executable candidate set; if none of the expected resource consumptions exceed the corresponding allowed upper limit, the candidate monitoring interval t is retained. Even under resource-scarce conditions, candidate monitoring intervals that cause the prevention event coverage rate C(t) to be lower than the preset event coverage threshold or the prevention response delay penalty P(t) to be higher than the preset allowed response delay value must not be selected. For each qualified candidate monitoring interval t, its resource consumption R(t) and the current remaining power, available storage capacity, and communication status of the intelligent insect monitoring device are used to determine whether the candidate monitoring interval t meets the resource budget constraints. If the expected number of data collections, expected data upload volume, expected storage usage, or expected device wake-up energy consumption corresponding to a candidate monitoring interval t exceeds the range allowed by the current edge resource budget status, then the candidate monitoring interval t will be removed from the resource executable candidate set. If its expected resource consumption does not exceed the corresponding resource constraint, then the candidate monitoring interval t will be retained as a resource executable candidate monitoring interval. When the edge resource budget status is in a resource-sufficient state, the resource budget constraint is used to confirm that the qualified candidate monitoring interval t can be executed by the device without further narrowing the candidate range. When the edge resource budget status is in a resource-constrained state, candidate monitoring intervals with lower resource consumption R(t) and still meeting the screening conditions of F(t), C(t), P(t), and G(t) will be retained first. When the edge resource budget status is in a resource-scarce state, longer candidate monitoring intervals that meet the minimum requirements of prevention and control event coverage C(t) and prevention and control response delay penalty P(t) will be retained first, but candidate monitoring intervals that cause prevention and control event coverage C(t) to be lower than the preset event coverage threshold or prevention and control response delay penalty P(t) to be higher than the preset response delay allowable value shall not be selected.

[0055] For candidate monitoring intervals t that meet the eligibility screening and resource budget constraints, calculate their comprehensive monitoring quality value Q(t). In one specific implementation, the information loss degree L(t), resource consumption degree R(t), and prevention and control response delay penalty degree P(t) are converted into corresponding positive evaluation terms, and the trend fidelity degree F(t), prevention and control event coverage C(t), and phase transition retention degree G(t) are directly used as positive evaluation terms. The comprehensive monitoring quality value Q(t) can be determined as follows: Q(t) = ·[1-L(t)]+ · F(t)+ · [1-R(t)]+ · C(t)+ ·[1-P(t)]+ ·G(t); where, , , , , and These are the weighting coefficients for information loss L(t), trend fidelity F(t), resource consumption R(t), prevention and control event coverage C(t), prevention and control response delay penalty P(t), and phase transition retention G(t), respectively, and the sum of each weighting coefficient is 1. When P(t) is not normalized to the range of 0 to 1, P(t) is first normalized according to the preset allowable value of response delay or the maximum penalty value, and then it is used in the calculation of the comprehensive monitoring quality value Q(t).

[0056] The penalty for delay in prevention and control response, P(t), is determined according to the following formula: P(t) = Σ[ · ] / Σ ; in, Let j be the event weight of the j-th original prevention and control response event segment. The number of traps can be determined based on the maximum number of traps per segment, the duration of the segment, or the event number of the segment. When all event segments are of equal importance, Take the same value. The single-event response penalty value is the j-th original prevention and control response event segment under the candidate monitoring interval t. It is used to characterize the degree to which the resampling response time of the event segment is earlier or later than its trigger start time. When the resampling response time is later than the trigger start time, the first penalty coefficient is used to calculate the delayed response penalty. When the resampling response time is earlier than or equal to the trigger start time, the second penalty coefficient is used to calculate the earlier or synchronous response penalty. The first penalty coefficient is greater than the second penalty coefficient.

[0057] After the calculation is completed, the candidate monitoring interval with the largest comprehensive monitoring quality value Q(t) is selected from the set of executable resource candidates as the optimal monitoring interval for this round. When two or more candidate monitoring intervals have the same or nearly the same comprehensive monitoring quality value Q(t), the selection rule is determined according to the monitoring application requirements: if the monitoring application requirement is to trigger a prevention and control threshold, the candidate monitoring interval with a smaller prevention and control response delay penalty P(t) is preferred; if the monitoring application requirement is to study population dynamics, the candidate monitoring interval with a smaller information loss L(t) or a larger stage transition retention G(t) is preferred; if the monitoring application requirement is to conduct long-term low-density monitoring, the candidate monitoring interval with a smaller resource consumption R(t) is preferred. The determined candidate monitoring interval is taken as the optimal monitoring interval for this round. And output it to the subsequent S6.

[0058] If the set of qualified candidate monitoring intervals is empty, or if the set of qualified candidate monitoring intervals is not empty but no executable candidate monitoring interval exists after resource budget constraints, a safety rollback is performed. The safety rollback process includes: determining the smaller of the preset minimum sampling interval or the previous round's effective optimal monitoring interval as the safe monitoring interval, and using this safe monitoring interval as the optimal monitoring interval t_opt for the current round. If no previous round's effective optimal monitoring interval exists, the preset minimum sampling interval is directly determined as the safe monitoring interval. During the safety rollback process, a rollback reason flag is generated simultaneously. The rollback reason flag records the reason why the normal optimal monitoring interval cannot be determined from the candidate monitoring interval set T in this round. The rollback reason flag includes at least one of the following: insufficient trend fidelity flag, insufficient event coverage flag, excessive response delay flag, insufficient stage transition retention flag, and insufficient resource budget flag. If a candidate monitoring interval is removed because F(t) is lower than the preset trend fidelity threshold, a trend fidelity deficiency flag is generated; if a candidate monitoring interval is removed because C(t) is lower than the preset event coverage threshold, an event coverage deficiency flag is generated; if a candidate monitoring interval is removed because P(t) is higher than the preset allowable response delay value, a response delay exceedance flag is generated; if a candidate monitoring interval is removed because G(t) is lower than the preset stage transition retention threshold, a stage transition retention deficiency flag is generated; if a candidate monitoring interval meets the event protection quality requirements but does not meet the resource constraints corresponding to the edge resource budget state, a resource budget deficiency flag is generated.

[0059] After generating the rollback reason flag, the optimal monitoring interval for this round will be... The rollback reason marker, the monitoring interval of the removed candidate and its corresponding F(t), C(t), P(t), G(t) and R(t) indicators are saved together for subsequent confidence hysteresis dynamic updates in S6 and recovery judgment in the next round of optimization.

[0060] During the next dynamic update, the monitoring quality indicators corresponding to the rollback reason markers are checked first. If the rollback reason marker is a trend fidelity insufficiency marker, it is first determined whether F(t) corresponding to the candidate monitoring interval has reached the preset trend fidelity recovery threshold; if the rollback reason marker is an event coverage insufficiency marker, it is first determined whether C(t) has reached the preset event coverage recovery threshold; if the rollback reason marker is a response latency exceeding the limit marker, it is first determined whether P(t) is not higher than the preset response latency recovery threshold; if the rollback reason marker is a phase transition maintenance insufficiency marker, it is first determined whether G(t) has reached the preset phase transition maintenance recovery threshold; if the rollback reason marker is a resource budget insufficiency marker, it is first determined whether the remaining power, available storage capacity, or communication status has recovered to the corresponding resource recovery threshold. When the monitoring quality indicators corresponding to the rollback reason markers continuously reach the corresponding preset recovery thresholds, the safe rollback state is exited, and the optimal monitoring interval selection based on the qualified candidate monitoring interval set is restored. Continuous achievement can be set to meet the corresponding recovery thresholds for two or more consecutive update cycles to avoid frequent exits and re-entries into the safe rollback state due to single data fluctuations. When the monitoring quality indicator corresponding to the rollback reason marker does not reach the corresponding preset recovery threshold, the safe monitoring interval will continue to be maintained, and the corresponding indicator will continue to be verified in the next round of dynamic updates.

[0061] S6, Device-side execution and confidence hysteresis dynamic update, specifically includes the following steps: After completing S5 and obtaining the optimal monitoring interval for this round. Then, the optimal monitoring interval will be determined. Output to intelligent insect pest monitoring equipment. Optimal monitoring interval. The optimal monitoring interval can be the candidate monitoring interval with the largest comprehensive monitoring quality value Q(t) among the qualified candidate monitoring interval sets, or it can be the safe monitoring interval determined when the qualified candidate monitoring interval set is empty or the resource budget is not met. The intelligent insect pest monitoring equipment receives the optimal monitoring interval. Then, based on the equipment type and controllable components, the optimal monitoring interval is determined. Converted into device-side execution parameters.

[0062] For color-sensitive intelligent monitoring devices, the optimal monitoring interval will be determined. This involves adjusting at least two of the following: the wake-up cycle of the high-definition camera, the image acquisition cycle of the sticky insect board, the operation cycle of the image recognition module, the data upload cycle, and the local cache clearing cycle. Specifically, when When the time frame is short, increase the wake-up frequency of the high-definition camera and the operating frequency of the image recognition module, and correspondingly shorten the data upload cycle; when For extended periods, reduce the wake-up frequency of the high-definition camera and the operating frequency of the image recognition module, and correspondingly extend the data upload cycle or increase the proportion of local cached batch uploads. For sexually induced intelligent monitoring devices, the optimal monitoring interval should be... This involves adjusting at least two of the following: the data aggregation cycle, the counting result upload cycle, the local cache clearing cycle, and the auxiliary image acquisition or status detection cycle of the automatic counting device. Specifically, the automatic counting device can continuously record the counting signal generated when pests enter the trap or counting area, but according to... The counting signals are summarized, encapsulated, and uploaded; when When the time is short, shorten the cycle for summarizing and uploading the counting results; when For longer periods, extend the period for summarizing and uploading the counting results to reduce the number of communications and storage pressure.

[0063] For light-induced intelligent monitoring equipment, the optimal monitoring interval period will be determined. This involves adjusting at least two of the following: the insect-attracting light source's operation cycle, the image acquisition component's or sensor counting component's operating cycle, the data upload cycle, and the local cache clearing cycle. If the light-attracting device needs to operate within a fixed attraction period, then, without changing the necessary attraction period, according to... Adjust the execution cycle for data collection, identification, counting, aggregation, and uploading. After updating the execution parameters on the device side, record the current execution monitoring interval as... If the optimal monitoring interval for this round is If adopted directly, then = If the switch is not initiated due to subsequent confidence hysteresis judgment, then The monitoring interval from the previous round of equipment execution remains unchanged. The equipment will follow the same procedures during subsequent monitoring processes. Collect or summarize the target pest trapping count data, and record the newly added trapping count data along with the corresponding sampling time, equipment identification, monitoring interval, and equipment resource status.

[0064] Dynamic updates are initiated when the number of newly added trap counts reaches a preset update data volume or a preset update cycle. The preset update data volume can be a preset number of new sampling points or a preset number of consecutive monitoring days; the preset update cycle can be a fixed period, such as every 7 days, every 15 days, or every 30 days. After dynamic updates are initiated, the newly added trap counts are incorporated into the original monitoring sequence. The updated original monitoring sequences were processed according to the preprocessing rules in S1. The process involves time-series sorting, removal of invalid sampling points, marking or imputing missing sampling points, and marking single-point outliers. This results in the updated original monitoring sequence. Then, based on the updated original monitoring sequence Calculate the population growth rate, peak-to-trough variation characteristics, fluctuation intensity, and stage identification confidence level of the target pest. The population growth rate can be determined based on the change between the average number of traps in the current time window and the average number of traps in the previous time window; the peak-to-trough variation characteristics can be determined based on the maximum number of traps, the minimum number of traps, and the difference between them in the current time window; the fluctuation intensity can be determined based on the coefficient of variation, standard deviation, or peak-to-trough difference of the number of traps in the current time window.

[0065] Using data from the most recent update cycle as the current window and data from the previous update cycle as the control window, the following criteria are used to determine whether a population is entering a growth phase or a second peak: If the average number of traps in the current window is significantly higher than that in the control window, and the growth rate reaches a preset growth threshold, it is considered a peak period; if the number of traps in the current window remains consistently high, and there is an original control response event segment or a segment peak, it is considered a peak period; if the number of traps in the current window changes little, and no new control response event segments appear, it is considered a stable period; if the number of traps in the current window remains consistently low, below the control threshold or below the preset low-density threshold, it is considered a trough period; if continuous growth resumes after the trough period and approaches or reaches the control threshold, it is considered a second peak period. The confidence level for stage identification is determined based on the population growth rate, peak and trough variation characteristics, fluctuation intensity, occurrence of control response event segments, and data quality. Specifically, when the population growth rate, peak and trough variation characteristics, and fluctuation intensity give a consistent judgment on the same population occurrence stage, and the updated original monitoring sequence When the number of valid sampling points reaches the preset number of valid samples, the confidence level of stage identification is increased; when the device is offline, image recognition fails, automatic counting is abnormal, there is a lot of data during the replacement of the trap, or the proportion of missing sampling points exceeds the preset proportion, the confidence level of stage identification is decreased; when the judgment criteria of different stages are contradictory, the confidence level of stage identification is decreased.

[0066] After identifying the current population stage of the target pest and the corresponding stage identification confidence level, the candidate monitoring interval set T and indicator weights are adjusted based on the stage identification results. When the target pest is in the rising, peak, or secondary peak stage of the population, and the stage identification confidence level reaches the preset confidence threshold, the upper limit of the candidate monitoring interval values ​​in the candidate monitoring interval set T is reduced. Specifically, the candidate monitoring interval values ​​in the candidate monitoring interval set T can be limited to candidate values ​​no greater than the previous round's effective optimal monitoring interval, or the upper limit of the candidate monitoring interval values ​​can be adjusted to a shorter time boundary corresponding to the allowable time limit for control response. At the same time, the weights corresponding to the control event coverage rate C(t), control response delay penalty P(t), and trend fidelity F(t) are increased, so that in subsequent S5 screening and comprehensive evaluation, candidate monitoring intervals that can cover the control response event segment, reduce response delay, and maintain the pest trend are prioritized. When the target pest is in a stable or trough stage, and the stage identification confidence level reaches the preset confidence threshold, the upper limit of the candidate monitoring interval values ​​in the candidate monitoring interval set T is expanded. Specifically, the candidate monitoring interval set T can be expanded to include candidate values ​​greater than the previous round's optimal effective monitoring interval. Alternatively, longer candidate monitoring intervals can be added within the limits of the maximum allowable monitoring interval and the allowable time limit for prevention and control response. Simultaneously, the weight corresponding to resource consumption R(t) can be increased, ensuring that subsequent S5 prioritizes candidate monitoring intervals with lower resource consumption, provided that the screening constraints of prevention and control event coverage C(t), prevention and control response delay penalty P(t), trend fidelity F(t), and stage transition retention G(t) are met.

[0067] When the confidence level of stage identification is lower than the preset confidence threshold, the candidate monitoring interval set T is not directly expanded according to the identified population occurrence stage. In this case, the current monitoring interval is maintained. Alternatively, the current monitoring interval can be rolled back to a shorter safe monitoring interval. If the device is currently in a safe rollback state, the safe monitoring interval will continue to be maintained; if the device is not currently in a safe rollback state, but the stage identification confidence is low and the coverage of prevention and response events is uncertain, the candidate monitoring interval set T will be limited to no more than the current monitoring interval. Within the candidate value range, to avoid relaxing the monitoring interval due to uncertainty in stage identification. After completing the adjustment of the candidate monitoring interval set T and weight coefficients, steps S2 to S5 are re-executed. When re-executing S2, based on the updated original monitoring sequence... The updated original set of control response event segments, the adjusted range of candidate monitoring interval values, and the allowable time limit for control responses are used to generate a new set of candidate monitoring intervals T. When S3 is re-executed, each candidate monitoring interval t in the new set of candidate monitoring intervals T undergoes event-protected resampling. When S4 is re-executed, new L(t), F(t), R(t), C(t), P(t), and G(t) are calculated. When S5 is re-executed, a new optimal monitoring interval is determined based on the updated edge resource budget status, the adjusted weighting coefficients, and the new event-protected quality index. After obtaining the new optimal monitoring interval, it is not immediately replaced as the equipment-executed monitoring interval; instead, a confidence hysteresis judgment is performed. The confidence hysteresis judgment is used to avoid frequent switching of the equipment-executed monitoring interval due to short-term pest fluctuations, data anomalies, or stage identification errors.

[0068] The confidence hysteresis judgment includes stage consistency judgment and change magnitude judgment. Stage consistency judgment is: if the population identified in two consecutive update cycles is in the same stage, and the stage identification confidence level in both update cycles reaches a preset confidence threshold, then the stage consistency update condition is met. Change magnitude judgment is: if the new optimal monitoring interval is relative to the current execution monitoring interval... If the change in the magnitude reaches a preset switching magnitude threshold, the change magnitude update condition is met. The change magnitude can be determined by the absolute difference between the two or by their relative change ratio. When either the phase consistency update condition or the change magnitude update condition is met, the new optimal monitoring interval is updated to the device's execution monitoring interval, and the updated device execution monitoring interval is sent to the intelligent insect pest monitoring device, causing it to readjust at least two of the following: camera wake-up cycle, image recognition module operation cycle, automatic counting cycle, data upload cycle, and local cache clearing cycle. When neither the phase consistency update condition nor the change magnitude update condition is met, the current execution monitoring interval is maintained. The process remains unchanged, and the new optimal monitoring interval is recorded locally or on the server as a candidate update result for comparison during the next round of dynamic updates. When S5 generates a rollback reason flag, S6 prioritizes verifying the monitoring quality indicators corresponding to the rollback reason flag during the next dynamic update. If the rollback reason flag is a trend fidelity insufficient flag, then it prioritizes verifying whether the trend fidelity F(t) reaches the preset trend fidelity recovery threshold; if the rollback reason flag is an event coverage insufficient flag, then it prioritizes verifying whether the prevention event coverage rate C(t) reaches the preset event coverage recovery threshold; if the rollback reason flag is a response delay exceeding the limit flag, then it prioritizes verifying whether the prevention response delay penalty P(t) is not higher than the preset response delay recovery threshold; if the rollback reason flag is a phase transition retention insufficient flag, then it prioritizes verifying whether the phase transition retention rate G(t) reaches the preset phase transition retention recovery threshold; if the rollback reason flag is a resource budget insufficient flag, then it prioritizes verifying whether the remaining power, available storage capacity, or communication status has recovered to the corresponding resource recovery threshold. When the monitoring quality indicator corresponding to the rollback reason marker continuously reaches the corresponding preset recovery threshold, the safe rollback state is exited, and the optimal monitoring interval selection based on the set of qualified candidate monitoring intervals is restored. When the monitoring quality indicator corresponding to the rollback reason marker does not reach the corresponding preset recovery threshold, the safe monitoring interval is maintained, and the corresponding indicator continues to be prioritized for verification in the next round of dynamic updates.

[0069] Example 1: Tea Green Leafhopper (Intelligent Color-Taxis Monitoring Device) This embodiment uses the tea green leafhopper as the monitoring target. Color-attracting intelligent monitoring equipment is deployed within the tea garden monitoring area. The equipment includes a yellow sticky insect board, a high-definition camera, and an image recognition module. The high-definition camera captures images of the sticky insect board at preset minimum sampling intervals. The image recognition module identifies and counts adult tea green leafhoppers in the images, forming an initial monitoring sequence. For the original monitoring sequence After performing time-series sorting, invalid sampling point removal, missing sampling point marking or imputation, and single-point outlier marking, an original set of control response event segments is constructed based on the control threshold corresponding to the tea garden green leafhopper. Each original control response event segment includes the trigger start time, segment peak time, segment end time, segment maximum trapping quantity, and segment event number.

[0070] The candidate monitoring interval set T is set to 1 day, 2 days, 3 days, 4 days, 5 days, 6 days, 7 days, 10 days, 14 days, 21 days, and 30 days. For each candidate monitoring interval t, the original monitoring sequence is... The system is divided into multiple temporally consecutive resampling windows. For resampling windows that overlap with the original control response event segment, priority is given to retaining the number of pests trapped and the sampling time corresponding to the trigger start time, the segment peak time, or the segment release time. For windows that do not overlap with the original control response event segment but are in a state of increased fluctuation, the maximum number of pests trapped within the window and its corresponding sampling time are retained. For stable windows, stationary aggregation is performed to obtain the resampling monitoring sequence X(t) for the corresponding candidate monitoring interval t. The resampling monitoring sequence X(t) is then mapped back to the original monitoring sequence. The time axis forms an interval restoration sequence. The parameters L(t), F(t), R(t), C(t), P(t), and G(t) are calculated. Then, qualified candidate monitoring intervals are selected based on preset trend fidelity thresholds, preset event coverage thresholds, preset response delay allowable values, and preset stage transition retention thresholds. Under edge resource budget constraints, the comprehensive monitoring quality value Q(t) is calculated, and the candidate monitoring interval with the largest Q(t) is selected as the optimal monitoring interval. The optimal monitoring interval will be determined. The data is output to the color-attracting intelligent monitoring device to adjust the wake-up cycle of the high-definition camera, the operation cycle of the image recognition module, the data upload cycle, and the local cache clearing cycle. Once the newly added trapping count data reaches the preset update cycle, the new data is incorporated into the original monitoring sequence. The population emergence stage and stage identification confidence level were re-identified, and the candidate monitoring interval set T and weight coefficients were adjusted according to the population emergence stage.

[0071] Example 2: Gray Tea Looper (Sexual Temptation Intelligent Monitoring Device) This embodiment uses the brown geometrid moth (Ectropis grisescens Warren) as the monitoring target and employs a sex-induced intelligent monitoring device to provide a method for determining the optimal interval period for intelligent pest monitoring. The method specifically includes the following steps: Step 1: Acquisition of Raw Monitoring Data: A pheromone-attracting intelligent monitoring device was deployed in a tea garden in Nanchang County, Jiangxi Province. This device uses a pheromone lure specifically for the tea geometrid moth, combined with a barrel-shaped trap, and automatically records the number of male adult tea geometrid moths captured using an infrared beam counting device. The minimum sampling interval was set to one day (i.e., the device automatically records and uploads data daily). Raw monitoring data was continuously collected for 180 days during the tea geometrid moth occurrence period from May to October 2025, totaling 180 data points.

[0072] Step 2: Similar to Example 1, calculate the information loss, trend fidelity, and resource consumption for each candidate interval.

[0073] Step 3: Calculate the monitoring quality evaluation indicators: For each candidate interval t, calculate the information loss degree L(t), trend fidelity F(t), and resource consumption degree R(t).

[0074] The calculation results show that: (1) Information loss: L=0.005 when t=2 hours; L=0.022 when t=6 hours; L=0.051 when t=12 hours; L=0.098 when t=24 hours; L=0.189 when t=48 hours; L=0.285 when t=72 hours. (2) Trend fidelity: F=0.999 when t=2 hours; F=0.991 when t=6 hours; F=0.974 when t=12 hours; F=0.942 when t=24 hours; F=0.882 when t=48 hours; F=0.814 when t=72 hours. (3) Resource consumption: Same as in Example 1, R(t)=1 / t (t in hours) Step 4: Set weights based on prevention and control decision requirements: In this embodiment, the main purpose of monitoring the brown tea geometrid moth is to trigger the prevention and control threshold (a prevention and control warning needs to be issued when the number of moths attracted per day exceeds 15 per trap). Therefore, the decision requirement type is "prevention and control threshold trigger". The weight coefficients are set as follows: w1=0.2, w2=0.6, w3=0.2. Step 5: Solve for the optimal monitoring interval. Table 1 Interval period t (days) 1-L(t) F(t) 1-R(t) Q(t) 1 1.000 1.000 0.000 0.600 2 0.981 0.988 0.500 0.895 3 0.964 0.974 0.667 0.928 4 0.948 0.962 0.750 0.936 5 0.932 0.948 0.800 0.935 6 0.918 0.934 0.833 0.931 7 0.905 0.921 0.857 0.927 10 0.882 0.892 0.900 0.909 14 0.851 0.856 0.929 0.877 21 0.814 0.816 0.952 0.839 30 0.775 0.774 0.967 0.797 The calculation results show that the comprehensive monitoring quality Q(t) is the largest (0.936) when t=4 days. Therefore, the optimal monitoring interval for the gray tea geometrid moth is t* = 4 days. The specific results are shown in Table 1.

[0075] Step 6: Effectiveness Verification: After adjusting the sampling interval of the sexually induced intelligent monitoring device from 1 day to 4 days, it was run continuously for 60 days and compared with the control data collected at a 1-day interval. The results showed that the consistency of the prevention threshold trigger time reached 97.2%, the daily average data upload volume of the device decreased by 75.0%, and energy consumption decreased by approximately 68%.

[0076] Example 3: Comparison of Optimal Interval Periods under Different Decision-Making Needs For the tea green leafhopper, weight coefficients were set for three different decision-making needs, and the optimal interval was calculated (see Table 2 for specific results): Table 2 Decision-making needs <![CDATA[w1]]> <![CDATA[w2]]> <![CDATA[w3]]> Optimal interval (days) Prevention threshold triggering 0.2 0.6 0.2 4 Population dynamics research 0.4 0.4 0.2 7 Short-term forecast and early warning 0.5 0.3 0.2 10 The results show that when the decision-making need is to trigger the prevention and control threshold, trend fidelity is the most important, and the optimal interval is 4 days. When the decision-making need is to study population dynamics, the interval can be appropriately relaxed to 7 days to save resources. When the decision-making need is to predict and warn for the short term, higher frequency data is required (a 10-day interval is sufficient because the daily data fluctuates greatly).

[0077] Example 4: Dynamic Update Verification: During tea garden monitoring, after accumulating new data for every 30 days, steps S1 to S5 were re-executed to optimize the interval. The dynamic update results of the tea green leafhopper are shown in Table 3 (details are shown in Table 3): Table 3 Time period Population characteristics Optimal interval (days) April to early May Early stage of population growth 7 Mid-May to early June Population peak 3 Mid-June to late July Population stabilization period 14 Early August to early September Population trough 21 Mid-September to early October Second peak of population 4 During dynamic updates, if insufficient sampling points are available due to device offlineness or communication interruption within a certain update cycle, causing the prevention and control event coverage C(t) corresponding to some candidate monitoring intervals to fall below the preset event coverage threshold, or the prevention and control response delay penalty P(t) to exceed the preset allowable response delay value, the system generates an insufficient event coverage flag or an excessive response delay flag, and determines the smaller of the preset minimum sampling interval or the previous round's effective optimal monitoring interval as the safe monitoring interval. In the subsequent two consecutive update cycles, if C(t) recovers to above the preset event coverage recovery threshold, and P(t) is not higher than the preset response delay recovery threshold, the system exits the safe rollback state and resumes the selection of the optimal monitoring interval based on the set of qualified candidate monitoring intervals.

Claims

1. A method for determining the optimal interval period for intelligent pest monitoring, characterized in that, Includes the following steps: S1. Collect target pest trapping and counting data to form the original monitoring sequence. And construct a set of prevention and control response event segments based on prevention and control thresholds; S2. Generate a set of candidate monitoring intervals T based on the minimum sampling interval, the maximum monitoring interval, the allowable time limit for prevention and control response, and the original set of prevention and control response event segments; S3. Divide the resampling window for each candidate interval t, and then calculate and correspond to obtain the resampling monitoring sequence X(t); S4. First, form the interval reduction sequence. Then, the information loss degree, trend fidelity, resource consumption degree, prevention and control event coverage, prevention and control response delay penalty degree, and phase transition retention degree are calculated. S5. First, obtain the set of qualified candidate monitoring intervals, then obtain the comprehensive monitoring quality value Q(t), and select the optimal monitoring interval. ; S6. Optimal monitoring interval The data is output to an intelligent insect monitoring device, and the optimal monitoring interval is updated when new trap count data meets the update criteria. ; In step S4, based on the original monitoring sequence Resampled monitoring sequence X(t), interval restored sequence And the original prevention and control response event segment set, respectively calculate the information loss degree L(t), trend fidelity F(t), resource consumption degree R(t), prevention and control event coverage C(t), prevention and control response delay penalty degree P(t) and stage transition retention degree G(t) corresponding to the candidate monitoring interval t; The prevention and control event coverage rate C(t) is determined by the ratio between the number of original prevention and control response event segments successfully covered by the resampled monitoring sequence X(t) and the total number of original prevention and control response event segments; The penalty for the delay in the prevention and control response is determined according to the delay time of the resampling response time in the resampling monitoring sequence X(t) that matches the original prevention and control response event segment relative to the trigger start time of the original prevention and control response event segment. When the resampling response time is later than the trigger start time of the original prevention and control response event segment, the first penalty coefficient is used, and when the resampling response time is earlier than or equal to the trigger start time of the original prevention and control response event segment, the second penalty coefficient is used. The first penalty coefficient is greater than the second penalty coefficient. In step S4, the resampled monitoring sequence X(t) corresponding to each candidate monitoring interval t is mapped back to the original monitoring sequence according to its representative sampling time. The timeline forms a sequence with the original monitoring sequence. Time-aligned interval restoration sequence Based on the original monitoring sequence Resampled monitoring sequence X(t), interval restored sequence And the original prevention and control response event segment set, respectively calculate the information loss degree L(t), trend fidelity F(t), resource consumption degree R(t), prevention and control event coverage C(t), prevention and control response delay penalty degree P(t) and stage transition retention degree G(t) corresponding to the candidate monitoring interval t; Information loss L(t) according to the original monitoring sequence The normalized difference between the information entropy of the resampled monitoring sequence X(t) and the information entropy of the resampled monitoring sequence is determined; Trend fidelity F(t) according to the original monitoring sequence Restored sequence with interval The weighted result of the Pearson correlation coefficient and the normalized dynamic time-normalized distance between them was determined; The stage transition retention G(t) is restored according to the interval sequence. Identified population stage transition points and original monitoring sequences The matching ratio between the identified population stage transition points is determined; The resource consumption R(t) is determined by weighting at least two of the following: the proportion of sampling times, the proportion of data upload volume, the proportion of storage usage, and the proportion of device wake-up energy consumption. The information loss L(t) is used to characterize the resampled monitoring sequence X(t) relative to the original monitoring sequence. The degree of information entropy loss; trend fidelity F(t) is used to characterize the reconstructed sequence at intervals. For the original monitoring sequence The degree to which the population change trend, peak position and fluctuation pattern of pests are maintained; the resource consumption degree R(t) is used to characterize at least two of the resource consumption degree of intelligent pest monitoring equipment under the candidate monitoring interval t: number of data collections, data upload volume, storage occupation and device wake-up energy consumption; the control event coverage rate C(t) is used to characterize the proportion of the resampled monitoring sequence X(t) that successfully covers the original control response event segment; The delay penalty P(t) for prevention and control response is used to characterize the degree of delayed response of the resampled monitoring sequence X(t) relative to the original prevention and control response event segment; the stage transition preservation G(t) is used to characterize the interval-restored sequence. The degree to which the target pest population maintains its transition points between the rising, peak, stable, trough, or secondary peak phases.

2. The method for determining the optimal interval period for intelligent pest monitoring according to claim 1, characterized in that: Intelligent insect monitoring equipment collects continuous trapping and counting data of target pests at preset minimum sampling intervals to form an original monitoring sequence. Based on the control threshold corresponding to the target pest, in the original monitoring sequence The original prevention and control response event segment set is constructed. Each original prevention and control response event segment includes the trigger start time, segment peak time, segment release time, segment maximum number of traps, and segment event number. Intelligent insect monitoring equipment includes color-attracting intelligent monitoring equipment, sex-attracting intelligent monitoring equipment, or light-attracting intelligent monitoring equipment; The color-attracting intelligent monitoring device includes a sticky insect board, a high-definition camera, and an image recognition module. The sticky insect board is used to attract target pests that are attracted to color, the high-definition camera is used to collect images of the sticky insect board, and the image recognition module is used to identify the target pests in the sticky insect board images and output the number of pests captured at the corresponding sampling time. The sex-attracting intelligent monitoring device includes a sex pheromone lure, a trap, and an automatic counting device. The sex pheromone lure and trap are used to attract target pests that are attracted to sex pheromones, and the automatic counting device is used to output the number of pests captured at the corresponding sampling time. The light-induced intelligent monitoring equipment is used to attract target pests based on light sources and output the number of pests captured at the corresponding sampling time.

3. The method for determining the optimal interval period for intelligent pest monitoring according to claim 2, characterized in that: In step S1, the original monitoring sequence This includes the number of pests trapped at each sampling time. In forming the original monitoring sequence This also includes the original monitoring sequence. Preprocessing is performed, including: The number of pests trapped was sorted chronologically according to the sampling time. Invalid sampling points generated during device offline, image recognition failure, automatic counting anomalies, or during trap replacement are removed. Missing sampling points are marked as missing or interpolation is performed based on adjacent valid sampling points; Single-point outliers exceeding a preset outlier multiple are marked as outliers; And based on the preprocessed original monitoring sequences Construct a set of original prevention and control response event segments.

4. The method for determining the optimal interval period for intelligent pest monitoring according to claim 3, characterized in that: The set of original prevention and control response event segments includes at least one original prevention and control response event segment; The construction of the original prevention and control response event segment set includes: When the original monitoring sequence When the number of pests trapped changes from below the control threshold corresponding to the target pest to reaching or exceeding the control threshold, the sampling time is determined as the trigger start time. When the number of pests trapped changes from reaching or exceeding the control threshold to falling below the control threshold and remaining below the control threshold for a preset release duration, the corresponding sampling time is determined as the segment release time. Between the trigger start time and the section release time, the sampling time with the largest number of pests trapped is determined as the section peak time; In step S2, each candidate monitoring interval t in the candidate monitoring interval set T satisfies the event coverage constraint for the original prevention and control response event segment; Each candidate monitoring interval t in the candidate monitoring interval set T is not less than the preset minimum sampling interval and not greater than the maximum allowed monitoring interval, and the candidate monitoring interval t satisfies the event coverage constraint for each original prevention and control response event segment in the original prevention and control response event segment set.

5. The method for determining the optimal interval period for intelligent pest monitoring according to claim 4, characterized in that: For each candidate monitoring interval t, the original monitoring sequence The system is divided into multiple time-continuous resampling windows. For each resampling window, a window fluctuation discrimination vector is calculated, consisting of the window growth rate, window coefficient of variation, window peak-to-valley difference, and proximity of the control threshold. A fluctuation enhancement label for the corresponding resampling window is generated based on the window fluctuation discrimination vector. When the resampling window overlaps with the original control response event segment in time, the corresponding number of pests trapped and the sampling time are retained. When the resampling window does not overlap with the original control response event segment, a representative number of pests trapped and a representative sampling time are generated based on the fluctuation state of the resampling window to obtain the resampling monitoring sequence X(t). In step S3, the proximity of the prevention and control threshold is determined based on the ratio between the maximum number of pests trapped in the corresponding resampling window and the prevention and control threshold of the target pest. When the proximity of the prevention and control threshold reaches the preset threshold proximity coefficient, it is determined that there are sampling points close to the prevention and control threshold within the corresponding resampling window; When the corresponding resampling window overlaps with multiple key sampling times within the same original prevention and control response event segment, the corresponding number of pests trapped and the sampling time are retained according to the priority of trigger start time, segment peak time, and segment release time. When the corresponding resampling window does not overlap with any original control response event segment in time, and the fluctuation enhancement marker indicates that the resampling window is in a state of fluctuation enhancement, the maximum number of pests trapped in the resampling window and its corresponding sampling time are retained. When the corresponding resampling window does not overlap with any original control response event segment in time, and the fluctuation enhancement marker indicates that the resampling window is in a stable state, the number of pests trapped in the resampling window is uniformly aggregated to generate the representative number of trapped pests and the representative sampling time of the resampling window.

6. The method for determining the optimal interval period for intelligent pest monitoring according to claim 5, characterized in that: In step S3, the window growth rate is determined based on the change between the number of terminal pests trapped and the number of initial pests trapped within the corresponding resampling window; The window coefficient of variation is determined by the ratio between the standard deviation and the mean of the number of pests trapped in the corresponding resampling window. The peak-to-valley difference in the window is determined based on the difference between the maximum and minimum number of pests trapped within the corresponding resampling window; When at least two of the window growth rate, window variation coefficient, window peak-to-valley difference, and prevention threshold proximity meet the corresponding preset threshold, a fluctuation enhancement label is generated to characterize that the resampling window is in a fluctuation enhancement state.

7. The method for determining the optimal interval period for intelligent pest monitoring according to claim 1, characterized in that: In step S5, the candidate monitoring interval set T is first screened according to the preset trend fidelity threshold, preset event coverage threshold, preset response delay allowable value and preset stage transition retention threshold to obtain a qualified candidate monitoring interval set. The edge resource budget status of the intelligent insect monitoring device is obtained, which includes at least two of the following: remaining power, available storage capacity, and communication status. Based on the trapping method of the target pest, the stage of population occurrence, the control threshold, the monitoring application requirements, and the edge resource budget status, the weight coefficients corresponding to the information loss degree L(t), trend fidelity F(t), resource consumption degree R(t), control event coverage C(t), control response delay penalty P(t), and stage transition retention degree G(t) are determined respectively. From the candidate monitoring interval set T, remove those with trend fidelity F(t) lower than the preset trend fidelity threshold, prevention event coverage C(t) lower than the preset event coverage threshold, prevention response delay penalty P(t) higher than the preset response delay allowable value, or stage transition retention G(t) lower than the preset stage transition retention threshold to obtain the qualified candidate monitoring interval set; if the qualified candidate monitoring interval set is not empty, then under the condition of satisfying the resource constraints corresponding to the edge resource budget state, perform a comprehensive evaluation on each candidate monitoring interval t in the qualified candidate monitoring interval set to obtain the comprehensive monitoring quality value Q(t); In step S6, the optimal monitoring interval is determined. The output is sent to the intelligent pest monitoring device, enabling the device to monitor according to the optimal monitoring interval. Adjust the collection, counting, identification, uploading, or cache cleanup cycle; when the newly added trap count data reaches the preset update condition, merge the newly added trap count data into the original monitoring sequence. Then, steps S2 to S5 are repeated to update the optimal monitoring interval. ; The intelligent pest monitoring device adjusts at least two of the following: camera wake-up cycle, image recognition module operation cycle, automatic counting cycle, data upload cycle, and local cache clearing cycle, in order to perform subsequent pest monitoring; When the newly added trap count data reaches the preset update data volume or preset update cycle, the newly added trap count data will be incorporated into the original monitoring sequence. ; The newly added trap count data will be incorporated into the original monitoring sequence. Then, based on the updated original monitoring sequence Calculate the population growth rate, peak and trough variation characteristics, fluctuation intensity, and stage identification confidence of the target pest to identify the current population development stage of the target pest; When the target pest is in the rising, peak or secondary peak period of the population, and the confidence level of the stage identification reaches the preset confidence threshold, the upper limit of the candidate monitoring interval value in the candidate monitoring interval set T is reduced, and the weights corresponding to the control event coverage C(t), the control response delay penalty P(t) and the trend fidelity F(t) are increased. When the target pest is in a stable or low period and the confidence level of the stage identification reaches the preset confidence threshold, the upper limit of the candidate monitoring interval value in the candidate monitoring interval set T is expanded, and the weight corresponding to the resource consumption degree R(t) is increased. When the confidence level of the stage identification is lower than the preset confidence threshold, maintain the current execution monitoring interval or roll back the current execution monitoring interval to a shorter safe monitoring interval; and re-execute steps S2 to S5. After obtaining a new optimal monitoring interval, if the population occurrence stages identified in two consecutive update cycles are consistent, or if the change in the new optimal monitoring interval relative to the current execution monitoring interval reaches a preset switching threshold, then the new optimal monitoring interval will be updated to the device's execution monitoring interval; otherwise, the current execution monitoring interval will be maintained.

8. The method for determining the optimal interval period for intelligent pest monitoring according to claim 7, characterized in that: In step S5, if the set of qualified candidate monitoring intervals is empty, the smaller of the preset minimum sampling interval or the previous round's effective optimal monitoring interval is determined as the safe monitoring interval, and this safe monitoring interval is used as the optimal monitoring interval for this round. At the same time, a rollback reason flag is generated; In steps S5 and S6, after the rollback reason marker is generated, the monitoring quality indicators corresponding to the rollback reason marker are checked first during the next dynamic update. When the monitoring quality index corresponding to the rollback reason marker continuously reaches the corresponding preset recovery threshold, the safe rollback state is exited, and the optimal monitoring interval selection based on the qualified candidate monitoring interval set is restored. When the monitoring quality indicator corresponding to the rollback reason marker does not reach the corresponding preset recovery threshold, the safe monitoring interval will continue to be maintained.

Images