A method for identifying migration events of migratory pests based on pest time-series dynamic background
Patent Information
- Application Number
- CN202611158345.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-31
- Publication Date
- 2026-10-30
AI Technical Summary
[0006]本发明的目的在于提供一种基于虫情时序动态背景的迁飞性害虫迁入事件识别方法，解决现有技术中固定阈值适应性差、人工判读效率低、连续高虫量重复识别、缺测干扰以及事件结果难以直接服务虫源回推的问题
[0042]本发明具有以下有益效果：第一，本发明使用站点自身短期滚动中位数作为动态背景，能够适应不同站点和不同时段的虫情背景差异；第二，本发明同时设置相对突增、绝对突增和低值后重启触发，能够兼顾高本底站点、低本底站点以及近零后突然回升的迁入信号；第三，本发明通过连续触发去冗余避免同一轮高虫量过程被重复识别为多个事件；第四，本发明通过峰值支持阈值将单日触发信号转化为事件窗口，使输出结果更接近可用于回推和分析的迁入候选期；第五，本发明将事件强度分级和回推任务导出结合，能够直接服务大范围迁飞轨迹回推、虫源分析和测报复核。
Smart Images
Figure CN123012491A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of agricultural pest monitoring and early warning and pest time series analysis technology, and in particular to a method for identifying migration events of migratory pests based on the dynamic background of pest time series. It can be used for processing light-attracted data of agricultural pests with migratory behavior such as white-backed planthopper, brown planthopper, rice leaf roller, and fall armyworm, automatic identification of migration events, and generation of subsequent pest source trajectory backtracking tasks. Background Technology
[0002] Migratory pests can migrate long distances using monsoons, low-level jet streams, and weather processes, causing sudden population increases in their destination areas and posing a serious threat to the production of crops such as rice and corn. Light trapping is an important tool in my country's major crop pest monitoring system. It can record the daily number of insects trapped at different monitoring stations continuously over a long period, providing a crucial data foundation for identifying migration processes, analyzing pest source relationships, and developing regional early warning systems.
[0003] Current methods for analyzing insect populations using light traps largely rely on fixed insect population thresholds, manual interpretation based on experience, or traditional peak period division rules. For example, in rice planthopper monitoring, a peak period is often defined as the day from the sudden increase in adult population to the day of the sudden decrease after the peak, with the day with the highest insect population within the peak period being the peak day. While these methods are practical for manual verification at a single site, they still have significant shortcomings when processing large batches over multiple years and at multiple sites: First, the baseline insect population varies greatly between different sites, making it difficult to adapt a fixed threshold to sites with high and low baseline populations; Second, the background insect population at the same site fluctuates significantly in different years and seasons, and simply using historical averages or fixed thresholds can easily lead to missed or incorrect assessments; Third, the original light trap data often contains multiple records from the same site on the same day, broken lights, missing measurements, non-standard notes, and missing coordinates, directly affecting the stability of event identification; Fourth, during a continuous period of high insect populations, multiple days may meet the high-value criteria, and judging each day individually can easily lead to repeated splitting of the same migration process; Fifth, traditional peak period rules usually do not directly provide event priorities and standardized task tables that can be used for meteorological trajectory backtracking.
[0004] Furthermore, traditional manual delineation of peak periods typically only provides empirical descriptions of sudden increase days, peak days, and peak period ranges, without further specifying how to maintain consistency of event boundaries in situations such as missing data, continuous triggering, and overlapping of adjacent events. Therefore, a more complete calculation process is still needed to automatically convert the light-attracting day sequence into a standard event table that can be batch-verified and directly drive insect source backtracking.
[0005] Therefore, there is an urgent need for an automated method for identifying candidate migration events based on multi-year, multi-site light-attracted insect infestation data. This method should be able to identify sudden increase signals by utilizing the short-term insect infestation background of each site, take into account both low-background and high-background sites, handle issues such as continuous triggering, missing data, event extension, merging of adjacent events, and hierarchical output, and provide a consistent data interface for subsequent migration trajectory retrospection and insect source analysis. Summary of the Invention
[0006] The purpose of this invention is to provide a method for identifying migratory pest migration events based on the dynamic background of insect infestation time, which solves the problems of poor adaptability of fixed thresholds, low efficiency of manual interpretation, repeated identification of continuous high insect populations, interference from missing data, and difficulty in directly using event results to backtrack the source of insects in the existing technology.
[0007] The technical solution to achieve the above-mentioned objective of this invention is: a method for identifying migratory pest migration events based on the dynamic background of insect infestation time, comprising:
[0008] Step 1: Construct a daily sequence for the site;
[0009] The system reads raw light-attracting data from multiple years and sites. This raw data includes at least the following fields: date, province, site, longitude, latitude, pest species, number of attracted pests, weather conditions at the time of lighting, and remarks. The date field is converted to a date format, the text field is normalized and cleaned for null values, and the longitude, latitude, and number of attracted pests fields are numerically processed. Records of target migratory pests are then filtered according to the target pest species. Daily-scale aggregation is performed on the target pest records from the same date, province, and site to obtain a daily pest count sequence for each site.
[0010] Step 2: Identify the day that triggers the insect population;
[0011] For each effective observation day at a station, a dynamic background insect population is constructed using the median insect population of the previous N days. Let Wt be the set of effective insect populations of station s within the N days prior to day t.
[0012] Step 3: Construct the immigration candidate event window;
[0013] The sudden increase trigger and the low value restart trigger are combined into a total trigger, and redundancy is removed from the continuous trigger segments, retaining only the first day of the continuous trigger segment as the migration candidate day; let the sudden increase trigger on day t be marked as St, and the low value restart trigger be marked as Rt, then the total trigger mark Gt and the candidate day mark Ht are respectively:
[0014] Gt = St∨Rt;
[0015] Ht=Gt∧¬Gt-1;
[0016] Where ∨ represents the logical "OR"; ∧ represents the logical "AND". When Ht is true, day t is considered a candidate day for migration. When both Gt and Gt-1 are true, day t is a continuation of the continuous triggering segment and is no longer considered a new candidate day for migration. Adjacent candidate days are merged into an initial event according to the maximum allowed interval between candidate days. Let two adjacent candidate days be ta and tb, and the maximum allowed interval be G. If tb-ta≤G+1, then they are assigned to the same triggering group Q. For the same triggering group Q, the initial start date Ts, the initial end date Te, the triggering peak insect quantity P, and the peak support threshold T are respectively:
[0017] Ts = min(Q);
[0018] Te = max(Q);
[0019] P = max(Ct), t ∈ Q;
[0020] T = max(Cmin, δ·P);
[0021] Where Cmin is the minimum daily insect quantity threshold; δ is the peak support ratio; P is the peak insect quantity within the trigger group; δ·P represents the supported insect quantity obtained by multiplying the trigger peak insect quantity by the peak support ratio. Starting from Te, checks are performed daily. If the insect quantity on the subsequent Te+j day is not empty and C(Te+j)≥T, and j does not exceed the maximum extension days E, then that day is included in the event window; if the insect quantity on subsequent days is empty, the insect quantity is lower than T, or the extension days reach E, then the extension is stopped.
[0022] Step 4: Merge events and output pushback tasks;
[0023] Event windows that overlap or are adjacent after extension are merged again, and the event start date, end date, peak date, total number of insects, peak number of insects, number of candidate days, number of missing days, and trigger type are recalculated. Let the two extended event windows be [Sa,Ea] and [Sb,Eb], and Sa≤Sb; Sa,Ea,Sb,Eb represent the start and end points of the corresponding time windows, respectively. If Sb≤Ea+1, then the two event windows are merged into:
[0024] [S,E]=[Sa,max(Ea,Eb)];
[0025] For the merged event window [S,E], the total number of insects Csum, the peak number of insects Cpeak, and the peak date Tpeak are respectively:
[0026] Csum = ΣCt, t ∈ [S, E] and Ct is a valid value;
[0027] Cpeak = max(Ct), t ∈ [S, E];
[0028] Tpeak=argmax(Ct),t∈[S,E];
[0029] Events are prioritized based on the peak insect count (Cpeak) and total insect count (Csum). If Cpeak ≥ PA or Csum ≥ TA, the event is classified as a Level A priority event for backtracking. If Cpeak ≥ PB and Csum ≥ TB, it is classified as a Level B event that can be backtracked. If Cpeak ≥ PC and Csum ≥ TC, it is classified as a Level C event requiring manual review. All others are classified as Level D events where backtracking is not recommended. Output fields include at least the year, event start date, event end date, peak date, total insect count, peak insect count, number of candidate days, number of missing days, trigger type, province, station, longitude, latitude, review level, and backtracking suggestion. Based on this, a migration candidate event table and a trajectory backtracking task table are generated.
[0030] Furthermore, step 1 can be specifically expressed as follows:
[0031] Specifically: Ct = Σxi, i ∈ R(s,t);
[0032] Where Ct is the daily insect count at station s on date t, R(s,t) is the set of target pest records at station s on date t, xi is the number of target pests attracted in the i-th record, and Σ is the summation; if the number of insects attracted at the same station on the same date is all empty, then Ct is recorded as an empty value; if the remarks, weather, or quantity fields indicate that there are broken lights, missing measurements, or invalid observations on that day, then a missing or invalid observation mark is set. For records with missing latitude and longitude, the valid latitude and longitude from other records at the same station are used to fill in the missing latitude and longitude first. If the missing latitude and longitude cannot be filled in, the coordinate table of external stations is used to fill in the missing latitude and longitude; and a continuous date set Ds={tmin,tmin+1,...,tmax} is constructed for each station. Dates that do not exist in the original data are retained as empty values and marked as missing days.
[0033] Furthermore, the specific method for step 2 is as follows: Wt = {Ct-1, Ct-2, ..., Ct-N | Ct-k is the effective observation value};
[0034] When the number of valid observations in Wt is not less than M, the dynamic background insect population Bt is defined as:
[0035] Bt = median(Wt);
[0036] Where median(·) represents the median function; median(Wt) represents taking the median of the effective insect count in set Wt. When the number of effective observations in Wt is less than M, Bt is recorded as a null value or not included in the relative surge determination dependent on the background; surge triggering and low-value restart triggering are determined based on Bt respectively; the determination condition for surge triggering is:
[0037] Ct≥Cmin, and [(Bt>0 and Ct≥α·Bt) or (Ct-max(Bt,0)≥β)];
[0038] Where Cmin is the minimum daily insect population threshold, α is the relative dynamic background surge factor, and β is the absolute increment threshold; if day t is a valid observation day and meets the above conditions, then day t is marked as the surge trigger day; the judgment condition for restarting the trigger after a low value is:
[0039] Ct-1≤L, Ct-2≤L, Ct≥Cmin, and [Ct≥Cr or Ct≥max(Bt,0)+γ];
[0040] Where L is the low-value threshold, Cr is the restart insect population threshold, and γ is the relative background restart increment. If day t is a valid observation day and meets the above conditions, then day t is marked as the low-value restart trigger day.
[0041] Furthermore, N is set to 7, M to 3, minimum daily insect quantity threshold Cmin to 10, relative dynamic background surge multiple α to 1.5, and absolute increment threshold β to 5; in the low-value restart trigger, the low-value threshold L for the previous two days is set to 1, the insect quantity threshold Cr for the current day restart is set to 12, and the relative background restart increment γ to 3; the allowed interval G between candidate days is set to 1 day, that is, at most one non-candidate day is allowed between two candidate days; the peak support ratio δ is set to 0.4; the maximum number of days the event is extended backward is set to 3; the A-level threshold PA is set to 50, TA to 200, the B-level threshold PB to 20, TB to 50, and the C-level threshold PC to 15, TC to 30.
[0042] This invention has the following beneficial effects: First, it uses the short-term rolling median of the site itself as a dynamic background, which can adapt to the differences in insect infestation backgrounds at different sites and at different times; Second, it simultaneously sets relative surges, absolute surges, and restart triggers after low values, which can take into account high-background sites, low-background sites, and migration signals that suddenly rebound after near zero; Third, it avoids the same high insect population process being repeatedly identified as multiple events by continuous triggering to remove redundancy; Fourth, it transforms the daily trigger signal into an event window through a peak support threshold, making the output results closer to the migration candidate period that can be used for backtracking and analysis; Fifth, it combines event intensity grading and backtracking task export, which can directly serve large-scale migration trajectory backtracking, insect source analysis, and measurement return verification. Attached Figure Description
[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art are briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0044] Figure 1 The flowchart of a method for identifying migratory pest migration events based on the dynamic background of insect infestation time provided by the present invention is shown.
[0045] Figure 2 This is a schematic diagram illustrating candidate day triggering, continuous triggering for redundancy removal, and event window extension in an embodiment of the present invention. Detailed Implementation
[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be noted that the described embodiments are only some embodiments of the present invention, not all embodiments, and are only used to illustrate the present invention, and are not intended to limit the scope of protection of the present invention. The scope of protection of this application should be determined by the claims.
[0047] Figure 1 The flowchart illustrates a method for identifying migratory pest migration events based on the dynamic background of insect infestation timelines, as provided by this invention. Figure 1 As shown, the method generally includes insect data preprocessing S1 and migration event identification S2. Insect data preprocessing S1 is used to convert multi-source, discontinuous, and possibly missing light-attracted raw records into a continuous daily insect population sequence and dynamic background at the site scale; migration event identification S2 is used to identify trigger signals based on the dynamic background and convert the trigger signals into outputtable and retrospective migration candidate event windows.
[0048] S11, Read the raw data from the light trap and filter the target species. The raw data includes fields such as date, province, station, longitude, latitude, pest species, quantity, weather at the time of lighting, and remarks. Text fields are standardized by removing spaces and null values, date fields are converted to date format, and latitude / longitude and quantity fields are numerically processed. Records of target migratory pests are then filtered. In this embodiment, the target migratory pest is the white-backed planthopper.
[0049] S12, Complete Station Coordinates. For records with missing latitude and longitude, prioritize using valid latitude and longitude coordinates of the same station that have appeared in the original light-tracking data to complete the coordinates; if they are still missing, use an external station coordinate table to complete them. This step reduces the impact of missing coordinates for local years or local records on the generation of subsequent trajectory backtracking tasks.
[0050] S13 performs daily-scale aggregation of target pest records for the same site on the same day and marks missing measurements. Specifically, using date, province, and site as grouping fields, multiple pest quantity records for the same day are summed. If all quantities in the same group are empty, they are left empty. The first valid value for latitude and longitude is used. The first non-empty text in the weather field is used. The remarks field is deduplicated and then concatenated. If any record has a light malfunction, missing quantity, or invalid observation mark, that day is designated as a risk day for light malfunction, missing measurement, or invalid observation.
[0051] S14, Construct a continuous daily series and calculate the dynamic background. For each station, construct a continuous daily index from the earliest observation date to the latest observation date. Dates not present in the original data are left blank and marked as missing based on insect population status. Let the insect population at a station on day t be Ct. The dynamic background insect population Bt is defined as the median of the insect populations over the N days prior to day t, with the number of valid observations within the background window not less than M, i.e., Bt = median(Ct-1, Ct-2, ..., Ct-N). In this embodiment, N = 7, M = 3. Using the median instead of the mean can reduce the impact of short-term abnormally high values on background estimation.
[0052] S21, trigger judgment is performed. Trigger judgment includes sudden increase trigger and low value restart trigger. If day t is a valid observation day, and Ct is not less than the minimum daily insect quantity threshold Cmin, and meets any of the following conditions, then day t is marked as sudden increase trigger: when Bt is greater than 0, Ct is not less than α·Bt; or Ct-max(Bt,0) is not less than β. In this embodiment, Cmin=10, α=1.5, β=5. If day t is a valid observation day, and the insect quantity of the previous two days is not greater than the low value threshold L, Ct is not less than Cmin, and meets any of the conditions that Ct is not less than the restart insect quantity threshold Cr or Ct is not less than max(Bt,0)+γ, then day t is marked as low value restart trigger. In this embodiment, L=1, Cr=12, γ=3.
[0053] S22, perform continuous trigger redundancy removal. Perform a logical OR operation on the sudden increase trigger and the low-value restart trigger to obtain the total trigger `raw_trigger`. If a total trigger exists on day t and not on day t-1, then day t is marked as a migration candidate day; if a total trigger exists on both day t and day t-1, then day t is not considered a new migration candidate day. Through this step, when the triggering conditions are met for multiple consecutive days, only the first day of the continuous trigger segment is retained, thus avoiding repeated identification of the same round of high insect population migration.
[0054] S23, perform candidate day merging and event window extension. Scan incoming candidate days in chronological order. If the interval between adjacent candidate days does not exceed the maximum allowed interval, they are grouped into the same trigger group. In this embodiment, the maximum allowed interval is 1 day, meaning that at most one non-candidate day is allowed between two candidate days. For each trigger group, the earliest candidate day is taken as the initial start date, and the latest candidate day is taken as the initial end date. The peak insect count P within the trigger group is calculated. The peak support threshold T = max(Cmin, δ·P) is further calculated, where δ is the peak support ratio. In this embodiment, δ = 0.4. Check each day from the initial end date. If the insect count on subsequent days is not empty and is not lower than T, then that day is merged into the event window; if the insect count is empty, lower than T, or the extension days reach the maximum extension days E, then the extension stops. In this embodiment, E = 3.
[0055] Following S23, the extended event windows are merged again. If the start date of the later event is no later than one day after the end date of the earlier event, the two events are merged into one event window. After merging, the peak date, peak insect count, total insect count, candidate days, missing days, and trigger type within the event window are recalculated. Figure 2 The process of dynamic background calculation, trigger judgment, continuous triggering to remove redundancy, candidate day merging, and window extension and merging is shown.
[0056] After completing the event window construction, the event metrics are output and categorized. Each migration candidate event outputs the year, start date, end date, peak date, total number of bugs, peak number of bugs, number of candidate days, number of missing days, trigger type, province, station, longitude, and latitude. Further, events are classified into four levels (A, B, C, and D) based on the peak number of bugs and the total number of bugs. In one embodiment, if the peak number of bugs is not less than 50 or the total number of bugs is not less than 200, it is classified as A (Priority Backtracking); if the peak number of bugs is not less than 20 and the total number of bugs is not less than 50, it is classified as B (Backtracking Possible); if the peak number of bugs is not less than 15 and the total number of bugs is not less than 30, it is classified as C (Manual Review); the rest are classified as D (Not Recommended). Level A and B events can be exported as trajectory backtracking tasks, while level C events can be submitted for manual review to determine whether to backtrack.
[0057] It should be noted that the output of this invention is an influx event window obtained based on the identification of sudden increases in the number of insects attracted by light traps. The start date, end date, and peak date are event indicators constructed by the algorithm and should not be interpreted as the unique and definite boundaries of the entire actual biological migration process. For those skilled in the art, N, M, Cmin, α, β, L, Cr, γ, δ, E, and the grading threshold can be adjusted according to the target pest species, regional monitoring habits, and data quality. All equivalent substitutions, parameter adjustments, or process combination improvements made under the technical concept of this invention should be included within the protection scope of this invention.
Claims
1. A method for identifying migratory pest migration events based on the dynamic background of insect infestation time, comprising: Step 1: Construct a daily sequence for the site; The system reads raw light-attracting data from multiple years and sites. This raw data includes at least the following fields: date, province, site, longitude, latitude, pest species, number of attracted pests, weather conditions at the time of lighting, and remarks. The date field is converted to a date format, the text field is normalized and cleaned for null values, and the longitude, latitude, and number of attracted pests fields are numerically processed. Records of target migratory pests are then filtered according to the target pest species. Daily-scale aggregation is performed on the target pest records from the same date, province, and site to obtain a daily pest count sequence for each site. Step 2: Identify the day that triggers the insect population; For each effective observation day at a station, a dynamic background insect population is constructed using the median insect population of the previous N days. Let Wt be the set of effective insect populations of station s within the N days prior to day t. Step 3: Construct the immigration candidate event window; The sudden increase trigger and the low value restart trigger are combined into a total trigger, and redundancy is removed from the continuous trigger segments, retaining only the first day of the continuous trigger segment as the migration candidate day; let the sudden increase trigger on day t be marked as St, and the low value restart trigger be marked as Rt, then the total trigger mark Gt and the candidate day mark Ht are respectively: Gt = St∨Rt; Ht=Gt∧¬Gt-1; Where ∨ represents the logical "OR"; ∧ represents the logical "AND". When Ht is true, day t is considered a candidate day for migration. When both Gt and Gt-1 are true, day t is a continuation of the continuous triggering segment and is no longer considered a new candidate day for migration. Adjacent candidate days are merged into an initial event according to the maximum allowed interval between candidate days. Let two adjacent candidate days be ta and tb, and the maximum allowed interval be G. If tb-ta≤G+1, then they are assigned to the same triggering group Q. For the same triggering group Q, the initial start date Ts, the initial end date Te, the triggering peak insect quantity P, and the peak support threshold T are respectively: Ts = min(Q); Te = max(Q); P = max(Ct), t ∈ Q; T = max(Cmin, δ·P); Where Cmin is the minimum daily insect quantity threshold; δ is the peak support ratio; P is the peak insect quantity within the trigger group; δ·P represents the supported insect quantity obtained by multiplying the trigger peak insect quantity by the peak support ratio. Starting from Te, checks are performed daily. If the insect quantity on the subsequent Te+j day is not empty and C(Te+j)≥T, and j does not exceed the maximum extension days E, then that day is included in the event window; if the insect quantity on subsequent days is empty, the insect quantity is lower than T, or the extension days reach E, then the extension is stopped. Step 4: Merge events and output pushback tasks; Event windows that overlap or are adjacent after extension are merged again, and the event start date, end date, peak date, total number of insects, peak number of insects, number of candidate days, number of missing days, and trigger type are recalculated. Let the two extended event windows be [Sa,Ea] and [Sb,Eb], and Sa≤Sb; Sa,Ea,Sb,Eb represent the start and end points of the corresponding time windows, respectively. If Sb≤Ea+1, then the two event windows are merged into: [S,E]=[Sa,max(Ea,Eb)]; For the merged event window [S,E], the total number of insects Csum, the peak number of insects Cpeak, and the peak date Tpeak are respectively: Csum = ΣCt, t ∈ [S, E] and Ct is a valid value; Cpeak = max(Ct), t ∈ [S, E]; Tpeak=argmax(Ct),t∈[S,E]; Events are prioritized based on the peak insect count (Cpeak) and total insect count (Csum). If Cpeak ≥ PA or Csum ≥ TA, the event is classified as a Level A priority event for backtracking. If Cpeak ≥ PB and Csum ≥ TB, it is classified as a Level B event that can be backtracked. If Cpeak ≥ PC and Csum ≥ TC, it is classified as a Level C event requiring manual review. All others are classified as Level D events where backtracking is not recommended. Output fields include at least the year, event start date, event end date, peak date, total insect count, peak insect count, number of candidate days, number of missing days, trigger type, province, station, longitude, latitude, review level, and backtracking suggestion. Based on this, a migration candidate event table and a trajectory backtracking task table are generated.
2. The method for identifying migratory pest migration events based on the dynamic background of insect infestation time as described in claim 1, characterized in that, The specific expression of step 1 is as follows: Specifically: Ct = Σxi, i ∈ R(s,t); Where Ct is the daily insect count at station s on date t, R(s,t) is the set of target pest records at station s on date t, xi is the number of target pests attracted in the i-th record, and Σ is the summation; if the number of insects attracted at the same station on the same date is all empty, then Ct is recorded as an empty value; if the remarks, weather, or quantity fields indicate that there are broken lights, missing measurements, or invalid observations on that day, then a missing or invalid observation mark is set. For records with missing latitude and longitude, the valid latitude and longitude from other records at the same station are used to fill in the missing latitude and longitude first. If the missing latitude and longitude cannot be filled in, the coordinate table of external stations is used to fill in the missing latitude and longitude; and a continuous date set Ds={tmin,tmin+1,...,tmax} is constructed for each station. Dates that do not exist in the original data are retained as empty values and marked as missing days.
3. The method for identifying migratory pest migration events based on the dynamic background of insect infestation time as described in claim 1, characterized in that, The specific method for step 2 is as follows: Wt={Ct-1,Ct-2,...,Ct-N|Ct-k is the effective observation value}; When the number of valid observations in Wt is not less than M, the dynamic background insect population Bt is defined as: Bt = median(Wt); Where median(·) represents the median function; median(Wt) represents taking the median of the effective insect count in set Wt. When the number of effective observations in Wt is less than M, Bt is recorded as a null value or not included in the relative surge determination dependent on the background; surge triggering and low-value restart triggering are determined based on Bt respectively; the determination condition for surge triggering is: Ct≥Cmin, and [(Bt>0 and Ct≥α·Bt) or (Ct-max(Bt,0)≥β)]; Where Cmin is the minimum daily insect population threshold, α is the relative dynamic background surge factor, and β is the absolute increment threshold; if day t is a valid observation day and meets the above conditions, then day t is marked as the surge trigger day; the judgment condition for restarting the trigger after a low value is: Ct-1≤L, Ct-2≤L, Ct≥Cmin, and [Ct≥Cr or Ct≥max(Bt,0)+γ]; Where L is the low-value threshold, Cr is the restart insect population threshold, and γ is the relative background restart increment. If day t is a valid observation day and meets the above conditions, then day t is marked as the low-value restart trigger day.
4. The method for identifying migratory pest migration events based on the dynamic background of insect infestation time as described in claim 3, characterized in that, N is set to 7, M to 3, minimum daily insect quantity threshold Cmin to 10, relative dynamic background surge multiple α to 1.5, absolute increment threshold β to 5; in low-value restart triggering, the low-value threshold L for the previous two days is set to 1, the insect quantity threshold Cr for restarting on the current day is set to 12, and the relative background restart increment γ to 3; the allowed interval G between candidate days is set to 1 day, that is, at most one non-candidate day is allowed between two candidate days; the peak support ratio δ is set to 0.4; the maximum number of days the event is extended backward is set to 3; the A-level threshold PA is set to 50, TA to 200, the B-level threshold PB to 20, TB to 50, and the C-level threshold PC to 15, TC to 30.