Pest ecological regulation auxiliary decision-making method and system based on multi-modal data fusion

Abstract

The invention relates to the technical field of pest control management, in particular to a pest ecological regulation auxiliary decision-making method and system based on multi-modal data fusion. The method comprises the following steps: firstly, acquiring multi-modal data of each partition, screening all reference time periods of each partition in a current time period, and acquiring the confidence coefficient of each monitoring means; and then acquiring a pest development similarity coefficient under each monitoring means between the current time period of each partition and each corresponding reference time period, and further acquiring a confidence time sequence of the quantity of pests in each reference time period and a preset future time period thereof, thereby predicting pest development in the region to be regulated and controlled and assisting a pest ecological regulation and control decision. According to the method, the reference time period for historical reference is preliminarily screened through the multi-modal data, and then the confidence time sequence is obtained through analysis and evaluation from the aspects of the environment and the pests, so that the accurate historical reference is provided, the pest development prediction accuracy is improved, and then the auxiliary decision-making effect of pest ecological regulation and control is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pest control management, and in particular to a pest ecological regulation decision-making auxiliary method and system based on multimodal data fusion. Background Art

[0002] In the face of the problem of pest resistance and the need for sustainable development, pest ecological control strategies are gradually being applied. Pest ecological control mainly predicts the number and spread of pests, and then adjusts and controls the biological and non-biological factors in the ecosystem to maintain the pest population at a low equilibrium density fluctuation. Therefore, before determining the decision on pest ecological control, it is crucial to accurately assess the population size and spread information of pests.

[0003] Currently, multimodal pest monitoring methods are mainly used to collect and monitor the number of pests in the ecological area to be regulated, and fusion analysis is used to predict the temporal changes in the pest population and its diffusion behavior, thereby assisting in making relevant ecological regulation decisions. However, the data formats of different modal pest monitoring methods are different, the fusion process is relatively complex, and it is easily affected by sensor interference, resulting in inaccurate pest population monitoring results. At the same time, in complex ecosystems, the behavior of pest populations is easily affected by the environment. For example, different environmental indicators have different effects on the migration of pests. The analysis of pest population changes and diffusion behavior based on multimodal pest monitoring methods may not be accurate enough, thereby affecting the auxiliary decision-making effect of pest ecological regulation. Summary of the Invention

[0004] In order to solve the technical problem that the analysis of pest population changes and diffusion behavior is not accurate enough, which leads to poor auxiliary decision-making effect for pest ecological control, the purpose of the present invention is to provide a method and system for auxiliary decision-making for pest ecological control based on multimodal data fusion. The technical solutions adopted are as follows: A pest ecological control decision-making support method based on multimodal data fusion, the method comprising: Obtain multimodal data for each sub-area within the area to be regulated at each monitoring moment within a preset historical period at the current moment, the multimodal data including environmental parameters under each environmental indicator and the number of pests under each monitoring method; In the preset historical period, the preset sub-period containing the current moment is used as the current period, and all historical periods of the same length as the current period are obtained. For each partition, the similarity of changes between the multimodal data corresponding to each partition in the current period and each other partition in each historical period is analyzed, and all reference periods for the current period are filtered out from all historical periods corresponding to each other partition. In each subarea, the deviations in pest populations under all monitoring methods at the same monitoring time are compared to obtain the confidence level of each monitoring method. Based on the temporal variation differences between environmental parameters under the same environmental indicators and the temporal variation differences between pest populations under the same monitoring method, the similarity coefficient of pest development under each monitoring method is obtained. Between the current period of each partition and each corresponding reference period in the other partitions, according to the similarity coefficient and confidence of pest development under each monitoring method, the time series of the number of pests under each monitoring method in each reference period and its preset future period is adjusted to obtain the confidence time series sequence; based on the confidence time series sequence of all reference periods and their preset future periods corresponding to the current period of each partition, the development of pests in the area to be regulated is predicted to assist in pest ecological regulation decision-making.

[0005] Furthermore, the method for obtaining the reference period includes: Take any environmental indicator or monitoring method as the target mode, any sub-area as the target sub-area, any non-target sub-area as the reference sub-area, and any historical period as the comparison period; Under the target mode, the difference between the corresponding modal data at the same monitoring time sequence number between the current period of the target partition and the comparison period of the reference partition is used to obtain the modal sub-deviation. The modal sub-deviation under all environmental indicators and monitoring methods is fused, and the negative correlation normalization result of the fusion result is used as the reference factor for the comparison period of the reference partition; Among all historical periods corresponding to all non-target partitions, all historical periods in which the reference factor is greater than a preset reference threshold are used as reference periods for the current period of the target partition.

[0006] Furthermore, the method for obtaining the confidence level includes: In each partition and at each monitoring moment, the reference pest number is obtained based on the concentrated characteristics of the pest number under all monitoring means, and the monitoring accuracy factor under each monitoring means is obtained based on the deviation of the pest number under each monitoring means relative to the reference pest number. The confidence level of each monitoring means is obtained based on the amplitude level and time series fluctuation of the monitoring accuracy factor within a preset historical period.

[0007] Furthermore, according to the amplitude level and time series fluctuation of the monitoring accuracy factor within a preset historical period, a method for obtaining the confidence level of each monitoring means includes: For each monitoring method, the normalized result of the extreme negative correlation of the monitoring accuracy factor in the preset historical period is used as the stability factor. The stability factor is used to weight the mean of the monitoring accuracy factor in the preset historical period to obtain the confidence level.

[0008] Furthermore, the method for obtaining the pest development similarity coefficient includes: Taking any subarea as the target subarea, under each environmental indicator and each monitoring method, obtain the environmental parameter sequence and pest quantity sequence of the target subarea in the current period, as well as the reference environmental parameter sequence and reference pest quantity sequence of the corresponding non-target subarea in each reference period of the current period; Under each environmental indicator, based on the difference between the environmental parameter sequence and each reference environmental parameter sequence, the environmental sub-deviation between the target partition in the current period and the non-target partition in each reference period is obtained. The environmental sub-deviation under all environmental indicators is combined to obtain the environmental deviation between the target partition in the current period and the non-target partition in each reference period. Under each monitoring method, based on the difference in pest numbers under the same sequence number between the pest number sequence and each reference pest number sequence, the pest change similarity parameters between the target partition in the current period and the non-target partition in each reference period are obtained; The pest change similarity parameter is weighted using the negative correlation mapping result of the environmental deviation to obtain the pest development similarity coefficient between the target partition in the current time period and the non-target partition in each reference time period.

[0009] Furthermore, the method for obtaining the environmental sub-deviation includes: Under each environmental indicator, the normalized result of the DTW distance between the environmental parameter sequence and each reference environmental parameter sequence is used as the environmental sub-deviation.

[0010] Furthermore, the method for obtaining the trusted time series includes: Taking any subarea as the target subarea, between the current time period of the target subarea and each corresponding reference time period in the non-target subarea, the pest development similarity coefficient is weighted using the confidence level under each monitoring method, and the weighted result is used as the confidence reference weight of each monitoring method in each corresponding reference time period in the non-target subarea; Under each monitoring method, the time series of the number of pests in each reference period of the current period and the corresponding non-target partition in the preset future period is used as the sequence to be adjusted; the confidence reference weight under each monitoring method is used to weightedly average the number of each pest in the corresponding sequence to be adjusted, and the weighted average results of all monitoring methods are used as sequence elements to construct a time series sequence to obtain a confidence time series sequence.

[0011] Furthermore, the method for predicting the development of pests in the area to be regulated includes: A training set of a preset prediction model is constructed based on all the confidence time series sequences, and the time series sequence of the number of pests in each partition in the current time period is used as the input of the trained preset prediction model. The time series sequence of the future number of pests in each partition in the preset future time period at the current moment is output, and based on the time series sequence of the future number of pests in each partition, the future spread rate of pests in each partition is obtained.

[0012] Furthermore, the preset prediction model is an ARIMA-SVM model.

[0013] A pest ecological control decision-making support system based on multimodal data fusion, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the pest ecological control decision-making support method based on multimodal data fusion are implemented.

[0014] The present invention has the following beneficial effects: The present invention first obtains the multimodal data of each monitoring moment in the preset historical period of each sub-area in the area to be regulated at the current moment, in preparation for subsequent analysis; then, for each sub-area, analyzes the similarity of changes between the multimodal data corresponding to each sub-area in the current period and each of the other sub-areas in each historical period, and preliminarily screens out all the reference periods of the current period from all the historical periods corresponding to each of the other sub-areas; and obtains the confidence of each monitoring method in each sub-area, in preparation for the subsequent analysis of the time series of the number of pests to provide accurate historical references; between the current period of each sub-area and each of the reference periods corresponding to the other sub-areas, analyzes the relationship between environmental parameters under the same environmental indicators. The difference in time series changes, and the difference in time series changes between the number of pests under the same monitoring means, analyze the impact of the environment on pest behavior from an environmental perspective alone, and then combine the actual changes of pests to accurately obtain the pest development similarity coefficient under each monitoring means; further, between the current time period of each partition and each reference time period corresponding to it in the other partitions, adjust the time series sequence of the number of pests under each monitoring means in each reference time period and its preset future time period, and obtain a confidence time series sequence to provide an accurate historical reference; finally, based on the confidence time series sequence of all reference time periods and their preset future time periods corresponding to the current time period of each partition, predict the development of pests in the area to be regulated, and assist in pest ecological regulation decision-making. The present invention preliminarily screens the reference time periods for historical reference through multimodal data, and further analyzes and evaluates from the perspective of the environment and the perspective of pest development changes to adjust the time series sequence of the number of pests in the reference time period, i.e., its future time period, thereby providing a reliable historical reference, and ultimately predicting pest development, improving the accuracy of pest development prediction, and thus improving the auxiliary decision-making effect of pest ecological regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A flowchart of a pest ecological control decision-making support method based on multimodal data fusion provided by one embodiment of the present invention; Figure 2 A flow chart of a method for obtaining pest development similarity coefficients provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0017] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of a pest ecological control decision-making support method and system based on multimodal data fusion proposed by the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The following describes in detail a pest ecological control decision-making support method and system based on multimodal data fusion provided by the present invention with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a flowchart of a pest ecological control decision-making support method based on multimodal data fusion provided by an embodiment of the present invention, specifically including: Step S1: Obtain multimodal data of each sub-area in the area to be regulated at each monitoring moment within a preset historical period at the current moment. The multimodal data includes environmental parameters under each environmental indicator and the number of pests under each monitoring method.

[0021] In one embodiment of the present invention, a region to be regulated, such as a piece of farmland, is first determined for pest ecological analysis and regulation. The region to be regulated is then evenly divided into zones to monitor and analyze pest development. Specifically, the region to be regulated is evenly divided into 5m x 5m zones in a grid-like manner. The implementer may also adjust the divisions according to actual applications. After the zones are set up, environmental monitoring sensors such as temperature sensors, humidity sensors, and light sensors are installed in each zone to collect environmental parameters under each environmental indicator, where environmental indicators include at least temperature, humidity, and light intensity. The number of pests in each zone is evaluated based on multiple monitoring methods, where the monitoring methods mainly include: (1) Install an insect collecting tank with an integrated infrared sensor. When pests enter the insect collecting tank, they block the infrared rays emitted from the tank mouth, thereby triggering an electrical signal change for counting to obtain the number of pests in the partition; (2) Arrange sound sensors (electret microphones, etc.) to emit specific sound waves when pests pass through the partition to obtain the number of pests in the partition; (3) Control a drone with an integrated camera to photograph each partition and, using image recognition technology, obtain the number of pests in that partition; (4) Install traps to obtain the number of trapped pests in the partition within a certain period of time, and use the interpolation method to obtain the number of pests in the partition at different monitoring times; It should be noted that the above monitoring methods are all well-known technologies, and implementers can also set other pest monitoring methods or other environmental monitoring indicators such as soil nutrients, etc., which will not be described in detail. After setting up various environmental monitoring sensors and pest population monitoring devices, the monitoring frequency is set to every half hour. At each monitoring moment, the environmental parameters monitored by each environmental monitoring sensor and the number of pests monitored by each monitoring method are synchronously collected. These data are used as multimodal data and transmitted to the cloud data center for analysis and processing via wireless communication. Then, at the current moment, the preset historical period is set to a historical year, and multimodal data at each monitoring moment in the historical year (including the current moment) is obtained to prepare for subsequent analysis and prediction of the future development of pests at the current moment, and to assist in pest control decision-making; implementers can also adjust the monitoring frequency and preset historical period on their own.

[0022] Step S2: In the preset historical period, the preset sub-period containing the current moment is taken as the current period, and all historical periods of the same length as the current period are obtained; for each partition, the similarity of the changes between the multimodal data corresponding to each partition in the current period and each remaining partition in each historical period is analyzed, and all reference periods of the current period are filtered out from all historical periods corresponding to each remaining partition.

[0023] The embodiment of the present invention will first determine a current period in the preset historical period for analysis, and at the same time determine a historical period of the same length as the current period. The historical period provides a historical reference for the current period, so that based on the similar environmental parameter changes and pest quantity changes between the current period and the historical period, similar references can be provided for each partition in each historical period, thereby predicting the future development of pests in each partition at the current moment.

[0024] In one embodiment of the present invention, the preset sub-period (which should be much smaller than the preset historical period) is set to 12 historical hours, and the time period with the current moment as the end point and a duration of 12 hours is obtained as the current period. At the same time, all historical periods (also 12 hours in length, but excluding the current moment) are obtained within the preset historical period, and different historical periods may overlap.

[0025] After obtaining the current period and all historical periods, we can further filter out historical periods for historical reference based on the similarity of changes between the corresponding multimodal data in the current period and historical periods in different partitions; Based on this, for each partition, the multimodal data in each partition in the current period is analyzed, and the changes in the multimodal data in each of the remaining partitions in each historical period are similar. All reference periods for the current period are filtered out from all historical periods corresponding to each of the remaining partitions.

[0026] It should be noted that the reference period is not a time period from a simple time domain perspective, but a historical period corresponding to similar changes in multimodal data provided on the basis of each partition, and the current period of each partition may have multiple reference periods in each of the other partitions, that is, there are multiple similar change references.

[0027] Preferably, in one embodiment of the present invention, considering that the analysis and acquisition methods of the corresponding reference time periods of the current time period of each partition in the other partitions are consistent, a target partition is first determined, and then any non-target partition is used as a reference partition, and any historical time period is used as a comparison time period for analysis and expression. Finally, the target partition is changed, so that the reference time period of the current time period of each partition can be obtained; and considering that multimodal data has monitoring data under multiple dimensions or modalities, any environmental indicator or monitoring method is used as the target mode for analysis to evaluate the modal sub-deviation, and then the modal sub-deviations under all modes are integrated to comprehensively evaluate the change differences between the multimodal data, and the reference factor of the comparison time period of the reference partition to the current time period of the target partition is determined, so as to screen the reference time period; Based on this, methods for obtaining reference time periods include: Take any environmental indicator or monitoring method as the target mode, any sub-area as the target sub-area, any non-target sub-area as the reference sub-area, and any historical period as the comparison period; Under the target mode, the difference between the corresponding modal data at the same monitoring time sequence number between the current period of the target partition and the comparison period of the reference partition is used to obtain the modal sub-deviation. The modal sub-deviations under all environmental indicators and monitoring methods are fused, and the negative correlation normalization result of the fusion result is used as the reference factor for the comparison period of the reference partition. Among all historical periods corresponding to all non-target partitions, all historical periods with reference factors greater than a preset reference threshold are used as reference periods for the current period of the target partition.

[0028] As an example, take the temperature in the environmental indicators as the target mode for analysis. Obtain the temperature parameter sequence of the target partition in the current period and the temperature parameter sequence of the reference partition in the comparison period. Then, at the same monitoring time sequence number (the same monitoring time sequence number may not correspond to the same time), that is, at the same sequence element sequence number of the two sequences, measure the difference by the absolute value of the difference. Average the absolute values ​​of the temperature differences at all the same monitoring time sequence numbers to obtain the temperature sub-deviation between the current period of the target partition and the comparison period of the reference partition (the modal sub-deviation under the temperature corresponding to the target mode). By changing the target mode, we can obtain the modal sub-deviations for each environmental indicator and each monitoring method. Then, we accumulate and average all the modal sub-deviations, and use the mean as the x in the exponential function exp(-x) with the natural constant e as the base, perform negative correlation mapping and normalization, and obtain the reference factor for the comparison period of the reference partition. By changing the reference partition and comparison period, the reference factors of all non-target partitions corresponding to all historical periods can be obtained; since the reference factor value range is 0-1, the preset reference threshold can be set to 0.6, and the implementer can also customize it, so that all reference periods corresponding to the current period of the target partition can be filtered out from all historical periods corresponding to all non-target partitions; finally, by changing the target partition, all reference periods corresponding to the current period of each partition can be obtained.

[0029] Step S3, in each partition, compare the deviation of the number of pests under all monitoring methods at the same monitoring time to obtain the confidence of each monitoring method; between the current time period of each partition and each corresponding reference time period in the other partitions, according to the time series change difference between environmental parameters under the same environmental indicators, and the time series change difference between the number of pests under the same monitoring method, obtain the pest development similarity coefficient under each monitoring method.

[0030] Considering that at each monitoring moment, there may be certain deviations in the number of pests obtained by different monitoring methods, which will lead to the accuracy of prediction of subsequent pest development, by comparing the number of pests under all monitoring methods, we can preliminarily evaluate the relative deviation of each monitoring method, and thus evaluate the confidence of each monitoring method.

[0031] Preferably, in one embodiment of the present invention, considering that the concentrated characteristics of pest populations under all monitoring methods can provide a general reference, the deviation of the pest population under each monitoring method from the general reference is evaluated, and the monitoring accuracy factor of each monitoring method can be provided; and considering that there is a certain degree of contingency at a single monitoring moment, if the monitoring accuracy factor can maintain a high amplitude and be stable within a preset historical period, it indicates that the monitoring accuracy or reliability of the monitoring method is high; therefore, the method for obtaining the confidence level includes: In each partition and at each monitoring moment, the reference pest number is obtained based on the concentrated characteristics of the pest number under all monitoring methods, and the monitoring accuracy factor under each monitoring method is obtained based on the deviation of the pest number under each monitoring method relative to the reference pest number. The confidence level of each monitoring method is obtained based on the amplitude level and time series fluctuation of the monitoring accuracy factor within the preset historical period.

[0032] As an example, in each partition and at each monitoring moment, the mode is used to represent the concentrated feature, the mode of the number of pests under all monitoring methods is used as the reference number of pests, and then the absolute value of the difference is used to measure the deviation, and the deviation is mapped to Negative correlation normalization is performed in the data to obtain the monitoring accuracy factor under each monitoring method; implementers can also use other negative correlation normalization methods such as mapping to the exponential function exp(-x) with the natural constant e as the base, which will not be repeated here; Among them, in a preferred embodiment of the present invention, considering that the mean reflects the average level of amplitude, and the range can reflect the drastic change of the monitoring accuracy factor in the time series to a certain extent, the larger the range, the less universal the monitoring accuracy factor is, so it can be subjected to negative correlation normalization adjustment logic; on the basis of a higher average level, if the monitoring accuracy factor changes stably, that is, the change amplitude is small, the confidence level is higher; therefore, according to the amplitude level and time series fluctuation of the monitoring accuracy factor in the preset historical period, the method of obtaining the confidence level of each monitoring means includes: For each monitoring method, the normalized result of the extreme negative correlation of the monitoring accuracy factor in the preset historical period is used as the stability factor. The stability factor is used to weight the mean of the monitoring accuracy factor in the preset historical period to obtain the confidence level.

[0033] As an example, the range of the monitoring accuracy factor in the preset historical period is mapped to The negative correlation normalization is performed to obtain the stability factor; then the mean of the monitoring accuracy factor in the preset historical period is obtained, and the mean is multiplied by the stability factor to obtain the confidence level.

[0034] Considering that for each partition, the more similar the pest population changes in the current period are to those in the corresponding reference periods of each other partition, the more similar the pest development conditions are in the current period of the partition and the other partitions in the corresponding reference periods. Considering that the environment will affect the aggregation of pests to a certain extent, the pest development conditions should also be highly similar when the environment is similar, the environmental differences between the current period of the partition and the other partitions in the corresponding reference periods are analyzed separately to eliminate the impact of multimodal data fusion on environmental assessment. Therefore, the embodiment of the present invention obtains the pest development similarity coefficient under each monitoring method based on the temporal change differences between environmental parameters under the same environmental indicators and the temporal change differences between the number of pests under the same monitoring method between the current time period of each partition and each corresponding reference time period in the other partitions.

[0035] Preferably, in one embodiment of the present invention, the method for obtaining the pest development similarity coefficient includes: See also Figure 2 , which shows a flow chart of a method for obtaining a pest development similarity coefficient provided by one embodiment of the present invention, specifically comprising: In step S201, any partition is taken as the target partition, and under each environmental indicator and each monitoring method, the environmental parameter sequence and pest quantity sequence of the target partition in the current period are obtained, as well as the reference environmental parameter sequence and reference pest quantity sequence of the corresponding non-target partition in each reference period of the current period.

[0036] In order to facilitate the subsequent differences in environmental deviations and temporal changes in pest numbers, it is first necessary to construct a sequence to provide a basis for analysis.

[0037] As an example, first determine a target partition. Taking any environmental indicator, such as temperature, as an example, all temperature data of the target partition in the current period are sorted in chronological order to construct a temperature sequence. Similarly, we can obtain the environmental parameter sequence under each other environmental indicator and the pest population sequence under each monitoring method, such as the humidity sequence, light intensity sequence, pest population sequence under the insect collection tank monitoring method, and pest population sequence under the sound sensor monitoring method. Then, all reference time periods corresponding to the current time period of the target partition are obtained; based on the above-mentioned temperature sequence construction method, all temperature data in the non-target partition corresponding to each reference time period are sorted in chronological order to construct a reference temperature sequence. Similarly, a reference environmental parameter sequence and a reference pest quantity sequence can be constructed.

[0038] Step S202: Under each environmental indicator, based on the difference between the environmental parameter sequence and each reference environmental parameter sequence, obtain the environmental sub-deviation between the target partition in the current period and the non-target partition in each reference period, and comprehensively consider the environmental sub-deviations under all environmental indicators to obtain the environmental deviation between the target partition in the current period and the non-target partition in each reference period.

[0039] In a preferred embodiment of the present invention, considering that the DTW algorithm can help evaluate the differences between time series, the method for obtaining the environmental sub-deviation includes: Under each environmental indicator, the normalized result of the DTW distance between the environmental parameter sequence and each reference environmental parameter sequence is used as the environmental sub-deviation.

[0040] As an example, the DTW distance is mapped to It should be noted that the acquisition of DTW distance is a well-known technology, and the implementer can also use other difference measurement methods such as cosine distance, Euclidean norm, etc. instead of DTW distance, or other normalization methods such as mapping to , no more details; Under each environmental indicator, if the DTW distance between the environmental parameter sequence corresponding to the target partition in the current period and the reference environmental parameter sequence corresponding to the non-target partition in each reference period is larger, it means that the environmental difference is greater, and the environmental sub-deviation between the target partition in the current period and the non-target partition in each reference period is larger; then the mean of the environmental sub-deviations under all environmental indicators is used as the environmental deviation between the target partition in the current period and the non-target partition in each reference period; in other examples, implementers can also replace the mean with the median.

[0041] Step S203 , under each monitoring method, based on the difference in pest numbers under the same sequence number between the pest number sequence and each reference pest number sequence, obtain pest change similarity parameters between the target partition in the current time period and the non-target partition in each reference time period.

[0042] As an example, the implementer can adopt the method of evaluating environmental sub-deviations in step S202, first evaluate the pest population deviation under each monitoring method, and then negatively map the pest population deviation, such as mapping it to an exponential function exp(-x) with the natural constant e as the base, and use the negative correlation mapping result as the pest change similarity parameter.

[0043] As another example, implementers can also directly measure the difference in pest numbers under the same sequence element number using the absolute value of the difference between the pest number sequence and each reference pest number sequence, and then negatively map the absolute value of the difference to the exponential function exp(-x) with the natural constant e as the base, and then use the mean of the corresponding negative correlation mapping results under all the same sequences as the pest change similarity parameter; implementers can also use other negative correlation mapping methods, which will not be repeated here.

[0044] Step S204 : weighting the pest change similarity parameter using the negative correlation mapping result of the environmental deviation to obtain the pest development similarity coefficient between the target partition in the current time period and the non-target partition in each reference time period.

[0045] Considering that the smaller the environmental deviation, the greater the environmental similarity, the logic is adjusted through negative correlation mapping so that when the environmental similarity is high, if the pest change similarity parameter is also large, it means that the pest development situation between the target partition in the current period and the non-target partition in each reference period is more similar; Therefore, as an example, the environmental deviation is first negatively correlated and mapped, such as by performing an inverse operation; then, between the target partition in the current period and the non-target partition in each reference period, the inverse of the environmental deviation is multiplied by the pest change similarity parameter to obtain the pest development similarity coefficient; implementers can also use other negative correlation mapping methods, such as mapping the negative correlation of the environmental deviation to the exponential function exp(-x) with the natural constant e as the base.

[0046] Step S4, between the current time period of each partition and each corresponding reference time period in the other partitions, according to the similarity coefficient and confidence of pest development under each monitoring method, adjust the time series of the number of pests under each reference time period and its preset future time period to obtain a confidence time series sequence; based on the confidence time series sequence of all reference time periods and their preset future time periods corresponding to the current time period of each partition, predict the development of pests in the area to be regulated, and assist in pest ecological regulation decision-making.

[0047] Considering that for each sub-area, if the pest development similarity coefficient under a certain monitoring method is higher between the current period of each sub-area and any corresponding reference period in any other sub-area, and the confidence level of the monitoring method is higher, then the development and changes of the pest population in the reference period and its future periods under this monitoring method will provide greater historical reference value for the future development and changes of the pest population in the current period of this sub-area. Based on the historical reference value provided by each monitoring method, the temporal variation sequence of the pest population in the reference period and its future periods can be adjusted to provide an accurate historical reference for subsequent prediction of pest development. Therefore, the embodiment of the present invention adjusts the time series of pest numbers under each monitoring method in each reference period and its preset future period based on the pest development similarity coefficient and confidence between the current period of each partition and each corresponding reference period in other partitions to obtain a confidence time series. The confidence time series is based on the pest development similarity coefficient and confidence level provided by different monitoring methods. Both reflect the reference value of the monitoring method. The number of pests obtained by different monitoring methods at each monitoring moment is integrated to construct a new confidence time series, which improves the monitoring accuracy of the number of pests after the fusion of multimodal monitoring methods and provides an accurate historical reference for subsequent analysis and prediction.

[0048] Preferably, in one embodiment of the present invention, the method for obtaining a trusted time series sequence includes: Taking any subarea as the target subarea, the confidence level of each monitoring method is used to weight the pest development similarity coefficient between the current period of the target subarea and each corresponding reference period in the non-target subarea. The weighted result is used as the confidence reference weight of each monitoring method in each corresponding reference period in the non-target subarea. Under each monitoring method, the time series of the number of pests in each reference period of the current period and the corresponding non-target partition in the preset future period is used as the sequence to be adjusted; the confidence reference weight under each monitoring method is used to weight the number of pests in the corresponding sequence to be adjusted and average it, and the weighted average result is used as the sequence element to construct a time series sequence to obtain a confidence time series sequence.

[0049] As an example, first, take any subarea as the target subarea, and multiply and combine the confidence level of each monitoring method and the pest development similarity coefficient between the current period of the target subarea and each corresponding reference period in the non-target subarea to obtain the confidence reference weight of each monitoring method in each reference period corresponding to the non-target subarea. Then, the duration of the preset future period is set to 2 hours, that is, under each monitoring method, in each reference period of the current period and within 2 hours after the reference period, the time series of the number of pests in the corresponding non-target partition is used as the sequence to be adjusted; the confidence reference weight under each monitoring method is used to weightedly average the number of each pest in the corresponding sequence to be adjusted, and the weighted average result of all monitoring methods is used as the corresponding confident pest number. All confident pest numbers are used as sequence elements and a time series sequence is constructed to obtain a confident time series sequence.

[0050] After changing the target partition, the confidence time series of the current period of each partition corresponding to all reference periods and their preset future periods can be obtained, which can then predict the development of pests in the area to be regulated and assist in pest ecological regulation decisions.

[0051] Preferably, in one embodiment of the present invention, considering that for each partition, the confidence time series sequence reflects the confidence pest development and change trend within the reference period in the other partitions, providing a rich historical reference for pest changes in the current period and future development period of each partition; and considering that time series prediction can predict the future development and change of the current moment when only the current pest development and change are known, a training set training time series prediction model can be constructed based on the confidence time series sequence, thereby predicting the future pest population changes in each partition, and further evaluating the pest spread, etc.; based on this, the method for predicting the development of pests in the area to be regulated includes: A training set of a preset prediction model is constructed based on all confidence time series sequences. The time series sequence of pest numbers in each partition in the current period is used as the input of the trained preset prediction model. The time series sequence of future pest numbers in each partition is output. Based on the time series sequence of future pest numbers in each partition, the future spread rate of pests in each partition is obtained.

[0052] Among them, the preset prediction model is the ARIMA-SVM model, which splits all confidence time series sequences into two segments, one is the subsequence corresponding to the reference period, and the other is the subsequence corresponding to the preset future period (2 hours) of the reference period. The subsequence corresponding to the reference period is used as the training input, and the prediction period is set to the next 2 hours. The subsequence corresponding to the preset future period (2 hours) of the reference period provides prediction verification, thereby obtaining a trained preset prediction model; finally, the time series sequence of the number of pests in each partition in the current period is input, and the model will automatically output the prediction result of the current period in the next 2 hours, that is, the pest number prediction sequence in the next two hours.

[0053] It should be noted that the training and application of the ARIMA-SVM model are well-known technologies and will not be described in detail. Implementers may also use other time series prediction algorithms and models.

[0054] For each partition, the pest population prediction sequence reflects the change in pest population within the preset next two hours from the current moment, and the first-order difference sequence of the pest population prediction sequence reflects the change rate of pests in the partition, that is, the future diffusion speed.

[0055] After obtaining the pest development prediction results for each sub-area within the area to be regulated, it can further assist in pest ecological regulation.

[0056] In one embodiment of the present invention, the predicted number of pests corresponding to each future monitoring moment in the pest quantity prediction sequence is mapped to a range of 0-255 as the pixel value of the partition; then, at each future monitoring moment, each partition is regarded as a pixel point, so that the area to be regulated is constructed as an image; in the corresponding image at each future monitoring moment, the larger the pixel value, the greater the degree of pest threat; between every two adjacent future monitoring moments, the corresponding images can also be subjected to frame difference analysis to evaluate the migration behavior of pests, thereby assisting in ecological regulation decisions; for example, when the pixel value of a certain pixel point in all images is greater than a preset threshold, a drone is used to spray biological agents such as NPV virus on the partition corresponding to the pixel point; when the pixel values ​​of pixels in a large area all exceed the preset threshold, physical harvesting of insect-source crops is performed, etc.

[0057] It should be noted that pest ecological regulation is a technical means well known to those skilled in the art, and the specific process will not be described in detail.

[0058] The present invention also proposes a pest ecological control decision-making support system based on multimodal data fusion. The system includes a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, it implements the steps of the pest ecological control decision-making support method based on multimodal data fusion.

[0059] In summary, the present invention first obtains multimodal data for each partition, screens all reference time periods for each partition in the current time period, and obtains the confidence level of each monitoring method; then, between the current time period of each partition and each corresponding reference time period, obtains the similarity coefficient of pest development under each monitoring method, and further obtains the confidence time series sequence of the number of pests in each reference time period and its preset future time period, thereby predicting the development of pests in the area to be regulated and assisting in pest ecological regulation decision-making. The present invention preliminarily screens reference time periods for historical reference through multimodal data, and then analyzes and evaluates from the perspectives of the environment and pests to obtain confidence time series sequences to provide accurate historical references, thereby improving the accuracy of pest development predictions and further improving the auxiliary decision-making effect of pest ecological regulation.

[0060] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0061] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A pest ecological control decision-making support method based on multimodal data fusion, characterized in that: The method comprises: Obtain multimodal data for each sub-area within the area to be regulated at each monitoring moment within a preset historical period at the current moment, the multimodal data including environmental parameters under each environmental indicator and the number of pests under each monitoring method; In the preset historical period, the preset sub-period containing the current moment is used as the current period, and all historical periods of the same length as the current period are obtained. For each partition, the similarity of changes between the multimodal data corresponding to each partition in the current period and each other partition in each historical period is analyzed, and all reference periods for the current period are filtered out from all historical periods corresponding to each other partition. In each subarea, the deviations in pest populations under all monitoring methods at the same monitoring time are compared to obtain the confidence level of each monitoring method. Based on the temporal variation differences between environmental parameters under the same environmental indicators and the temporal variation differences between pest populations under the same monitoring method, the similarity coefficient of pest development under each monitoring method is obtained. Between the current period of each partition and each corresponding reference period in the other partitions, according to the similarity coefficient and confidence of pest development under each monitoring method, the time series of the number of pests under each monitoring method in each reference period and its preset future period is adjusted to obtain the confidence time series sequence; based on the confidence time series sequence of all reference periods and their preset future periods corresponding to the current period of each partition, the development of pests in the area to be regulated is predicted to assist in pest ecological regulation decision-making.

2. The pest ecological control decision-making support method based on multimodal data fusion according to claim 1, characterized in that: The method for obtaining the reference period includes: Take any environmental indicator or monitoring method as the target mode, any sub-area as the target sub-area, any non-target sub-area as the reference sub-area, and any historical period as the comparison period; Under the target mode, the difference between the corresponding modal data at the same monitoring time sequence number between the current period of the target partition and the comparison period of the reference partition is used to obtain the modal sub-deviation. The modal sub-deviation under all environmental indicators and monitoring methods is fused, and the negative correlation normalization result of the fusion result is used as the reference factor for the comparison period of the reference partition; Among all historical periods corresponding to all non-target partitions, all historical periods in which the reference factor is greater than a preset reference threshold are used as reference periods for the current period of the target partition.

3. The pest ecological control decision-making support method based on multimodal data fusion according to claim 1, characterized in that: The method for obtaining the confidence level includes: In each partition and at each monitoring moment, the reference pest number is obtained based on the concentrated characteristics of the pest number under all monitoring means, and the monitoring accuracy factor under each monitoring means is obtained based on the deviation of the pest number under each monitoring means relative to the reference pest number. The confidence level of each monitoring means is obtained based on the amplitude level and time series fluctuation of the monitoring accuracy factor within a preset historical period.

4. The pest ecological control decision-making support method based on multimodal data fusion according to claim 3 is characterized in that: The method for obtaining the confidence level of each monitoring method based on the amplitude level and time series fluctuation of the monitoring accuracy factor within a preset historical period includes: For each monitoring method, the normalized result of the extreme negative correlation of the monitoring accuracy factor in the preset historical period is used as the stability factor. The stability factor is used to weight the mean of the monitoring accuracy factor in the preset historical period to obtain the confidence level.

5. The pest ecological control decision-making support method based on multimodal data fusion according to claim 1, characterized in that: The method for obtaining the pest development similarity coefficient includes: Taking any subarea as the target subarea, under each environmental indicator and each monitoring method, obtain the environmental parameter sequence and pest quantity sequence of the target subarea in the current period, as well as the reference environmental parameter sequence and reference pest quantity sequence of the corresponding non-target subarea in each reference period of the current period; Under each environmental indicator, based on the difference between the environmental parameter sequence and each reference environmental parameter sequence, the environmental sub-deviation between the target partition in the current period and the non-target partition in each reference period is obtained. The environmental sub-deviation under all environmental indicators is combined to obtain the environmental deviation between the target partition in the current period and the non-target partition in each reference period. Under each monitoring method, based on the difference in pest numbers under the same sequence number between the pest number sequence and each reference pest number sequence, the pest change similarity parameters between the target partition in the current period and the non-target partition in each reference period are obtained; The pest change similarity parameter is weighted using the negative correlation mapping result of the environmental deviation to obtain the pest development similarity coefficient between the target partition in the current time period and the non-target partition in each reference time period.

6. The pest ecological control decision-making support method based on multimodal data fusion according to claim 5, characterized in that: The method for obtaining the environmental sub-deviation includes: Under each environmental indicator, the normalized result of the DTW distance between the environmental parameter sequence and each reference environmental parameter sequence is used as the environmental sub-deviation.

7. The pest ecological control decision-making support method based on multimodal data fusion according to claim 1, characterized in that: The method for obtaining the trusted time series includes: Taking any subarea as the target subarea, between the current time period of the target subarea and each corresponding reference time period in the non-target subarea, the pest development similarity coefficient is weighted using the confidence level under each monitoring method, and the weighted result is used as the confidence reference weight of each monitoring method in each corresponding reference time period in the non-target subarea; Under each monitoring method, the time series of the number of pests in each reference period of the current period and the corresponding non-target partition in the preset future period is used as the sequence to be adjusted; the confidence reference weight under each monitoring method is used to weightedly average the number of each pest in the corresponding sequence to be adjusted, and the weighted average results of all monitoring methods are used as sequence elements to construct a time series sequence to obtain a confidence time series sequence.

8. The pest ecological control decision-making support method based on multimodal data fusion according to claim 1, characterized in that: Methods for predicting pest development in an area to be regulated include: A training set of a preset prediction model is constructed based on all the confidence time series sequences, and the time series sequence of the number of pests in each partition in the current time period is used as the input of the trained preset prediction model. The time series sequence of the future number of pests in each partition in the preset future time period at the current moment is output, and based on the time series sequence of the future number of pests in each partition, the future spread rate of pests in each partition is obtained.

9. The pest ecological control decision-making support method based on multimodal data fusion according to claim 8, characterized in that: The preset prediction model is the ARIMA-SVM model.

10. A pest ecological control decision-making support system based on multimodal data fusion, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the pest ecological control decision-making support method based on multimodal data fusion as described in any one of claims 1 to 9 are implemented.

Images