Method and system for detecting damage caused by important corn pests based on drone patrol

Abstract

The present invention discloses a method and system for detecting damage caused by important corn pests based on drone patrol. First, a two-dimensional map model is constructed, and initial image data of corn fields is obtained through drone inspections to identify pest types, quantities and damage symptoms, and at the same time evaluate the damage caused by them. Subsequently, sub-regions are fused based on the insect distribution and loss impact analysis, and changes are monitored over multiple cycles to generate insect distribution and loss impact change data. Sub-regions with linear correlation are marked using linear correlation analysis and association rule information is generated, and insect distribution and loss impact data are predicted based on the LSTM model. Finally, a prevention and control plan is generated based on the predicted data. Through the present invention, accurate insect distribution and loss impact prediction and analysis can be achieved, providing data support for subsequent information-based pest control.

Description

Technical Field

[0001] The present invention relates to the field of agricultural information technology, and more specifically, to a method and system for detecting damage caused by important corn pests based on unmanned aerial vehicle cruise. Background Art

[0002] With the rapid development of modern agriculture, timely and accurate detection of crop pests has become particularly important. Traditional methods of crop pest detection mainly rely on manual inspections and sampling analysis, which is time-consuming and labor-intensive, and is prone to missing key information. In recent years, the rapid development of drone technology has provided new possibilities for rapid and automated detection of agricultural pests. However, how to use drone technology efficiently and accurately for pest detection is still an urgent problem to be solved in the current field of agricultural information technology. And due to the constraints of traditional technology, there is currently a lack of pest correlation analysis in various planting areas of corn fields, and a lack of effective and precise prediction methods.

[0003] Therefore, it is necessary to develop a method for detecting damage caused by important corn pests based on drone patrol. Summary of the invention

[0004] The present invention overcomes the defects of the prior art and proposes a method and system for detecting damage caused by important corn pests based on drone cruising.

[0005] The first aspect of the present invention provides a method for detecting damage caused by important corn pests based on drone cruising, comprising:

[0006] Based on the target cornfield area, a two-dimensional visual map model is constructed;

[0007] The target cornfield area is inspected and monitored in real time by drones. The inspection is based on multiple sub-areas in the target cornfield area, and the corresponding initial image data is obtained. Pest identification and damage detection are performed based on the initial image data to obtain pest identification data and corn damage symptom data.

[0008] The pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact of each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions;

[0009] Acquire real-time image monitoring data of the target cornfield area within N preset periods, and perform periodic change analysis based on the image monitoring data, wherein the analysis dimensions include pest quantity distribution and pest damage impact distribution, and generate pest quantity change data and damage symptom change data for each sub-area through periodic change analysis;

[0010] Based on the pest quantity change data and damage symptom change data of each sub-region, linear correlation analysis is performed on adjacent sub-regions, sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information;

[0011] Construct a prediction model based on LSTM, serialize the pest quantity change data and damage symptom change data of each sub-region into time series and import them into the prediction model as training data for training, generate prediction data in a loop, perform parameter tuning once through the back propagation algorithm in each training process, judge the relevance of the prediction data generated by training through association rule information and perform secondary parameter tuning, and perform model training in a loop until the preset accuracy rate is reached;

[0012] Based on the trained prediction model, pest prediction data and impact prediction data are generated for each sub-region. The priority regulation analysis of the sub-region is carried out through the pest prediction data and impact prediction data, and a pest control plan for the corn field region is generated.

[0013] In this solution, a two-dimensional visualized map model is constructed based on the target cornfield area, specifically:

[0014] Acquire the planting area and regional contour information of the target corn field area, and construct a two-dimensional visualized map model based on the planting area and regional contour information;

[0015] Perform initial area division in the map model to form multiple sub-areas;

[0016] Generate the drone cruise route through the map model and ensure that it passes through all sub-areas.

[0017] In this solution, the target cornfield area is inspected and monitored in real time by drones. The inspection is based on multiple sub-areas in the target cornfield area, and the corresponding initial image data is obtained. Pest identification and damage detection are performed based on the initial image data to obtain pest identification data and corn damage symptom data, specifically:

[0018] In a preset cycle, the drone cruises the target cornfield area and stays in each sub-area for a preset time;

[0019] Build an image recognition model based on R-CNN;

[0020] Initial aerial image data and initial ground image data are obtained through drones, and the initial aerial image data and initial ground image data are imported into the image recognition model for pest identification and corn damage symptom identification. In the pest identification process, the target pests are identified, marked and counted in the image, and the number of pests in each sub-area is evaluated by the proportion of pests in the image, and pest identification data for each sub-area is obtained;

[0021] In the process of corn damage symptom recognition, corn area detection, corn growth integrity analysis and corn disease recognition analysis are performed on the initial ground image data to obtain corn damage symptom data for each sub-area;

[0022] The initial image data includes initial aerial image data and initial ground image data.

[0023] In this solution, the pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact of each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions, specifically:

[0024] Obtain pest quantity and impact level based on pest identification data and corn damage symptom data;

[0025] Sub-regions are used as analysis units, and sub-regions with similar pest distribution and impact distribution are merged. The similarity judgment process is to compare the number of pests and impact level in a current sub-region with the adjacent sub-region. If the difference between the two data is within the preset range, the adjacent sub-region is merged with the current sub-region to form a larger sub-region.

[0026] All sub-regions are fused and judged to obtain multiple fused sub-regions.

[0027] In this solution, real-time image monitoring data of the target cornfield area is obtained within N preset periods, and periodic change analysis is performed based on the image monitoring data. The analysis dimensions include pest quantity distribution and pest damage impact distribution. Through periodic change analysis, pest quantity change data and damage symptom change data of each sub-area are generated, specifically:

[0028] Acquire real-time image monitoring data of a target cornfield area within N preset cycles, wherein the real-time image monitoring data includes aerial image data and ground image data;

[0029] The image recognition model is introduced to evaluate the number and distribution of pests in each sub-area based on real-time image monitoring data. The distribution analysis is combined with the map model to analyze the distribution of pest numbers in different sub-areas. Through periodic data change analysis, the pest number change data is obtained.

[0030] An image recognition model is introduced to analyze the pest damage impact and periodic damage changes in each sub-area of ​​the real-time image monitoring data to obtain the impact change data.

[0031] In this solution, the adjacent sub-regions are subjected to linear correlation analysis based on the pest quantity change data and the damage symptom change data of each sub-region, the sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information, specifically:

[0032] Taking the sub-region as the analysis unit, mark a current sub-region and randomly select an adjacent sub-region with the current sub-region as the center;

[0033] The pest population change data of the current sub-region is used as the independent variable data, and the pest population change data of the adjacent sub-region is used as the dependent variable data to perform linear correlation analysis and calculate the correlation coefficient to obtain the correlation coefficient;

[0034] Determine whether the correlation coefficient meets the preset range. If not, it is determined that there is no correlation between the two.

[0035] If it is in accordance with the above, a linear regression-based equation fitting is performed based on the independent variable data and the dependent variable data to obtain the fitting equation;

[0036] The influence change data of the current sub-region is used as the independent variable data, and the influence change data of the adjacent sub-region is used as the dependent variable data to calculate the correlation coefficient, and whether there is a correlation is determined through a preset interval, and the equation is fitted to obtain the corresponding correlation coefficient and fitting equation;

[0037] Taking the current sub-region as the center, linear correlation analysis is performed on the remaining adjacent sub-regions;

[0038] The adjacent sub-regions with linear correlation are recorded, and the obtained correlation coefficients are integrated with the corresponding fitting equations to form the correlation information of the current sub-region;

[0039] Perform linear calculation analysis on adjacent sub-regions for all sub-regions and obtain the correlation information of each sub-region;

[0040] The association information of each sub-region is integrated to form association rule information for the sub-region.

[0041] In this scheme, the prediction model based on LSTM is constructed, the pest quantity change data and the damage symptom change data of each sub-area are time-seriesized and imported into the prediction model as training data for training, and the prediction data is generated cyclically. In each training process, the parameters are tuned once by the back propagation algorithm, the correlation of the training-generated prediction data is judged by the association rule information, and the parameters are tuned twice, and the model training is cyclically performed until the preset accuracy is reached, specifically:

[0042] Build a prediction model based on LSTM;

[0043] The pest quantity change data and damage symptom change data of each sub-region are time-seriesized and imported into the prediction model as training data for training, and the prediction data of each sub-region is generated cyclically;

[0044] In each training process, the model parameters are tuned once through the back propagation algorithm. After the first parameter tuning, the predicted data of the current sub-region and the predicted data of the adjacent sub-region are linearly correlated and analyzed to obtain the predicted association information. It is judged whether the predicted association information conforms to the association rule information, and the judgment result is used as the deviation information between the predicted data and the real data for secondary parameter tuning.

[0045] The model training is repeated until the preset accuracy is reached.

[0046] In this solution, the trained prediction model is used to generate pest prediction data and impact prediction data for each sub-region, and the priority regulation analysis of the sub-region is performed through the pest prediction data and impact prediction data, and a corn field regional pest control plan is generated, specifically:

[0047] Generate pest prediction data and impact prediction data for each sub-region based on the trained prediction model;

[0048] Through pest prediction data and impact prediction data, each sub-area is evaluated for damage detection, and the sub-areas are prioritized for prevention and control based on the evaluation results;

[0049] Based on pest prediction data, impact prediction data and sub-region priorities, drone-based pest control analysis is conducted on the target cornfield area to generate a pest control plan.

[0050] The second aspect of the present invention further provides a corn major pest damage detection system based on drone cruise, the system comprising: a memory, a processor, the memory comprising a corn major pest damage detection program based on drone cruise, the corn major pest damage detection program based on drone cruise being executed by the processor to implement the following steps:

[0051] Based on the target cornfield area, a two-dimensional visual map model is constructed;

[0052] The target cornfield area is inspected and monitored in real time by drones. The inspection is based on multiple sub-areas in the target cornfield area, and the corresponding initial image data is obtained. Pest identification and damage detection are performed based on the initial image data to obtain pest identification data and corn damage symptom data.

[0053] The pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact of each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions;

[0054] Acquire real-time image monitoring data of the target cornfield area within N preset periods, and perform periodic change analysis based on the image monitoring data, wherein the analysis dimensions include pest quantity distribution and pest damage impact distribution, and generate pest quantity change data and damage symptom change data for each sub-area through periodic change analysis;

[0055] Based on the pest quantity change data and damage symptom change data of each sub-region, linear correlation analysis is performed on adjacent sub-regions, sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information;

[0056] Construct a prediction model based on LSTM, serialize the pest quantity change data and damage symptom change data of each sub-region into time series and import them into the prediction model as training data for training, generate prediction data in a loop, perform parameter tuning once through the back propagation algorithm in each training process, judge the relevance of the prediction data generated by training through association rule information and perform secondary parameter tuning, and perform model training in a loop until the preset accuracy rate is reached;

[0057] Based on the trained prediction model, pest prediction data and impact prediction data are generated for each sub-region. The priority regulation analysis of the sub-region is carried out through the pest prediction data and impact prediction data, and a pest control plan for the corn field region is generated.

[0058] The present invention discloses a method and system for detecting important corn pests based on drone patrol. First, a two-dimensional map model is constructed, and initial image data of corn fields is obtained through drone inspections to identify pests and evaluate their impacts. Subsequently, sub-regions are fused based on pest distribution and impact analysis, and changes are monitored over multiple cycles to generate pest and damage symptom change data. Sub-regions with linear correlation are marked using linear correlation analysis and association rule information is generated, and pest and impact data are predicted based on an LSTM model. Finally, a prevention and control plan is generated based on the predicted data. Through the present invention, accurate prediction and analysis of pest and damage impacts can be achieved, providing data support for subsequent information-based pest control. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A flow chart of a method for detecting damage caused by important corn pests based on drone cruising is shown in the present invention;

[0060] Figure 2 A block diagram of a major corn pest damage detection system based on drone cruising according to the present invention is shown. DETAILED DESCRIPTION

[0061] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0063] Figure 1 A flow chart of a method for detecting damage caused by important corn pests based on drone cruising according to the present invention is shown.

[0064] like Figure 1 As shown, the first aspect of the present invention provides a method for detecting damage caused by important corn pests based on drone cruising, comprising:

[0065] S102, constructing a two-dimensional visualization map model based on the target cornfield area;

[0066] S104, patrolling and real-time monitoring the target cornfield area by using a drone, the patrolling is based on multiple sub-areas in the target cornfield area, and corresponding initial image data is obtained, pest identification and damage detection are performed based on the initial image data, and pest identification data and corn damage symptom data are obtained;

[0067] S106, performing pest distribution and corn growth impact analysis on each sub-region using pest identification data and corn damage symptom data, and merging sub-regions with similar pest distribution and impact distribution to obtain multiple fused sub-regions;

[0068] S108, obtaining real-time image monitoring data of the target cornfield area within N preset periods, performing periodic change analysis based on the image monitoring data, wherein the analysis dimensions include pest quantity distribution and pest damage impact distribution, and generating pest quantity change data and damage symptom change data for each sub-area through periodic change analysis;

[0069] S110, performing linear correlation analysis on adjacent sub-regions based on the pest quantity change data and the damage symptom change data of each sub-region, marking the sub-regions with linear correlation, and recording the calculated correlation coefficients to form association rule information;

[0070] S112, constructing a prediction model based on LSTM, time-series the pest quantity change data and the damage symptom change data of each sub-region and importing them into the prediction model as training data for training, cyclically generating prediction data, performing parameter tuning once through the back propagation algorithm during each training process, judging the relevance of the prediction data generated by the training through the association rule information and performing secondary parameter tuning, and cyclically performing model training until a preset accuracy rate is reached;

[0071] S114, generating pest prediction data and impact prediction data for each sub-region based on the trained prediction model, performing priority regulation analysis on the sub-regions through the pest prediction data and impact prediction data, and generating a pest control plan for the cornfield region.

[0072] According to an embodiment of the present invention, the two-dimensional visualized map model is constructed based on the target cornfield area, specifically:

[0073] Acquire the planting area and regional contour information of the target corn field area, and construct a two-dimensional visualized map model based on the planting area and regional contour information;

[0074] Perform initial area division in the map model to form multiple sub-areas;

[0075] Generate the drone cruise route through the map model and ensure that it passes through all sub-areas.

[0076] It should be noted that the initial regional division in the map model is to form multiple sub-regions in order to divide the smallest research unit area. In the subsequent process, appropriate regional fusion will be carried out based on the similarity of pest characteristics between regions to form larger sub-regions and improve the analysis efficiency. The drone conducts patrol and pest control work covering the entire corn field through the patrol route. The drone includes a camera and a spraying device.

[0077] According to an embodiment of the present invention, the target cornfield area is inspected and monitored in real time by a drone, the inspection is based on multiple sub-areas in the target cornfield area, and corresponding initial image data is obtained, pest identification and damage detection are performed based on the initial image data, and pest identification data and corn damage symptom data are obtained, specifically:

[0078] In a preset cycle, the drone cruises the target cornfield area and stays in each sub-area for a preset time;

[0079] Build an image recognition model based on R-CNN;

[0080] Initial aerial image data and initial ground image data are obtained through drones, and the initial aerial image data and initial ground image data are imported into the image recognition model for pest identification and corn damage symptom identification. In the pest identification process, the target pests are identified, marked and counted in the image, and the number of pests in each sub-area is evaluated by the proportion of pests in the image, and pest identification data for each sub-area is obtained;

[0081] In the process of corn damage symptom recognition, corn area detection, corn growth integrity analysis and corn disease recognition analysis are performed on the initial ground image data to obtain corn damage symptom data for each sub-area;

[0082] The initial image data includes initial aerial image data and initial ground image data.

[0083] It should be noted that the target pest is generally the fall armyworm, which has a great harmful impact on corn fields. The pest identification process includes the identification of aerial images (pests in flight) and the identification of pests on the ground in corn fields, such as pests resting on corn stems and leaves.

[0084] The corn damage symptom data includes information such as the degree of damage and the impact level. The corn growth integrity analysis and corn disease identification analysis include corn stem and leaf integrity analysis, corn growth analysis after pests have eaten, etc. The initial image data is used to perform an initial pest analysis in an early cycle, conduct a preliminary pest impact assessment on the corresponding sub-area, and conduct a preliminary pest impact assessment based on the number of pests and the damage detection of ground corn, and further merge similar impact areas.

[0085] The R-CNN is a regional convolutional neural network, which has better effect in target monitoring and analysis.

[0086] According to an embodiment of the present invention, the pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact of each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions, specifically:

[0087] Obtain pest quantity and impact level based on pest identification data and corn damage symptom data;

[0088] Sub-regions are used as analysis units, and sub-regions with similar pest distribution and impact distribution are merged. The similarity judgment process is to compare the number of pests and impact level in a current sub-region with the adjacent sub-region. If the difference between the two data is within the preset range, the adjacent sub-region is merged with the current sub-region to form a larger sub-region.

[0089] All sub-regions are fused and judged to obtain multiple fused sub-regions.

[0090] It should be noted that if the data difference between the two is within the preset range, it is determined whether the difference between the number of pests in the current sub-area and the number of pests in the adjacent sub-area is within the preset range, and the impact level judgment is the same. The preset range includes the pest difference value and the impact level difference value. The present invention performs similarity analysis on different areas based on two similarity dimensions, respectively based on the dimensions of pests and impacts, corresponding to the aerial pest analysis of the image and the corn planting impact analysis on the ground, which can effectively merge areas with similar comprehensive impacts, improve the system's ability to analyze complex areas and improve data analysis efficiency.

[0091] According to an embodiment of the present invention, real-time image monitoring data of the target cornfield area is obtained within N preset periods, and periodic change analysis is performed based on the image monitoring data. The analysis dimensions include pest quantity distribution and pest damage impact distribution. Through periodic change analysis, pest quantity change data and damage symptom change data of each sub-area are generated, specifically:

[0092] Acquire real-time image monitoring data of a target cornfield area within N preset cycles, wherein the real-time image monitoring data includes aerial image data and ground image data;

[0093] The image recognition model is introduced to evaluate the number and distribution of pests in each sub-area based on real-time image monitoring data. The distribution analysis is combined with the map model to analyze the distribution of pest numbers in different sub-areas. Through periodic data change analysis, the pest number change data is obtained.

[0094] An image recognition model is introduced to analyze the pest damage impact and periodic damage changes in each sub-area of ​​the real-time image monitoring data to obtain the impact change data.

[0095] It should be noted that N cycles correspond to N sets of real-time image monitoring data. The pest quantity change data includes pest quantity change data and distribution change data. The subsequent linear correlation analysis mainly uses the pest quantity change for calculation and analysis. The impact change data includes N hazard levels and impact level information of a sub-area in N cycles. The impact level information is used for subsequent sub-area linear correlation analysis.

[0096] According to an embodiment of the present invention, the linear correlation analysis is performed on adjacent sub-regions based on the pest quantity change data and the damage symptom change data of each sub-region, the sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information, specifically:

[0097] Taking the sub-region as the analysis unit, mark a current sub-region and randomly select an adjacent sub-region with the current sub-region as the center;

[0098] The pest population change data of the current sub-region is used as the independent variable data, and the pest population change data of the adjacent sub-region is used as the dependent variable data to perform linear correlation analysis and calculate the correlation coefficient to obtain the correlation coefficient;

[0099] Determine whether the correlation coefficient meets the preset range. If not, it is determined that there is no correlation between the two.

[0100] If it is in accordance with the above, a linear regression-based equation fitting is performed based on the independent variable data and the dependent variable data to obtain the fitting equation;

[0101] The influence change data of the current sub-region is used as the independent variable data, and the influence change data of the adjacent sub-region is used as the dependent variable data to calculate the correlation coefficient, and whether there is correlation and the generation of the fitting equation is determined through the preset interval to obtain the corresponding correlation coefficient and fitting equation;

[0102] Taking the current sub-region as the center, linear correlation analysis is performed on the remaining adjacent sub-regions;

[0103] The adjacent sub-regions with linear correlation are recorded, and the obtained correlation coefficients are integrated with the corresponding fitting equations to form the correlation information of the current sub-region;

[0104] Perform linear calculation analysis on adjacent sub-regions for all sub-regions and obtain the correlation information of each sub-region;

[0105] The association information of each sub-region is integrated to form association rule information for the sub-region.

[0106] It should be noted that the association information includes adjacent sub-region labels for the existence of correlation in a sub-region, correlation coefficients and equations based on pest (quantity) changes, correlation coefficients and equations based on impact (level) changes, and other information. By integrating all the association information, association rule information for the entire corn field area can be formed, and the data correlation between sub-regions can be grasped based on information technology.

[0107] In the embodiment of the present invention, the linear changes in the number of pests and the impact level are mainly studied and the degree of linear correlation between adjacent sub-areas is calculated. However, based on research needs, other similar change data can also be used as basic data for linear correlation, such as changes in distribution data, changes in the number of pest species, etc. The linear correlation coefficient calculation can be calculated using the Pearson correlation coefficient, which is conducive to rapid linear change calculation. The preset interval is set by the user. Within the preset interval, it means that there is a certain linear correlation. The corresponding two adjacent sub-areas need to be marked and related information recorded. Exceeding the preset interval means that there is no correlation. In addition, linear correlation has positive correlation and negative correlation; if the current sub-area is linearly correlated with the adjacent sub-area and the relevant information has been recorded, when analyzing the adjacent sub-area, there is no need to repeat the linear correlation calculation and analysis with the current sub-area. The independent variable data and the dependent variable data are both data sets of N cycles.

[0108] It is worth mentioning that in the complex cornfield environment, due to environmental influences, the pests in the cornfield often have certain migration changes, quantity changes, and changes in the proportion of different insect ages, which leads to differences in the impact of damage in different regions. The changes may often have certain linear correlations, especially for the impact changes in adjacent areas. Therefore, the present invention performs change analysis on the data of N cycles and performs linear correlation analysis based on adjacent areas to obtain association rule information that can reflect different sub-areas. The association rule information can effectively reflect the degree of association and correlation between pests and damage impacts between sub-areas. Through the effective acquisition of this information, the relationship between sub-areas can be accurately grasped. At the same time, the environmental change trend of different sub-areas can be analyzed based on information technology.

[0109] According to an embodiment of the present invention, the prediction model based on LSTM is constructed, the pest quantity change data and the damage symptom change data of each sub-region are time-series and imported into the prediction model as training data for training, and the prediction data is generated cyclically. In each training process, the back propagation algorithm is used to perform parameter tuning, and the correlation of the training-generated prediction data is judged by the association rule information and the secondary parameter tuning is performed, and the model training is cyclically performed until the preset accuracy rate is reached, specifically:

[0110] Build a prediction model based on LSTM;

[0111] The pest quantity change data and damage symptom change data of each sub-region are time-seriesized and imported into the prediction model as training data for training, and the prediction data of each sub-region is generated cyclically;

[0112] In each training process, the model parameters are tuned once through the back propagation algorithm. After the first parameter tuning, the predicted data of the current sub-region and the predicted data of the adjacent sub-region are linearly correlated and analyzed to obtain the predicted association information. It is judged whether the predicted association information conforms to the association rule information, and the judgment result is used as the deviation information between the predicted data and the real data for secondary parameter tuning.

[0113] The model training is repeated until the preset accuracy is reached.

[0114] It should be noted that the prediction data for each sub-region includes two types of prediction data, namely, pest change prediction data and impact change prediction data. The subsequent linear correlation analysis of the prediction data also includes linear analysis of these two types of prediction data, namely, linear correlation analysis of the two dimensions of pest and impact changes.

[0115] In the linear correlation analysis of the predicted data of the current sub-region and the predicted data of the adjacent sub-region, the linear correlation analysis is consistent with the method of linear correlation analysis of the pest quantity change data and the damage symptom change data of each sub-region within N preset periods.

[0116] The judgment of whether the predicted association information conforms to the association rule information includes the linear correlation judgment of the two dimensions of pests and impacts. If it conforms to the association rule information, it means that the correlation between the predicted data is consistent with the correlation between the previous N cycles, and the authenticity of the predicted data is high. For example, based on the predicted data of a current sub-region, the specific content of the calculated predicted association information is: the current sub-region has a positive data correlation with a certain adjacent sub-region, and the correlation coefficient is K. Based on the association rule information, it is judged whether the current sub-region actually exists and is positively correlated with the data of a certain adjacent sub-region, and K and the corresponding real correlation coefficient are within the preset error range, then it is considered to conform to the association rule information, and the association deviation is recorded to obtain the judgment result. Through the judgment result, it is possible to analyze whether the predicted result data conforms to the preset association rule information, and use this as the deviation from the real data, so as to correct the prediction model. In the present invention, the association rule information is used as the real data relationship.

[0117] Based on the embodiments of the present invention, in the traditional LSTM prediction model, association rule information is introduced to perform prediction training correction, and the traditional back propagation algorithm correction model parameters are retained. When performing the first parameter tuning, the model of the back propagation algorithm is corrected based on the real input data. During the secondary parameter tuning, based on the association rule information, the prediction data is constrained and optimized, so that the model is close to the prediction trend of the real data and at the same time meets the association rules between certain prediction data. If the association rules are not met, it means that the prediction results are quite different. If the association rules are met, it means that the prediction results are more accurate. The prediction deviation analysis and secondary model optimization are performed through the judgment results. The present invention introduces association rules for training optimization, which can effectively make the prediction data consistent with the actual prediction effect, and at the same time meet the correlation between sub-regions, so as to realize accurate prediction and analysis of pests and damage impacts, and provide data support for subsequent information-based pest control.

[0118] According to an embodiment of the present invention, the trained prediction model is used to generate pest prediction data and impact prediction data for each sub-region, and the priority control analysis of the sub-region is performed through the pest prediction data and the impact prediction data, and a pest control plan for the corn field region is generated, specifically:

[0119] Generate pest prediction data and impact prediction data for each sub-region based on the trained prediction model;

[0120] Through pest prediction data and impact prediction data, each sub-area is evaluated for damage detection, and the sub-areas are prioritized for prevention and control based on the evaluation results;

[0121] Based on pest prediction data, impact prediction data and sub-region priorities, drone-based pest control analysis is conducted on the target cornfield area to generate a pest control plan.

[0122] It should be noted that the hazard detection assessment is a comprehensive assessment based on the predicted number of pests and the predicted impact level. The pest control plan includes information such as drone-based monitoring plans, pesticide spraying plans, and regional priority control parameters.

[0123] Figure 2 A block diagram of a major corn pest damage detection system based on drone cruising according to the present invention is shown.

[0124] The second aspect of the present invention further provides a corn major pest damage detection system 2 based on drone cruising, the system comprising: a memory 21, a processor 22, the memory comprising a corn major pest damage detection program based on drone cruising, the corn major pest damage detection program based on drone cruising implementing the following steps when executed by the processor:

[0125] Based on the target cornfield area, a two-dimensional visual map model is constructed;

[0126] The target cornfield area is inspected and monitored in real time by drones. The inspection is based on multiple sub-areas in the target cornfield area, and the corresponding initial image data is obtained. Pest identification and damage detection are performed based on the initial image data to obtain pest identification data and corn damage symptom data.

[0127] The pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact of each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions;

[0128] Acquire real-time image monitoring data of the target cornfield area within N preset periods, and perform periodic change analysis based on the image monitoring data, wherein the analysis dimensions include pest quantity distribution and pest damage impact distribution, and generate pest quantity change data and damage symptom change data for each sub-area through periodic change analysis;

[0129] Based on the pest quantity change data and damage symptom change data of each sub-region, linear correlation analysis is performed on adjacent sub-regions, sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information;

[0130] Construct a prediction model based on LSTM, serialize the pest quantity change data and damage symptom change data of each sub-region into time series and import them into the prediction model as training data for training, generate prediction data in a loop, perform parameter tuning once through the back propagation algorithm in each training process, judge the relevance of the prediction data generated by training through association rule information and perform secondary parameter tuning, and perform model training in a loop until the preset accuracy rate is reached;

[0131] Based on the trained prediction model, pest prediction data and impact prediction data are generated for each sub-region. The priority regulation analysis of the sub-region is carried out through the pest prediction data and impact prediction data, and a pest control plan for the corn field region is generated.

[0132] According to an embodiment of the present invention, the two-dimensional visualized map model is constructed based on the target cornfield area, specifically:

[0133] Acquire the planting area and regional contour information of the target corn field area, and construct a two-dimensional visualized map model based on the planting area and regional contour information;

[0134] Perform initial area division in the map model to form multiple sub-areas;

[0135] Generate the drone cruise route through the map model and ensure that it passes through all sub-areas.

[0136] It should be noted that the initial regional division in the map model is to form multiple sub-regions in order to divide the smallest research unit area. In the subsequent process, appropriate regional fusion will be carried out based on the similarity of pest characteristics between regions to form larger sub-regions and improve the analysis efficiency. The drone conducts patrol and pest control work covering the entire corn field through the patrol route. The drone includes a camera and a spraying device.

[0137] According to an embodiment of the present invention, the target cornfield area is inspected and monitored in real time by a drone, the inspection is based on multiple sub-areas in the target cornfield area, and corresponding initial image data is obtained, pest identification and damage detection are performed based on the initial image data, and pest identification data and corn damage symptom data are obtained, specifically:

[0138] In a preset cycle, the drone cruises the target cornfield area and stays in each sub-area for a preset time;

[0139] Build an image recognition model based on R-CNN;

[0140] Initial aerial image data and initial ground image data are obtained through drones, and the initial aerial image data and initial ground image data are imported into the image recognition model for pest identification and corn damage symptom identification. In the pest identification process, the target pests are identified, marked and counted in the image, and the number of pests in each sub-area is evaluated by the proportion of pests in the image, and pest identification data for each sub-area is obtained;

[0141] In the process of corn damage symptom recognition, corn area detection, corn growth integrity analysis and corn disease recognition analysis are performed on the initial ground image data to obtain corn damage symptom data for each sub-area;

[0142] The initial image data includes initial aerial image data and initial ground image data.

[0143] It should be noted that the target pest is generally the fall armyworm, which has a great impact on corn fields. The pest identification process includes the identification of aerial images (pests in flight) and the identification of pests on the ground in corn fields, such as pests resting on corn stems and leaves.

[0144] The corn damage symptom data includes information such as the degree of damage and the impact level. The corn growth integrity analysis and corn disease identification analysis include corn stem and leaf integrity analysis, corn growth analysis after pests have eaten, etc. The initial image data is used to perform an initial pest analysis in an early cycle, conduct a preliminary pest impact assessment on the corresponding sub-area, and conduct a preliminary pest impact assessment based on the number of pests and the damage detection of ground corn, and further merge similar impact areas.

[0145] The R-CNN is a regional convolutional neural network, which has better effect in target monitoring and analysis.

[0146] According to an embodiment of the present invention, the pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact of each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions, specifically:

[0147] Obtain pest quantity and impact level based on pest identification data and corn damage symptom data;

[0148] Sub-regions are used as analysis units, and sub-regions with similar pest distribution and impact distribution are merged. The similarity judgment process is to compare the number of pests and impact level in a current sub-region with the adjacent sub-region. If the difference between the two data is within the preset range, the adjacent sub-region is merged with the current sub-region to form a larger sub-region.

[0149] All sub-regions are fused and judged to obtain multiple fused sub-regions.

[0150] It should be noted that if the data difference between the two is within the preset range, it is determined whether the difference between the number of pests in the current sub-area and the number of pests in the adjacent sub-area is within the preset range, and the impact level judgment is the same. The preset range includes the pest difference value and the impact level difference value. The present invention performs similarity analysis on different areas based on two similarity dimensions, respectively based on the dimensions of pests and impacts, corresponding to the aerial pest analysis of the image and the corn planting impact analysis on the ground, which can effectively merge areas with similar comprehensive impacts, improve the system's ability to analyze complex areas and improve data analysis efficiency.

[0151] According to an embodiment of the present invention, real-time image monitoring data of the target cornfield area is obtained within N preset periods, and periodic change analysis is performed based on the image monitoring data. The analysis dimensions include pest quantity distribution and pest damage impact distribution. Through periodic change analysis, pest quantity change data and damage symptom change data of each sub-area are generated, specifically:

[0152] Acquire real-time image monitoring data of a target cornfield area within N preset cycles, wherein the real-time image monitoring data includes aerial image data and ground image data;

[0153] The image recognition model is introduced to evaluate the number and distribution of pests in each sub-area based on real-time image monitoring data. The distribution analysis is combined with the map model to analyze the distribution of pest numbers in different sub-areas. Through periodic data change analysis, the pest number change data is obtained.

[0154] An image recognition model is introduced to analyze the pest damage impact and periodic damage changes in each sub-area of ​​the real-time image monitoring data to obtain the impact change data.

[0155] It should be noted that N cycles correspond to N sets of real-time image monitoring data. The pest quantity change data includes pest quantity change data and distribution change data. The subsequent linear correlation analysis mainly uses the pest quantity change for calculation and analysis. The impact change data includes N hazard levels and impact level information of a sub-area in N cycles. The impact level information is used for subsequent sub-area linear correlation analysis.

[0156] According to an embodiment of the present invention, the linear correlation analysis is performed on adjacent sub-regions based on the pest quantity change data and the damage symptom change data of each sub-region, the sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information, specifically:

[0157] Taking the sub-region as the analysis unit, mark a current sub-region and randomly select an adjacent sub-region with the current sub-region as the center;

[0158] The pest population change data of the current sub-region is used as the independent variable data, and the pest population change data of the adjacent sub-region is used as the dependent variable data to perform linear correlation analysis and calculate the correlation coefficient to obtain the correlation coefficient;

[0159] Determine whether the correlation coefficient meets the preset range. If not, it is determined that there is no correlation between the two.

[0160] If it is in accordance with the above, a linear regression-based equation fitting is performed based on the independent variable data and the dependent variable data to obtain the fitting equation;

[0161] The influence change data of the current sub-region is used as the independent variable data, and the influence change data of the adjacent sub-region is used as the dependent variable data to calculate the correlation coefficient, and whether there is correlation and the generation of the fitting equation is determined through the preset interval to obtain the corresponding correlation coefficient and fitting equation;

[0162] Taking the current sub-region as the center, linear correlation analysis is performed on the remaining adjacent sub-regions;

[0163] The adjacent sub-regions with linear correlation are recorded, and the obtained correlation coefficients are integrated with the corresponding fitting equations to form the correlation information of the current sub-region;

[0164] Perform linear calculation analysis on adjacent sub-regions for all sub-regions and obtain the correlation information of each sub-region;

[0165] The association information of each sub-region is integrated to form association rule information for the sub-region.

[0166] It should be noted that the association information includes labels for adjacent sub-regions that are correlated with a sub-region, correlation coefficients and equations based on changes in pests (quantity), correlation coefficients and equations based on changes in impact (level), and other information. By integrating all the association information, association rule information for the entire corn field area can be formed. The association rule information can be used to grasp the data correlation between sub-regions based on information technology.

[0167] In the embodiment of the present invention, the linear changes in the number of pests and the impact level are mainly studied and the degree of linear correlation between adjacent sub-areas is calculated. However, based on research needs, other similar change data can also be used as basic data for linear correlation, such as changes in distribution data, changes in the number of pest species, etc. The linear correlation coefficient calculation can be calculated using the Pearson correlation coefficient, which is conducive to rapid linear change calculation. The preset interval is set by the user. Within the preset interval, it means that there is a certain linear correlation. The corresponding two adjacent sub-areas need to be marked and related information recorded. Exceeding the preset interval means that there is no correlation. In addition, linear correlation has positive correlation and negative correlation; if the current sub-area is linearly correlated with the adjacent sub-area and the relevant information has been recorded, when analyzing the adjacent sub-area, there is no need to repeat the linear correlation calculation and analysis with the current sub-area. The independent variable data and the dependent variable data are both data sets of N cycles.

[0168] It is worth mentioning that in the complex cornfield environment, due to environmental influences, the pests in the cornfield often have certain migration changes, quantity changes, and changes in the proportion of different insect ages, which leads to differences in the impact of damage in different regions. The changes may often have certain linear correlations, especially for the impact changes in adjacent areas. Therefore, the present invention performs change analysis on the data of N cycles and performs linear correlation analysis based on adjacent areas to obtain association rule information that can reflect different sub-areas. The association rule information can effectively reflect the degree of association and correlation between pests and damage impacts between sub-areas. Through the effective acquisition of this information, the relationship between sub-areas can be accurately grasped. At the same time, the environmental change trend of different sub-areas can be analyzed based on information technology.

[0169] According to an embodiment of the present invention, the prediction model based on LSTM is constructed, the pest quantity change data and the damage symptom change data of each sub-region are time-series and imported into the prediction model as training data for training, and the prediction data is generated cyclically. In each training process, the back propagation algorithm is used to perform parameter tuning, and the correlation of the training-generated prediction data is judged by the association rule information and the secondary parameter tuning is performed, and the model training is cyclically performed until the preset accuracy rate is reached, specifically:

[0170] Build a prediction model based on LSTM;

[0171] The pest quantity change data and damage symptom change data of each sub-region are time-seriesized and imported into the prediction model as training data for training, and the prediction data of each sub-region is generated cyclically;

[0172] In each training process, the model parameters are tuned once through the back propagation algorithm. After the first parameter tuning, the predicted data of the current sub-region and the predicted data of the adjacent sub-region are linearly correlated and analyzed to obtain the predicted association information. It is judged whether the predicted association information conforms to the association rule information, and the judgment result is used as the deviation information between the predicted data and the real data for secondary parameter tuning.

[0173] The model training is repeated until the preset accuracy is reached.

[0174] It should be noted that the prediction data for each sub-region includes two types of prediction data, namely, pest change prediction data and impact change prediction data. The subsequent linear correlation analysis of the prediction data also includes linear analysis of these two types of prediction data, namely, linear correlation analysis of the two dimensions of pest and impact changes.

[0175] In the linear correlation analysis of the predicted data of the current sub-region and the predicted data of the adjacent sub-region, the linear correlation analysis is consistent with the method of linear correlation analysis of the pest quantity change data and the damage symptom change data of each sub-region within N preset periods.

[0176] The judgment of whether the predicted association information conforms to the association rule information includes the linear correlation judgment of the two dimensions of pests and impacts. If it conforms to the association rule information, it means that the correlation between the predicted data is consistent with the correlation between the previous N cycles, and the authenticity of the predicted data is high. For example, based on the predicted data of a current sub-region, the specific content of the calculated predicted association information is: the current sub-region has a positive data correlation with a certain adjacent sub-region, and the correlation coefficient is K. Based on the association rule information, it is judged whether the current sub-region actually exists and is positively correlated with the data of a certain adjacent sub-region, and K and the corresponding real correlation coefficient are within the preset error range, then it is considered to conform to the association rule information, and the association deviation is recorded to obtain the judgment result. Through the judgment result, it is possible to analyze whether the predicted result data conforms to the preset association rule information, and use this as the deviation from the real data, so as to correct the prediction model. In the present invention, the association rule information is used as the real data relationship.

[0177] Based on the embodiments of the present invention, in the traditional LSTM prediction model, association rule information is introduced to perform prediction training correction, and the traditional back propagation algorithm correction model parameters are retained. When performing the first parameter tuning, the model of the back propagation algorithm is corrected based on the real input data. During the secondary parameter tuning, based on the association rule information, the prediction data is constrained and optimized, so that the model is close to the prediction trend of the real data and at the same time meets the association rules between certain prediction data. If the association rules are not met, it means that the prediction results are quite different. If the association rules are met, it means that the prediction results are more accurate. The prediction deviation analysis and secondary model optimization are performed through the judgment results. The present invention introduces association rules for training optimization, which can effectively make the prediction data consistent with the actual prediction effect, and at the same time meet the correlation between sub-regions, so as to realize accurate prediction and analysis of pests and damage impacts, and provide data support for subsequent information-based pest control.

[0178] According to an embodiment of the present invention, the trained prediction model is used to generate pest prediction data and impact prediction data for each sub-region, and the priority control analysis of the sub-region is performed through the pest prediction data and the impact prediction data, and a pest control plan for the corn field region is generated, specifically:

[0179] Generate pest prediction data and impact prediction data for each sub-region based on the trained prediction model;

[0180] Through pest prediction data and impact prediction data, each sub-area is evaluated for damage detection, and the sub-areas are prioritized for prevention and control based on the evaluation results;

[0181] Based on pest prediction data, impact prediction data and sub-region priorities, drone-based pest control analysis is conducted on the target cornfield area to generate a pest control plan.

[0182] The present invention discloses a method and system for detecting important corn pests based on drone patrol. First, a two-dimensional map model is constructed, and initial image data of corn fields is obtained through drone inspections to identify pests and evaluate their impacts. Subsequently, sub-regions are fused based on pest distribution and impact analysis, and changes are monitored over multiple cycles to generate pest and damage symptom change data. Sub-regions with linear correlation are marked using linear correlation analysis and association rule information is generated, and pest and impact data are predicted based on an LSTM model. Finally, a prevention and control plan is generated based on the predicted data. Through the present invention, accurate prediction and analysis of pest and damage impacts can be achieved, providing data support for subsequent information-based pest control.

[0183] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0184] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0185] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0186] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0187] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0188] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for detecting important corn pests based on drone cruise, characterized in that: include: Based on the target cornfield area, a two-dimensional visual map model is constructed; The target cornfield area is inspected and monitored in real time by drones. The inspection is based on multiple sub-areas in the target cornfield area, and the corresponding initial image data is obtained. Pest identification and damage detection are performed based on the initial image data to obtain pest identification data and corn damage symptom data. The pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact of each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions; Acquire real-time image monitoring data of the target cornfield area within N preset periods, and perform periodic change analysis based on the image monitoring data, wherein the analysis dimensions include pest quantity distribution and pest damage impact distribution, and generate pest quantity change data and damage symptom change data for each sub-area through periodic change analysis; Based on the pest quantity change data and damage symptom change data of each sub-region, linear correlation analysis is performed on adjacent sub-regions, sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information; Construct a prediction model based on LSTM, serialize the pest quantity change data and damage symptom change data of each sub-region into time series and import them into the prediction model as training data for training, generate prediction data in a loop, perform parameter tuning once through the back propagation algorithm in each training process, judge the relevance of the prediction data generated by training through association rule information and perform secondary parameter tuning, and perform model training in a loop until the preset accuracy rate is reached; Generate pest prediction data and impact prediction data for each sub-region based on the trained prediction model, conduct priority regulation analysis of the sub-regions through the pest prediction data and impact prediction data, and generate a pest control plan for the corn field region; Among them, the target corn field area is inspected and monitored in real time by a drone, the inspection is based on multiple sub-areas in the target corn field area, and the corresponding initial image data is obtained, pest identification and damage detection are performed based on the initial image data, and pest identification data and corn damage symptom data are obtained, specifically: In a preset cycle, the drone cruises the target cornfield area and stays in each sub-area for a preset time; Build an image recognition model based on R-CNN; Initial aerial image data and initial ground image data are obtained through drones, and the initial aerial image data and initial ground image data are imported into the image recognition model for pest identification and corn damage symptom identification. In the pest identification process, the target pests are identified, marked and counted in the image, and the number of pests in each sub-area is evaluated by the proportion of pests in the image, and pest identification data for each sub-area is obtained; In the process of corn damage symptom recognition, corn area detection, corn growth integrity analysis and corn disease recognition analysis are performed on the initial ground image data to obtain corn damage symptom data for each sub-area; The initial image data includes initial aerial image data and initial ground image data; The pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact on each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions, specifically: Obtain pest quantity and impact level based on pest identification data and corn damage symptom data; Sub-regions are used as analysis units, and sub-regions with similar pest distribution and impact distribution are merged. The similarity judgment process is to compare the number of pests and impact level in a current sub-region with the adjacent sub-region. If the difference between the two data is within the preset range, the adjacent sub-region is merged with the current sub-region to form a larger sub-region. All sub-regions are fused and judged to obtain multiple fused sub-regions.

2. The method for detecting important corn pests based on drone cruise according to claim 1, characterized in that: The two-dimensional visualized map model is constructed based on the target cornfield area, specifically: Acquire the planting area and regional contour information of the target corn field area, and construct a two-dimensional visualized map model based on the planting area and regional contour information; Perform initial area division in the map model to form multiple sub-areas; Generate the drone cruise route through the map model and ensure that it passes through all sub-areas.

3. The method for detecting important corn pests based on drone cruising according to claim 1, characterized in that: In the N preset periods, real-time image monitoring data of the target cornfield area is obtained, and periodic change analysis is performed based on the image monitoring data. The analysis dimensions include pest quantity distribution and pest damage symptom distribution. Through periodic change analysis, pest quantity change data and damage symptom change data of each sub-area are generated, specifically: Acquire real-time image monitoring data of a target cornfield area within N preset cycles, wherein the real-time image monitoring data includes aerial image data and ground image data; The image recognition model is introduced to evaluate the number and distribution of pests in each sub-area based on real-time image monitoring data. The distribution analysis is combined with the map model to analyze the distribution of pest numbers in different sub-areas. Through periodic data change analysis, the pest number change data is obtained. An image recognition model is introduced to analyze the pest damage impact and periodic damage changes in each sub-area of ​​the real-time image monitoring data to obtain the impact change data.

4. The method for detecting important corn pests based on drone cruising according to claim 3 is characterized in that: Based on the pest quantity change data and the damage symptom change data of each sub-region, a linear correlation analysis is performed on adjacent sub-regions, sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information, specifically: Taking the sub-region as the analysis unit, mark a current sub-region and randomly select an adjacent sub-region with the current sub-region as the center; The pest population change data of the current sub-region is used as the independent variable data, and the pest population change data of the adjacent sub-region is used as the dependent variable data to perform linear correlation analysis and calculate the correlation coefficient to obtain the correlation coefficient; Determine whether the correlation coefficient meets the preset range. If not, it is determined that there is no correlation between the two. If it is in accordance with the above, a linear regression-based equation fitting is performed based on the independent variable data and the dependent variable data to obtain the fitting equation; The influence change data of the current sub-region is used as the independent variable data, and the influence change data of the adjacent sub-region is used as the dependent variable data to calculate the correlation coefficient, and whether there is correlation and the generation of the fitting equation is determined through the preset interval to obtain the corresponding correlation coefficient and fitting equation; Taking the current sub-region as the center, linear correlation analysis is performed on the remaining adjacent sub-regions; The adjacent sub-regions with linear correlation are recorded, and the obtained correlation coefficients are integrated with the corresponding fitting equations to form the correlation information of the current sub-region; Perform linear calculation analysis on adjacent sub-regions for all sub-regions and obtain the correlation information of each sub-region; The association information of each sub-region is integrated to form association rule information for the sub-region.

5. The method for detecting important corn pests based on drone cruising according to claim 4 is characterized in that: The prediction model based on LSTM is constructed, the pest quantity change data and the damage symptom change data of each sub-region are time-seriesed and imported into the prediction model as training data for training, and the prediction data is generated cyclically. In each training process, the parameters are tuned once by the back propagation algorithm, the correlation of the training-generated prediction data is judged by the association rule information, and the parameters are tuned twice, and the model training is cyclically performed until the preset accuracy is reached, specifically: Build a prediction model based on LSTM; The pest quantity change data and damage symptom change data of each sub-region are time-seriesized and imported into the prediction model as training data for training, and the prediction data of each sub-region is generated cyclically; In each training process, the model parameters are tuned once through the back propagation algorithm. After the first parameter tuning, the predicted data of the current sub-region and the predicted data of the adjacent sub-region are linearly correlated and analyzed to obtain the predicted association information. It is judged whether the predicted association information conforms to the association rule information, and the judgment result is used as the deviation information between the predicted data and the real data for secondary parameter tuning. The model training is repeated until the preset accuracy is reached.

6. The method for detecting important corn pests based on drone cruising according to claim 5, characterized in that: The trained prediction model is used to generate pest prediction data and impact prediction data for each sub-region, and the priority control analysis of the sub-region is performed based on the pest prediction data and the impact prediction data, and a pest control plan for the corn field region is generated, specifically: Generate pest prediction data and impact prediction data for each sub-region based on the trained prediction model; Through pest prediction data and impact prediction data, each sub-area is evaluated for damage detection, and the sub-areas are prioritized for prevention and control based on the evaluation results; Based on pest prediction data, impact prediction data and sub-region priorities, drone-based pest control analysis is conducted on the target cornfield area to generate a pest control plan.

7. A corn major pest damage detection system based on drone cruise, characterized in that: The system includes: a memory and a processor, wherein the memory includes a maize major pest damage detection program based on drone cruise, and the maize major pest damage detection program based on drone cruise is executed by the processor to implement the following steps: Based on the target cornfield area, a two-dimensional visual map model is constructed; The target cornfield area is inspected and monitored in real time by drones. The inspection is based on multiple sub-areas in the target cornfield area, and the corresponding initial image data is obtained. Pest identification and damage detection are performed based on the initial image data to obtain pest identification data and corn damage symptom data. The pest identification data and corn damage symptom data are used to analyze the pest distribution and corn growth impact of each sub-region, and the sub-regions with similar pest distribution and impact distribution are merged to obtain multiple fused sub-regions; Acquire real-time image monitoring data of the target cornfield area within N preset periods, and perform periodic change analysis based on the image monitoring data, wherein the analysis dimensions include pest quantity distribution and pest damage impact distribution, and generate pest quantity change data and damage symptom change data for each sub-area through periodic change analysis; Based on the pest quantity change data and damage symptom change data of each sub-region, linear correlation analysis is performed on adjacent sub-regions, sub-regions with linear correlation are marked, and the calculated correlation coefficients are recorded to form association rule information; Construct a prediction model based on LSTM, serialize the pest quantity change data and damage symptom change data of each sub-region into time series and import them into the prediction model as training data for training, generate prediction data in a loop, perform parameter tuning once through the back propagation algorithm in each training process, judge the relevance of the prediction data generated by training through association rule information and perform secondary parameter tuning, and perform model training in a loop until the preset accuracy rate is reached; Based on the trained prediction model, pest prediction data and impact prediction data are generated for each sub-region. The priority regulation analysis of the sub-region is carried out through the pest prediction data and impact prediction data, and a pest control plan for the corn field region is generated.

Images