Forestry intelligent pest control system based on multi-sensor cooperation

Abstract

The application discloses a forestry intelligent pest control system based on multi-sensor cooperation and relates to the technical field of intelligent pest control. In order to solve the problems of low diagnosis accuracy and poor cooperation efficiency in the prior art, the application proposes a "hypothesis-verification" cooperation strategy. The system first generates a preliminary diagnosis hypothesis by using macro-sensor data and then adaptively adjusts and controls micro-sensor for targeted verification according to the preliminary diagnosis hypothesis, so as to realize high-confidence diagnosis. The system further establishes a pest spatial-temporal diffusion risk map to realize trend prediction and formulates a graded and zoned precision control strategy. Through closed-loop self-optimization learning of the prevention and control effect evaluation and the model, the application realizes intelligent control from passive response to active prediction, and significantly improves the diagnosis accuracy and resource utilization efficiency.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of intelligent pest prevention, and particularly relates to a forestry intelligent pest prevention system based on multi-sensor cooperation. BACKGROUND

[0002] Traditional forestry pest prevention relies on inefficient and lagging manual patrol. In order to improve efficiency, satellite remote sensing and unmanned aerial vehicles and other macro monitoring technologies have been introduced, but they can only identify abnormal stress of vegetation and cannot accurately diagnose specific pest types, resulting in poor diagnostic specificity and high false positive rate, and still requiring a large amount of manual review. Although existing technologies attempt to integrate multiple sensor information to improve accuracy, they generally use simple parallel data collection and "stacking" integration, that is, all data are collected without distinction for analysis. This approach lacks a cooperative logic for targeted detection based on preliminary diagnosis, resulting in high system energy consumption and high cost, and failing to form an efficient and interpretable diagnostic verification chain. In addition, existing prevention and control decision functions are relatively weak, mostly stopping at disease diagnosis and generally lacking dynamic prediction capability for the spatiotemporal diffusion trend of pests, so that it is impossible to develop a truly graded and zoned precision prevention and control strategy. More importantly, the entire "monitoring-diagnosis-prevention" process is often an open-loop system, lacking a feedback link for prevention and control effect evaluation and model self-optimization, making it difficult for the system to achieve self-evolution and continuous learning, and limiting the level of intelligence. SUMMARY

[0003] In view of the above problems existing in the prior art, the present application aims to provide a forestry intelligent pest prevention system based on multi-sensor cooperation, comprising:

[0004] A multi-dimensional perception module configured to collect multi-modal perception information of a forestry target area, the multi-modal perception information at least including first sensor data and second sensor data; the first sensor data is macro data reflecting the physiological state of vegetation, and the second sensor data is micro data representing direct or indirect characteristics of pests.

[0005] A central processing module electrically connected to the multi-dimensional perception module, configured to, based on the first sensor data, identify an abnormal vegetation area and generate at least one preliminary diagnosis hypothesis through a preset vegetation anomaly analysis model, the preliminary diagnosis hypothesis including a candidate pest type and a corresponding initial confidence; adaptively adjust the working parameters or data collection priority of the second sensor according to the preliminary diagnosis hypothesis; fuse the adjusted second sensor data and the first sensor data to construct a cooperative verification model, verify or correct the preliminary diagnosis hypothesis, and output a final diagnosis conclusion with a verified confidence.

[0006] An intelligent decision-making and control module is electrically connected with the central processing module and is configured to generate and execute a targeted pest control instruction according to the final diagnosis conclusion.

[0007] Preferably, the first type of sensor data is selected from at least one of satellite remote sensing images, unmanned aerial vehicle multi-spectral / hyper-spectral data, and chlorophyll fluorescence data; and the second type of sensor data is selected from at least one of high-resolution image data, in-forest acoustic vibration signals, pest pheromone concentration data, or specific volatile organic compound concentration data.

[0008] Preferably, the central processing module generates a preliminary diagnosis hypothesis, specifically, the first type of sensor data is input into a pre-trained feature matching knowledge base storing macroscopic data feature spectra of vegetation under different pest stress, and the preliminary diagnosis hypothesis and an initial confidence level are generated by calculating the similarity between the input data and the feature spectra in the knowledge base.

[0009] Preferably, the central processing module adaptively adjusts the working parameters of the second type of sensor according to the preliminary diagnosis hypothesis, including: if the preliminary diagnosis hypothesis is a boring drywood termite, the collection frequency and analysis sensitivity of the in-forest acoustic sensor are increased; and if the preliminary diagnosis hypothesis is a leaf disease, the unmanned aerial vehicle is instructed to conduct high-resolution visible light or thermal imaging detailed investigation on the target area.

[0010] Preferably, the collaborative verification model is a dynamic weight fusion model, and the central processing module dynamically adjusts the weight coefficient of the first type of sensor data in generating the final diagnosis conclusion by taking the verification result of the second type of sensor data as an adjustment factor when performing fusion; and when the second type of data supports the preliminary diagnosis hypothesis, the weight of the corresponding first type of data feature is increased.

[0011] Preferably, the method comprises the following steps:

[0012] In the method, first type of sensor data of a forestry target area is acquired, and the first type of sensor data is macroscopic data reflecting the physiological state of vegetation.

[0013] In the method, the first type of sensor data is processed, an abnormal vegetation area is identified through a preset vegetation anomaly analysis model, and at least one preliminary diagnosis hypothesis is generated, the preliminary diagnosis hypothesis including a candidate pest type and a corresponding initial confidence level.

[0014] In the method, in response to the preliminary diagnosis hypothesis, a second type of sensor is adaptively regulated to collect second type of sensor data, and the second type of sensor data is microcosmic data representing direct or indirect characteristics of pests.

[0015] The second type of sensor data and the first type of sensor data are input into a collaborative verification model, the preliminary diagnosis hypothesis is verified or corrected by the collaborative verification model to generate a final diagnosis conclusion with verified confidence.

[0016] Based on the final diagnosis conclusion, a prevention and control instruction is generated, and a prevention and control device is controlled to execute the prevention and control instruction.

[0017] Preferably, further comprising: based on the final diagnosis conclusion, and in combination with topographic data, meteorological data of the target area, and a biological propagation model of the diagnosed pest, a pest spatio-temporal diffusion risk map is established to predict the spread trend and influence range of the pest in a future period of time.

[0018] The step S5 comprises,

[0019] The spatio-temporal diffusion risk map is automatically divided into at least multiple functional areas including a core prevention and control area, a prevention buffer area, and a monitoring observation area according to a preset risk value threshold.

[0020] For the core prevention and control area and the prevention buffer area, the priority of the prevention and control means and the precise dosage of the pesticide or natural enemy are dynamically calculated according to the specific risk level, vegetation density data in the area, and the biological model of the diagnosed pest.

[0021] The topographic data and the energy consumption model of the prevention and control device are used as constraint conditions, and the shortest operation time, the lowest energy consumption, and the highest prevention and control coverage rate are used as optimization objectives to plan an optimal execution path for the prevention and control device to pass through the core prevention and control area and the prevention buffer area.

[0022] Preferably, the collaborative verification model in the step S4 adopts a Bayesian inference network, wherein the preliminary diagnosis hypothesis generated from the first type of sensor data is used as prior probability, the features collected by the second type of sensor are used as new evidence, the confidence in various pest types is updated by calculating posterior probability to obtain the final diagnosis conclusion.

[0023] Preferably, further comprising: after executing the prevention and control instruction, the first type of sensor data and the second type of sensor data are collected again, the prevention and control effect is evaluated, and the evaluation result is fed back to iteratively optimize the vegetation anomaly analysis model and the collaborative verification model.

[0024] Compared with the prior art, the application has the beneficial effects that: the application significantly improves the accuracy, efficiency and reduces the running cost of disease and pest diagnosis by constructing an innovative "hypothesis-verification" closed-loop logic. Unlike the parallel data fusion of the prior art, the application first generates a preliminary diagnosis hypothesis using macro data, and then targets the micro sensor for verification according to the hypothesis. This hierarchical and progressive cooperative mode avoids the waste of sensor resources and upgrades the diagnosis from fuzzy anomaly detection to high-confidence cause confirmation, providing a solid foundation for precision prevention and control.

[0025] The application improves disease and pest control from passive diagnosis to active prediction and dynamic planning, realizing intelligent closed-loop management. After accurate diagnosis, the system can establish a spatiotemporal diffusion risk map to predict the trend of disease and pest spread, and automatically generate a hierarchical and zoned precision prevention and control strategy and optimal execution path accordingly, ensuring efficient use of prevention and control resources. More importantly, the application introduces a feedback mechanism for prevention and control effect evaluation and model self-optimization, enabling the system to continuously learn and self-improve, demonstrating excellent adaptive ability and long-term application value. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 The application is a system module composition schematic diagram.

[0027] Figure 2 The application is an example step flowchart of the disease and pest control method.

[0028] Figure 3 The application is an example step flowchart of the spread prediction and planning method. DETAILED DESCRIPTION

[0029] The application will be further described below in conjunction with specific embodiments.

[0030] As Figure 1 The application provides a forestry intelligent disease and pest control system based on multi-sensor cooperation, which includes:

[0031] A multi-dimensional perception module is configured to collect multi-modal perception information of a forestry target area, including at least first sensor data and second sensor data; the first sensor data is macro data reflecting the physiological state of vegetation, and the second sensor data is micro data representing direct or indirect characteristics of diseases and pests.

[0032] The central processing module is electrically connected with the multi-dimensional perception module, and is configured to identify an abnormal vegetation area and generate at least one preliminary diagnosis hypothesis including a candidate pest type and a corresponding initial confidence level by a preset vegetation anomaly analysis model based on the first type of sensor data; adaptively adjust working parameters or data acquisition priorities of the second type of sensor according to the preliminary diagnosis hypothesis; fuse the adjusted second type of sensor data and the first type of sensor data to construct a collaborative verification model, verify or correct the preliminary diagnosis hypothesis, and output a final diagnosis conclusion with a verified confidence level.

[0033] The intelligent decision and control module is electrically connected with the central processing module, and is configured to generate and execute a targeted pest control instruction according to the final diagnosis conclusion.

[0034] It should be noted that in the embodiment, the multi-dimensional perception module is preferably composed of an air-based unmanned aerial vehicle platform, a ground-based fixed sensor network and an under-forest mobile node. The unmanned aerial vehicle platform is equipped with a hyperspectral and thermal infrared camera for periodic aerial macroscopic data; the ground-based sensor network includes a chlorophyll fluorescence detector, an information pheromone collector and a VOC detector; the under-forest mobile node includes a small automatic inspection vehicle equipped with a high-resolution camera and an acoustic sensor. Different sensors are networked through LoRa, NB-IoT or 4G / 5G communication methods, and all collected data are marked with a time stamp and gathered to the central processing module.

[0035] For example, in a certain pine forest monitoring scene, the hyperspectral image shows that the NDVI value of the target area decreases by about 40%, and the central processing module generates a preliminary diagnosis hypothesis of "suspected pine wood nematode disease" with an initial confidence level of 0.61. Then, the module issues an instruction to increase the sampling rate of the acoustic sensor from 2kHz to 6kHz to capture the typical activity soundprint of the borer type pest. Finally, after the collaborative verification model fuses the image data and acoustic data, the diagnosis conclusion is "pine wood nematode disease, confidence level 0.83".

[0036] The first type of sensor data is selected from at least one of satellite remote sensing images, unmanned aerial vehicle multispectral / hyperspectral data and chlorophyll fluorescence data; and the second type of sensor data is selected from at least one of high-resolution image data, forest acoustic vibration signals, pest pheromone concentration or specific volatile organic compound concentration data.

[0037] In this embodiment, the first type of macro data is collected at a period of 7-10 days, for example, the unmanned aerial vehicle hyperspectral data covers 400-1000 nm, with a resolution of 0.2 m; the satellite remote sensing data uses Sentinel-2 NDVI products (10 m resolution, collection interval 5 days); and the chlorophyll fluorescence array records the photosynthetic efficiency change at an interval of 10 minutes. The second type of micro data is collected at a period of several hours to real-time level, for example, the understory acoustic sensor captures the xylem micro-vibration signal at a bandwidth of 2-8 kHz, the VOC sensor samples at an interval of 5 minutes, detects the concentration of key volatile organic compounds such as methanol, acetaldehyde, pinene, etc., and the resolution is ≤5 ppb.

[0038] For example, in a monitoring task, the unmanned aerial vehicle NDVI value in the first type of data decreases to 0.35, combined with the chlorophyll fluorescence Fv / Fm downward trend (from 0.82 to 0.65), the system generates a preliminary diagnosis of leaf disease. Subsequently, the high-resolution image of the second type of data shows typical spot texture, and the methanol concentration in the VOC concentration increases to 12 ppb, thereby enhancing the diagnostic confidence of “leaf spot disease”.

[0039] The example data is shown in the following table:

[0040]

[0041] For example, in a monitoring task, the unmanned aerial vehicle NDVI value in the first type of data decreases to 0.35, combined with the chlorophyll fluorescence Fv / Fm downward trend (from 0.82 to 0.65), the system generates a preliminary diagnosis of leaf disease. Subsequently, the high-resolution image of the second type of data shows typical spot texture, and the methanol concentration in the VOC concentration increases to 12 ppb, thereby enhancing the diagnostic confidence of “leaf spot disease”.

[0042] In this embodiment, the central processing module generates a preliminary diagnosis hypothesis, specifically: inputting the first type of sensor data into a pre-trained feature matching knowledge base, the knowledge base stores the macro data feature spectrum of vegetation under different disease and pest stress, and generating a preliminary diagnosis hypothesis and an initial confidence by calculating the similarity of the input data and the feature spectrum in the knowledge base.

[0043] For example, when the unmanned aerial vehicle hyperspectral data has a similarity of 0.73 with the “leaf spot disease / early stage” template and a similarity of 0.58 with the “rust disease / medium stage” template, the system outputs a preliminary diagnosis hypothesis of “leaf spot disease” with an initial confidence of 0.62.

[0044] The central processing module adaptively adjusts the working parameters of the second type of sensor according to the preliminary diagnosis hypothesis, including: if the preliminary diagnosis hypothesis is a dry wood boring pest, the collection frequency and analysis sensitivity of the forest acoustic sensor are increased; if the preliminary diagnosis hypothesis is a leaf disease, the unmanned aerial vehicle is instructed to conduct high-resolution visible light or thermal imaging detailed investigation on the target area.

[0045] The collaborative verification model is a dynamic weight fusion model; when the central processing module performs fusion, the verification result of the second type of sensor data is used as an adjustment factor to dynamically adjust the weight coefficient of the first type of sensor data in generating the final diagnosis conclusion; when the second type of data supports the preliminary diagnosis hypothesis, the weight of the corresponding first type of data feature is increased. For example: when the diagnosis hypothesis is "wood boring pest", the sampling rate of the trunk acoustic sensor is increased from 2 kHz to 6 kHz, and the amplification gain is increased by 9 dB, which is used to capture the 2-8 kHz interval pest activity signal; when the diagnosis hypothesis is "leaf disease", the unmanned aerial vehicle is instructed to obtain 0.2 m resolution RGB image and collect leaf temperature distribution for subsequent thermal stress identification.

[0046] For example, in a pine forest detection task, the central processing module issues an acoustic sensor parameter patch according to the suspected pine wood nematode disease hypothesis, and completes the detailed data collection of the area within 30 minutes.

[0047] As shown in Figure 2 The embodiment performs the following steps:

[0048] Step S1, acquiring first type of sensor data of a forestry target area, the first type of sensor data being macroscopic data reflecting the physiological state of vegetation.

[0049] Step S2, processing the first type of sensor data, identifying an abnormal vegetation area through a predetermined vegetation anomaly analysis model, and generating at least one preliminary diagnosis hypothesis, the preliminary diagnosis hypothesis including a candidate disease and pest type and an initial confidence.

[0050] Step S3, in response to the preliminary diagnosis hypothesis, adaptively regulating the second type of sensor to collect second type of sensor data, the second type of sensor data being microscopic data representing direct or indirect characteristics of diseases and pests.

[0051] Step S4, inputting the second type of sensor data and the first type of sensor data into a collaborative verification model, and verifying or correcting the preliminary diagnosis hypothesis by the collaborative verification model to generate a final diagnosis conclusion with a verified confidence.

[0052] Step S5, generating a control instruction based on the final diagnosis conclusion, and controlling the control equipment to execute the control instruction.

[0053] It should be noted that in an example process: step S1 acquires a drone hyperspectral image matrix M ∈ ℝ^{1024×1024×64}; step S2 model outputs an abnormal area mask Mask, accounting for 12%, generates a diagnostic hypothesis “leaf spot disease” initial confidence 0.58; step S3 triggers the drone to collect 10 high-resolution images and infrared temperature distribution; step S4 obtains the conclusion “leaf spot disease (moderate)” after collaborative verification, with a confidence of 0.87; step S5 generates a task script, and the drone is instructed to execute 0.3 L / acre dose spraying in the core control area, and uploads the flight trajectory and drug amount record after the operation is completed.

[0054] After step S4, it further includes: based on the final diagnosis conclusion, and combined with the topographic and geomorphic data of the target area, meteorological data, and the biological transmission model of the diagnosed pests and diseases, a pest and disease spatio-temporal diffusion risk map is established, which is used to predict the spread trend and influence range of pests and diseases in a future period of time.

[0055] As shown in Figure 3 , step S5 in this embodiment includes a sub-step,

[0056] Step S501, the spatio-temporal diffusion risk map is automatically divided into at least a plurality of functional areas including a core control area, a prevention buffer area and a monitoring observation area according to a preset risk value threshold.

[0057] Step S502, for the core control area and the prevention buffer area, the priority of the control means and the precise dosage of the pesticide or natural enemy are dynamically calculated according to the specific risk level, vegetation density data in the area and the biological model of the diagnosed pests and diseases.

[0058] For example, the division standard is core area R≥0.8, buffer area 0.6≤R<0.8, and observation area 0.4≤R<0.6. The pesticide dosage is calculated by the formula , wherein is the host density, and Sev is the disease severity. The priority score is calculated by weighting the risk value, severity and wind direction exposure coefficient.

[0059] For example, the dosage in the core area A1 (host density 0.75, severity 0.6) is 0.30 L / acre; the dosage in the buffer area B3 is 0.18 L / acre, and a sex pheromone trapping belt with a width of 30m is arranged.

[0060] Step S503, taking the topographic and geomorphic data and the energy consumption model of the control equipment as constraint conditions, and taking the shortest operation time, the lowest energy consumption and the highest control coverage rate as optimization objectives, an optimal execution path is planned for the control equipment to pass through the core control area and the prevention buffer area.

[0061] For example, the path optimization objective function is Wherein T is time, E is energy consumption, Coverage is coverage rate, and the constraint conditions include no-fly zone, slope greater than 35° detour, and battery remaining amount not less than 20%. The solving method is A* heuristic search combined with genetic algorithm optimization.

[0062] For example, in a UAV operation task, the optimal path length is 3.6 km, the estimated time is 18.5 minutes, the coverage rate is 97.8%, and the remaining power after execution is 23%.

[0063] In the collaborative verification model in step S4, the preliminary diagnosis hypothesis generated by the first type of sensor data is used as the prior probability, the features collected by the second type of sensor are used as new evidence, the confidence in various types of pests and diseases is updated by calculating the posterior probability, and the final diagnosis conclusion is obtained.

[0064] After step S5, further comprising: after executing the prevention and control instruction, collecting the first type of sensor data and the second type of sensor data again, evaluating the prevention and control effect, and feeding back the evaluation result for iterative optimization of the vegetation anomaly analysis model and the collaborative verification model.

[0065] The above examples are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A forestry intelligent pest control system based on multi-sensor cooperation, characterized in that, The method comprises the following steps: A multi-dimensional perception module is configured to collect multi-modal perception information of a forestry target area, including at least first sensor data and second sensor data; The first sensor data is macro data reflecting the physiological state of vegetation, and the second sensor data is micro data representing direct or indirect characteristics of pests and diseases; A central processing module is electrically connected to the multi-dimensional perception module and is configured to, based on the first sensor data, identify an abnormal vegetation area by a preset vegetation anomaly analysis model and generate at least one preliminary diagnosis hypothesis, wherein the preliminary diagnosis hypothesis includes a candidate pest and disease type and a corresponding initial confidence level; adaptively adjust the working parameters or data collection priority of the second sensor according to the preliminary diagnosis hypothesis; fuse the second sensor data collected after adjustment with the first sensor data to construct a collaborative verification model, verify or correct the preliminary diagnosis hypothesis, and output a final diagnosis conclusion with a verified confidence level; An intelligent decision and control module is electrically connected to the central processing module and is configured to generate and execute targeted pest and disease control instructions based on the final diagnosis conclusion; The collaborative verification model is a dynamic weight fusion model; when performing fusion, the central processing module dynamically adjusts the weight coefficient of the first sensor data in generating the final diagnosis conclusion by taking the verification result of the second sensor data as an adjustment factor; when the second data supports the preliminary diagnosis hypothesis, the weight of the corresponding first data feature is increased.

2. The multi-sensor coordination-based intelligent forest pest control system according to claim 1, characterized in that: The first sensor data is selected from at least one of satellite remote sensing images, unmanned aerial vehicle multi-spectral / hyper-spectral data, and chlorophyll fluorescence data; and the second sensor data is selected from at least one of high-resolution image data, forest acoustic vibration signals, pest pheromone concentration data, or specific volatile organic compound concentration data.

3. The multi-sensor coordination-based intelligent forest pest control system of claim 1, wherein: The central processing module generates a preliminary diagnosis hypothesis by inputting the first sensor data into a pre-trained feature matching knowledge base, wherein the knowledge base stores vegetation macro data feature spectra under different pest and disease stresses, and generates the preliminary diagnosis hypothesis and an initial confidence level by calculating the similarity between the input data and the feature spectra in the knowledge base.

4. The multi-sensor coordination-based intelligent forest pest control system according to claim 1, characterized in that: The central processing module adaptively adjusts the working parameters of the second sensor according to the preliminary diagnosis hypothesis, including: if the preliminary diagnosis hypothesis is a boring pest, the collection frequency and analysis sensitivity of the forest acoustic sensor are increased; and if the preliminary diagnosis hypothesis is a leaf disease, the unmanned aerial vehicle is instructed to perform high-resolution visible light or thermal imaging detailed investigation on the target area.

5. The multi-sensor coordination-based intelligent forest pest control system according to claim 1, characterized in that: A prevention and control method is executed, comprising the following steps: Step S1: acquiring first sensor data of a forestry target area, wherein the first sensor data is macro data reflecting the physiological state of vegetation; Step S2: processing the first sensor data to identify an abnormal vegetation area by a preset vegetation anomaly analysis model and generate at least one preliminary diagnosis hypothesis, wherein the preliminary diagnosis hypothesis includes a candidate pest and disease type and a corresponding initial confidence level; Step S3, in response to the preliminary diagnosis hypothesis, adaptively regulating the second type of sensor to collect second type of sensor data, the second type of sensor data being microscopic data characterizing direct or indirect characteristics of the pest; Step S4, inputting the second type of sensor data and the first type of sensor data into a collaborative verification model, verifying or correcting the preliminary diagnosis hypothesis by the collaborative verification model to generate a final diagnosis conclusion with a verified confidence; Step S5, based on the final diagnosis conclusion, generating a control instruction, and controlling the control equipment to execute the control instruction.

6. The multi-sensor coordination based intelligent forest pest control system according to claim 5, characterized in that: After the step S4, further comprising: based on the final diagnosis conclusion, and combining topographic and geomorphic data of the target area, meteorological data, and a biological propagation model of the diagnosed pest, establishing a pest spatio-temporal diffusion risk map for predicting the spread trend and influence range of the pest in a future period of time.

7. The multi-sensor collaboration based intelligent forest pest control system according to claim 6, characterized in that: The step S5 comprises, Step S501, dividing the spatio-temporal diffusion risk map into at least multiple functional areas including a core control area, a prevention buffer area, and a monitoring observation area according to a preset risk value threshold; Step S502, for the core control area and the prevention buffer area, dynamically calculating the priority of the control means, and the precise dosage of the pesticide or natural enemy according to the specific risk level, the vegetation density data, and the biological propagation model of the diagnosed pest in the area; Step S503, taking the topographic and geomorphic data and the energy consumption model of the control equipment as constraint conditions, and taking the shortest operation time, the lowest energy consumption, and the highest control coverage rate as optimization objectives, to plan an optimal execution path for the control equipment to pass through the core control area and the prevention buffer area.

8. The multi-sensor collaboration based intelligent forest pest control system of claim 5, wherein: The collaborative verification model in the step S4 adopts a Bayesian inference network, wherein the preliminary diagnosis hypothesis generated by the first type of sensor data is taken as a prior probability, the features collected by the second type of sensor are taken as new evidence, the confidence in various pest types is updated by calculating a posterior probability to obtain the final diagnosis conclusion.

9. The multi-sensor collaboration based intelligent forest pest control system of claim 5, wherein: After the step S5, further comprising: after executing the control instruction, collecting the first type of sensor data and the second type of sensor data again, evaluating the control effect, and feeding back the evaluation result for iteratively optimizing the vegetation anomaly analysis model and the collaborative verification model.

Images