Hilly mountain natural rubber garden intelligent inspection method and device

Abstract

The application discloses a hilly and mountainous natural rubber plantation intelligent inspection method and device, relates to the technical field of forestry and agricultural intelligence equipment, and solves the problems of low artificial inspection efficiency of the existing rubber plantation and insufficient unmanned aerial vehicle under-forest monitoring. The device is based on a four-legged walking platform, integrates a navigation module, a data acquisition module and a pest condition detection module; the method generates an inspection route and a tree body database, fuses positioning walking, RGB preliminary screening-high spectrum accurate detection, multi-source data fusion to obtain a comprehensive risk value, divides levels and generates a report. The device can adapt to a 25-35 degree slope, the positioning error is less than or equal to 5 cm, accurate identification of diseases and insect pests and quantitative monitoring of insect conditions are realized, autonomous return is supported, and the inspection efficiency and accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent equipment technology for forestry and agriculture, and in particular to a smart inspection method and device for natural rubber plantations in hilly and mountainous areas. Background Technology

[0002] Natural rubber plantations are mostly located in tropical hilly and mountainous areas, where forest roads are narrow, slopes are steep, and the soil is loose and slippery. Current pest and disease inspections typically rely on manual sampling on foot and visual interpretation or single multi-rotor drone aerial photography. Manual inspections are labor-intensive, inefficient, and highly subjective; drones are limited by understory obstruction, dense canopy, and flight time, making it difficult to conduct close-range, fixed-point, and repetitive high-spectral fine-grained identification of tree trunks and quantitative monitoring of pests.

[0003] Regarding the aforementioned technologies, the inventors believe there is an urgent need for a smart inspection equipment and method that can move nimbly through complex forest terrain, maintain stable close-range perception, achieve rapid initial screening and high-precision classification of pests and diseases, and link with quantitative determination of pest infestation. Summary of the Invention

[0004] To address the issue of intelligent inspection and pest assessment in rubber plantations, this application provides a smart inspection method and device for natural rubber plantations in hilly and mountainous areas.

[0005] The intelligent inspection method and device for natural rubber plantations in hilly and mountainous areas provided in this application adopts the following technical solution:

[0006] A smart inspection method for natural rubber plantations in hilly and mountainous areas includes the following steps:

[0007] Step 1: Generate an inspection route and target natural rubber tree database based on the work area boundary and road / rubber tapping road information, and set the stopping posture for each target tree;

[0008] Step 2: The navigation module uses GNSS RTK, IMU and visual odometry to locate the current position and the target tree position. The panoramic camera collects information on the rubber forest operation environment in which the device is currently located. The electronic control unit plans the walking path of the quadrupedal walking platform and performs gait control and obstacle avoidance in real time.

[0009] Step 3: Use a depth camera to estimate the trunk normal and distance, and drive the robotic arm and moving guide rail to precisely move the hyperspectral camera to the target area;

[0010] Step 4: Use an RGB camera for rapid initial screening of pests and diseases. If the actual value of the first confidence level P1 is greater than the first confidence level threshold K1, the hyperspectral camera is triggered to perform line scan acquisition and complete the "white / dark" reference calibration. If the actual value of the first confidence level P1 is less than the first confidence level threshold K1, the system switches to the next tree.

[0011] Step 5: Preprocess and extract features from the spectral data collected by the hyperspectral camera. Use a classification model to output the pest type and the actual value of the second confidence level P2. When the actual value of the second confidence level P2 detected by the classification model is greater than the second confidence threshold K2, the electronic control unit is triggered to control the quadrupedal walking platform to switch to the parking mode. If the actual value of the second confidence level P2 is less than the second confidence threshold K2, switch to the next tree.

[0012] Step 6: The insect infestation detection module starts working, collecting insect infestation data and environmental parameters in the rubber plantation, and obtaining the insect population density and trend score per unit time;

[0013] Step 7: Integrate the first confidence level P1, the second confidence level P2, the insect population density, and environmental parameters to obtain a comprehensive risk value R, classify the risk level, and generate an inspection report that includes geographical location, tree number, and image / spectral summary;

[0014] Step 8: When the comprehensive risk value R reaches the warning threshold, output handling suggestions and decide whether to continue patrolling or return to base / recharge based on the task completion status or battery level. If the comprehensive risk value R does not reach the warning threshold, continue patrolling.

[0015] Optionally, in step two, in areas with weak RTK coverage, the target tree is repeatedly visited and located using manual markers for visual positioning, with a positioning error not exceeding 5 cm. In step five, spectral preprocessing employs Savitzky-Golay smoothing and standard normal variable transformation; feature extraction uses competitive adaptive reweighted sampling, ReliefF algorithm, and stepwise regression projection to select key bands; and the classification model uses partial least squares discrimination, support vector machine, and one-dimensional convolutional neural network. In step seven, fusion calculation employs Bayesian fusion or learning-based weighting strategies, and based on the numerical range of R, the risk level is divided into at least four levels: Level I Normal, Level II Attention, Level III Warning, and Level IV Response. The inspection report is uploaded to the cloud platform via 5G, Wi-Fi, and LoRa communication links, and a spatiotemporal heat map of pests and diseases is generated.

[0016] A smart inspection device for natural rubber plantations in hilly areas includes a quadrupedal walking platform. The platform is equipped with a data acquisition module and a pest detection module. The data acquisition module includes a robotic arm with an RGB camera and the data acquisition platform at its end. The data acquisition platform includes a movable guide rail, which is a spirally ascending semi-circular structure at the end of the robotic arm. A movable guide slide rail is mounted on the guide rail, and a track motor is slidably mounted on the slide rail. A hyperspectral camera is mounted on the track motor. The pest detection module includes a support frame mounted on the quadrupedal walking platform. A panoramic camera is mounted on the top of the support frame. Inside the support frame, from top to bottom, are arranged a pest trapping and killing component, a collection component, a detection component, and an insect storage box. A depth camera is mounted at the front end of the quadrupedal walking platform. The platform also includes a navigation module and an electronic control unit, which are electrically connected to the data acquisition platform, the panoramic camera, the robotic arm, the pest detection module, and the navigation module.

[0017] Optionally, the hyperspectral camera is equipped with a light shield and a supplementary light on its outer side.

[0018] Optionally, the hyperspectral camera is equipped with a floating track, which is arranged vertically, and the track motor is equipped with a floating motor, which drives the hyperspectral camera to slide on the floating track.

[0019] Optionally, the robotic arm includes a robotic arm mounting base, which is mounted on a quadrupedal walking platform. A first connecting gear is rotatably mounted on the robotic arm mounting base, and a first motor is mounted on the robotic arm mounting base to drive the first connecting gear to rotate. A first arm segment is mounted on the first connecting gear, and a second connecting gear is rotatably mounted at the end of the first arm segment. A second arm segment is mounted on the second connecting gear, and a second motor is mounted on the second arm segment to drive the second connecting gear to rotate. A third connecting gear is rotatably mounted at the end of the second arm segment, and a third arm segment is mounted on the third connecting gear. A third motor is mounted on the third arm segment to drive the third connecting gear to rotate. An RGB camera is mounted on the third arm segment, and a telescopic push rod is mounted on the third arm segment. A telescopic arm segment is mounted on the telescopic push rod, and the data acquisition platform is mounted on the telescopic arm segment.

[0020] Optionally, the insect trapping and killing component includes a trapping frame with a trapping lamp on it. Above the trapping lamp is a dosing box, and below the trapping frame is an insect collecting dish. The insect collecting dish is funnel-shaped and has an upper pipe below it. An upper baffle is slidably mounted on a support frame below the upper pipe. A pipe push rod is also mounted on the support frame. Below the upper baffle is a middle pipe corresponding to the upper pipe. A negative pressure motor is mounted on the middle pipe. A collection box is mounted on the support frame. A collection camera is mounted on the top of the collection box. The lower outlet of the middle pipe leads to the collection box. Below the collection box is a lower pipe with a lower baffle and a first motor. A lower baffle is mounted at the output end of the first motor. An insect storage box is located at the outlet of the lower pipe.

[0021] Optionally, the navigation module includes a GNSS RTK receiver, an inertial measurement unit, and a visual odometry unit, and the electronic control unit performs multi-sensor fusion positioning. The navigation module also includes a visual positioning unit that identifies artificial markers to achieve accurate alignment for repeated visits in areas where RTK signals are blocked. The quadrupedal walking platform can walk continuously on terrain with a slope of 25° to 35° and pass through terrain with a slope of 40° for a short period of time.

[0022] Optionally, the electronic control unit includes an industrial computer and a wireless communication module, used to perform fusion analysis on the above data and upload it to the cloud for management. The industrial computer has edge computing capabilities.

[0023] In summary, this application includes at least one of the following beneficial technical effects:

[0024] 1) The electronic control unit integrates the first confidence level P1, the second confidence level P2, the insect population density, and environmental parameters, and generates a comprehensive risk value R and risk levels I-IV through Bayesian fusion and other strategies. The inspection report is uploaded to the cloud via 5G / Wi-Fi / LoRa and a spatiotemporal heat map of pests and diseases is generated. It supports autonomous return after the task ends or when the battery is low. Edge computing can complete early identification locally, reducing manual intervention and providing accurate data support for pest and disease control in rubber plantations. The navigation module adopts GNSS RTK, IMU and visual odometry fusion positioning. In areas with weak RTK coverage, visual positioning is achieved through manual markers, and the error of repeated visits does not exceed 5cm. In the detection stage, RGB cameras are used for rapid initial screening of pests and diseases, and hyperspectral cameras are used for fine detection by combining Savitzky-Golay smoothing, CARS feature extraction and PLS-DA / SVM / 1D-CNN classification models. Combined with the insect detection module, the insect population density is quantitatively obtained, which greatly improves the accuracy of pest and disease identification and monitoring.

[0025] 2) The quadrupedal walking platform can continuously walk on slopes of 25° to 35° and briefly traverse slopes of 40°. It can adapt to the complex environment of narrow roads and soft, slippery soil in hilly rubber plantations, solving the problems of high physical exertion in traditional manual inspections and the limitations of drones due to forest cover and limited battery life. It enables all-terrain mobile inspections. By pre-setting inspection routes, automatically aligning with target trees, and linking initial screening with fine inspection, it shortens the inspection time for a single tree and enables time-series retesting of the same tree, improving overall inspection efficiency and reducing rubber plantation management costs. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of an intelligent inspection device for natural rubber plantations in hilly areas according to a preferred embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the robotic arm structure of a smart inspection device for natural rubber plantations in hilly areas according to a preferred embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram showing the location of the data acquisition platform of the intelligent inspection device for natural rubber plantations in hilly areas according to a preferred embodiment of the present invention.

[0029] Figure 4 This is a schematic diagram of the insect detection module of a smart inspection device for natural rubber plantations in hilly areas according to a preferred embodiment of the present invention.

[0030] Figure 5 This is a flowchart illustrating a preferred embodiment of the intelligent inspection device control method for natural rubber plantations in hilly and mountainous areas according to the present invention.

[0031] Explanation of reference numerals in the attached diagram: 1000, Data acquisition platform; 1001, Moving guide rail; 1002, Moving guide slide rail; 1104, Track motor; 1105, Limiting rubber pad; 1106, Supplemental light; 1107, Light shield; 1108, Floating track; 1109, Floating motor; 2000, RGB camera; 2001, Panoramic camera; 2002, Depth camera; 2003, Hyperspectral camera; 3000, Robotic arm; 3101, Third arm section; 3102, Third motor; 3103, Third connecting gear; 3104, Second arm section; 3105, Second motor; 3106, First arm section; 3107, First motor; 3108, First connecting gear; 3109, Robotic arm mounting base; 311 1. Second connecting gear; 3112. Telescopic push rod; 3113. Telescopic arm; 4000. Insect detection module; 4101. Support frame; 4102. Insect storage box; 4103. Lower section pipe; 4104. Collection box; 4105. Observation window; 4106. Collection camera; 4107. Middle section pipe; 4108. Upper section pipe; 4109. Insect collection dish; 4110. Trapping rack; 4111. Dosing box; 4112. Trapping light; 4113. Upper baffle; 4114. Pipe push rod; 4115. Negative pressure motor; 4116. Support frame; 4117. Lower baffle first motor; 4118. Lower baffle second motor; 5000. Four-legged walking platform; 6000. Navigation module; 7000. Electrical control unit. Detailed Implementation

[0032] The terminology used in the following embodiments is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to also include expressions such as “one or more,” unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of this application, “at least one” and “one or more” refer to one, two, or more than two. The term “and / or” is used to describe the relationship between related objects, indicating that three relationships may exist; for example, A and / or B can indicate: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character “ / ” generally indicates that the preceding and following related objects are in an “or” relationship.

[0033] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0034] The following is in conjunction with the appendix Figure 1-5 The present invention will be described in further detail below.

[0035] This application discloses a smart inspection device for natural rubber plantations in hilly and mountainous areas, referring to... Figure 1 , Figure 2 and Figure 3The system includes a quadrupedal walking platform 5000, on which a data acquisition module and an insect detection module 4000 are installed. The data acquisition module includes a robotic arm 3000, at the end of which is an RGB camera 2000 and a data acquisition platform 1000. The data acquisition platform 1000 includes a moving guide rail 1001, which is a spirally ascending semi-circular structure located at the end of the robotic arm 3000, used to accurately collect insect information on rubber trees. A moving guide slide rail 1002 is provided on the moving guide slide rail 1001, and a track motor 1104 is slidably mounted on the moving guide slide rail 1002. A hyperspectral camera 2003 is mounted on the track motor 1104, and a light shield 1107 and a supplementary light 1106 are provided on the outside of the hyperspectral camera 2003. A floating track 1108 is provided on the hyperspectral camera 2003, which is arranged vertically. A floating motor 1109 is provided on the track motor 1104. The motor 1109 can slide on the floating track 1108 to enable the hyperspectral camera 2003 to float vertically. The track motor 1104 is also equipped with a limiting rubber pad 1105 to limit the lowest position of the hyperspectral camera 2003 to prevent it from squeezing the track motor 1104. The insect detection module 4000 includes a support frame 4116, which is set on the quadrupedal walking platform 5000. A panoramic camera 2001 is installed on the top of the support frame 4116. The support frame 4116 contains, from top to bottom, an insect trapping and killing component, a collection component, a detection component, and an insect storage box 4102. A depth camera 2002 is installed at the front end of the quadrupedal walking platform 5000. The quadrupedal walking platform 5000 contains a navigation module 6000 and an electronic control unit 7000. The electronic control unit 7000 is electrically connected to the data acquisition platform 1000, the panoramic camera 2001, the robotic arm 3000, the insect detection module 4000, and the navigation module 6000.

[0036] Reference Figure 1 The navigation module 6000 includes a GNSS RTK receiver, an inertial measurement unit (IMU), and a visual odometry unit (VIO). The electronic control unit 7000 performs multi-sensor fusion positioning. The navigation module 6000 also includes a visual positioning unit that recognizes artificial markers to achieve accurate alignment for repeated visits in areas where RTK signals are blocked. The quadrupedal walking platform 5000 can walk continuously on terrain with a slope of 25° to 35° and pass through terrain with a slope of 40° for a short period of time. The electronic control unit 7000 includes an industrial computer and a wireless communication module for fusing and analyzing the above data and uploading it to the cloud for management. The industrial computer has edge computing capabilities and can complete early identification of pests and diseases locally.

[0037] Reference Figure 2The robotic arm 3000 includes a robotic arm mounting base 3109, which is mounted on a quadrupedal walking platform 5000. A first connecting gear 3108 is rotatably mounted on the robotic arm mounting base 3109. A first motor 3107 is mounted on the robotic arm mounting base 3109 to drive the first connecting gear 3108 to rotate. A first arm segment 3106 is mounted on the first connecting gear 3108. A second connecting gear 3111 is rotatably mounted at the end of the first arm segment 3106. A second arm segment 3104 is mounted on the second connecting gear 3111. The second arm segment 3104 is equipped with... The second motor 3105 is used to drive the second connecting gear 3111 to rotate. The end of the second arm 3104 is rotatably provided with a third connecting gear 3103. The third connecting gear 3103 is provided with a third arm 3101. The third arm 3101 is provided with a third motor 3102 for driving the third connecting gear 3103 to rotate. The third arm 3101 is provided with an RGB camera 2000. The third arm 3101 is provided with a telescopic push rod 3112. The telescopic push rod 3112 is provided with a telescopic arm 3113. The data acquisition platform 1000 is provided on the telescopic arm 3113.

[0038] Reference Figure 4 The insect trapping and killing component includes a trapping frame 4110, a trapping lamp 4112 on the trapping frame 4110, a dosing box 4111 above the trapping lamp 4112, an insect collection dish 4109 below the trapping frame 4110, the insect collection dish 4109 being funnel-shaped, an upper section pipe 4108 below the insect collection dish 4109, an upper baffle 4113 slidably mounted on a support frame 4116 below the upper section pipe 4108, and a pipe push rod 4114 on the support frame 4116 for pushing the upper baffle 4113 to move at the outlet below the upper section pipe 4108 to open and close the upper pipe. A middle section pipe 4107 corresponding to the upper section pipe 4108 is located below the upper baffle 4113. When the upper baffle 4113 is open, the upper section pipe 4108 and the middle section pipe 4107 are connected. A negative pressure motor 4 is mounted on the middle section pipe 4107. 115. A collection box 4104 is provided on the support frame 4116. An observation window 4105 is provided on the collection box 4104. A collection camera 4106 is provided on the top of the collection box 4104. The outlet of the middle section pipe 4107 leads to the collection box 4104. The negative pressure motor 4115 sucks the insects in the upper section pipe 4108 into the collection box 4104. A lower section pipe 4103 is provided below the collection box 4104. A lower baffle first motor 4117 is provided on the lower section pipe 4103. A lower baffle is provided at the output end of the lower baffle first motor 4117. The rotation of the lower baffle first motor 4117 controls the closing and opening of the lower section pipe 4103. A lower baffle second motor 4118 is also provided on the lower section pipe 4103. Its function is the same as that of the lower baffle first motor 4117. It is used as a backup. An insect storage box 4102 is provided at the outlet of the lower section pipe 4103.

[0039] The RGB camera 2000 is used to perform rapid detection and segmentation of diseased or insect-infested areas, and can make preliminary identification of diseases and pests such as rubber bark beetles and rubber tree anthracnose.

[0040] The depth camera 2002 is used to estimate the target tree trunk normal and distance. Based on this, the electronic control unit 7000 drives the robotic arm 3000 and the moving guide rail 1001 to work together to maintain the working distance of the hyperspectral camera 2003 in the range of 0.3 to 0.6 m.

[0041] Based on the preliminary information collected by the RGB camera 2000, the hyperspectral camera 2003 further collects spectral image features of multiple points on the problem rubber trees and combines them with a deep learning model to accurately identify rubber tree diseases and pests.

[0042] The pest detection module 4000 is used to conduct fixed-point monitoring of rubber bark beetle infestation in a certain area of ​​the rubber plantation to obtain the pest population density. When more than 3 rubber bark beetles are found in a certain area of ​​the rubber plantation, the pest detection module 4000 will stay and conduct fixed-point monitoring and analysis.

[0043] The panoramic camera 2001 is used to generate a global environment map and provide inspection path records, which enables the electronic control unit 7000 to perform time-series retesting of the same target tree.

[0044] The electronic control unit 7000 is equipped with a time synchronization module to align the timestamps of the RGB camera 2000, panoramic camera 2001, depth camera 2002, hyperspectral camera 2003 and navigation module 6000, with a time error of no more than ±2 ms.

[0045] Reference Figure 5 The control method for the intelligent inspection device of natural rubber plantation in hilly and mountainous areas provided by the present invention includes the following steps:

[0046] Step 1: Generate an inspection route and target natural rubber tree database based on the work area boundary and road / rubber tapping road information, and set the stopping posture for each target tree;

[0047] Step 2: The navigation module 6000 uses GNSS RTK, IMU and visual odometry to locate the current position and the position of the target tree. The panoramic camera 2001 collects information on the rubber plantation operation environment where the device is currently located. The electronic control unit 7000 plans the walking path of the quadrupedal walking platform 5000 and performs gait control and obstacle avoidance in real time.

[0048] Step 3: Use depth camera 2002 to estimate the trunk normal and distance, and drive robotic arm 3000 and moving guide rail 1001 to precisely move hyperspectral camera 2003 to the target area;

[0049] Step 4: Use RGB camera 2000 to quickly screen for pests and diseases. If the actual value of the first confidence level P1 is greater than the first confidence level threshold K1, then trigger hyperspectral camera 2003 to perform line scan acquisition and complete "white / dark" reference calibration. If the actual value of the first confidence level P1 is less than the first confidence level threshold K1, then switch to the next tree.

[0050] Step 5: Preprocess and extract features from the spectral data acquired by the hyperspectral camera 2003. Use a classification model to output the pest type and the actual value of the second confidence level P2. When the actual value of the second confidence level P2 detected by the classification model is greater than the second confidence threshold K2, the electronic control unit 7000 is triggered to control the quadrupedal walking platform 5000 to switch to the stationary mode. If the actual value of the second confidence level P2 is less than the second confidence threshold K2, switch to the next tree.

[0051] Step 6: The insect infestation detection module 4000 starts working, collecting insect infestation data and environmental parameters in the rubber plantation, and obtaining the insect population density and trend score per unit time;

[0052] Step 7: Integrate the first confidence level P1, the second confidence level P2, the insect population density, and environmental parameters to obtain a comprehensive risk value R, classify the risk level, and generate an inspection report that includes geographical location, tree number, and image / spectral summary;

[0053] Step 8: When the comprehensive risk value R reaches the warning threshold, output handling suggestions and decide whether to continue patrolling or return to base / recharge based on the task completion status or battery level. If the battery level is sufficient, switch to the next rubber plantation area for patrolling. If the task is completed or the battery level is low, return to base or recharge. If the comprehensive risk value R does not reach the warning threshold, switch to the next rubber plantation area for patrolling.

[0054] Furthermore, in step two, in areas with weak RTK coverage, visual positioning using manual markers is used to repeatedly visit and locate the target tree with an error not exceeding 5 cm.

[0055] Furthermore, in step five, the spectral preprocessing uses Savitzky-Golay smoothing and standard normal variable transformation, the feature extraction uses competitive adaptive reweighted sampling (CARS), ReliefF algorithm and stepwise regression projection (SPA) to select key bands, and the classification model uses partial least squares discriminant analysis (PLS-DA), support vector machine (SVM) and one-dimensional convolutional neural network (1D-CNN).

[0056] Furthermore, in step seven, the fusion computing adopts a Bayesian fusion or learning-based weighted strategy, and the risk level is divided into at least four levels based on the numerical range of R: Level I Normal, Level II Attention, Level III Early Warning, and Level IV Response. The inspection report is uploaded to the cloud platform via 5G, Wi-Fi, and LoRa communication links to generate a spatiotemporal heat map of pests and diseases.

[0057] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. Any equivalent modifications or variations made by those skilled in the art based on the content disclosed in the present invention should be included within the scope of protection set forth in the claims.

Claims

1. A smart inspection method for natural rubber plantations in hilly and mountainous areas, characterized in that, Includes the following steps: Step 1: Generate an inspection route and target natural rubber tree database based on the work area boundary and road information, and set the stopping pose for each target tree; Step 2: The navigation module (6000) uses GNSS RTK, IMU and visual odometry to locate the current position and the position of the target tree. The panoramic camera (2001) collects information on the rubber plantation operation environment where the device is currently located. The electronic control unit (7000) plans the walking path of the quadrupedal walking platform (5000) and performs gait control and obstacle avoidance in real time. Step 3: Use the depth camera (2002) to estimate the trunk normal and distance, drive the robotic arm (3000) and the moving guide rail (1001) so that the hyperspectral camera (2003) is aligned with the target area in the standing pose; Step 4: Use an RGB camera (2000) to perform a rapid initial screening of pests and diseases. If the actual value of the first confidence level P1 is greater than the first confidence level threshold K1, then trigger the hyperspectral camera (2003) to perform line scan acquisition and complete the "white and dark" reference calibration. If the actual value of the first confidence level P1 is less than the first confidence level threshold K1, then switch to the next tree. Step 5: Preprocess and extract features from the spectral data collected by the hyperspectral camera (2003), and use a classification model to output the pest type and the actual value of the second confidence level P2. When the actual value of the second confidence level P2 detected by the classification model is greater than the second confidence level threshold K2, the electronic control unit (7000) is triggered to control the quadrupedal walking platform (5000) to switch to the stationary mode. If the actual value of the second confidence level P2 is less than the second confidence level threshold K2, then switch to the next tree. Step 6: When the quadrupedal walking platform is in stationary mode, the insect detection module (4000) starts working, collecting rubber plantation insect data and environmental parameters, and obtaining the insect population density and trend score per unit time. Step 7: Integrate the first confidence level P1, the second confidence level P2, the insect population density, and environmental parameters to obtain a comprehensive risk value R, classify the risk level, and generate an inspection report that includes geographical location, tree number, and image spectral summary; Step 8: When the comprehensive risk value R reaches the warning threshold, output handling suggestions and decide whether to continue patrolling or return to base for refueling based on the task completion status or battery level. If the comprehensive risk value R does not reach the warning threshold, continue patrolling.

2. The intelligent inspection method for natural rubber plantations in hilly and mountainous areas according to claim 1, characterized in that: In step two, in areas with weak RTK coverage, visual positioning using manual markers is used to repeatedly visit and locate the target tree, with a positioning error not exceeding 5 cm. In step five, spectral preprocessing employs Savitzky-Golay smoothing and standard normal variable transformation; feature extraction uses competitive adaptive reweighted sampling, ReliefF algorithm, and stepwise regression projection to select key bands; and the classification model uses partial least squares discriminant analysis, support vector machine, and one-dimensional convolutional neural network. In step seven, fusion calculation uses Bayesian fusion or learning-based weighting strategies, and based on the numerical range of R, the risk level is divided into at least four levels: Level I Normal, Level II Attention, Level III Warning, and Level IV Response. Inspection reports are uploaded to the cloud platform via 5G, Wi-Fi, and LoRa communication links, generating a spatiotemporal heat map of pests and diseases.

3. A smart inspection device for natural rubber plantations in hilly and mountainous areas, characterized in that, A smart inspection method for natural rubber plantations in hilly areas, as described in any one of claims 1-2, includes a quadrupedal walking platform (5000). The quadrupedal walking platform (5000) is equipped with a data acquisition module and an insect detection module (4000). The data acquisition module includes a robotic arm (3000), with an RGB camera (2000) and a data acquisition platform (1000) at its end. The data acquisition platform (1000) includes a movable guide rail (1001), which is a spirally ascending semi-circular structure at the end of the robotic arm (3000). A movable guide slide rail (1002) is provided on the movable guide rail (1001), and a track motor (1104) is slidably mounted on the movable guide slide rail (1002). A hyperspectral camera (2003) is provided. The insect detection module (4000) includes a support frame (4116), which is set on a quadrupedal walking platform (5000). A panoramic camera (2001) is provided on the top of the support frame (4116). The support frame (4116) is provided with an insect trapping and killing component, a collection component, a detection component, and an insect storage box (4102) from top to bottom. A depth camera (2002) is provided at the front end of the quadrupedal walking platform (5000). A navigation module (6000) and an electronic control unit (7000) are provided inside the quadrupedal walking platform (5000). The electronic control unit (7000) is electrically connected to the data acquisition platform (1000), the panoramic camera (2001), the robotic arm (3000), the insect detection module (4000), and the navigation module (6000).

4. The intelligent inspection device for natural rubber plantations in hilly and mountainous areas according to claim 3, characterized in that: The hyperspectral camera (2003) is equipped with a light shield (1107) and a supplementary light (1106) on its outside.

5. The intelligent inspection device for natural rubber plantations in hilly and mountainous areas according to claim 4, characterized in that: The hyperspectral camera (2003) is provided with a floating track (1108), which is arranged vertically. The track motor (1104) is provided with a floating motor (1109), which drives the hyperspectral camera (2003) to slide on the floating track (1108).

6. The intelligent inspection device for natural rubber plantations in hilly and mountainous areas according to claim 3, characterized in that: The robotic arm (3000) includes a robotic arm mounting base (3109), which is mounted on a quadrupedal walking platform (5000). A first connecting gear (3108) is rotatably mounted on the robotic arm mounting base (3109). A first motor (3107) is mounted on the robotic arm mounting base (3109) to drive the first connecting gear (3108) to rotate. A first arm segment (3106) is mounted on the first connecting gear (3108). A second connecting gear (3111) is rotatably mounted at the end of the first arm segment (3106). A second arm segment (3104) is mounted on the second connecting gear (3111). A second arm segment (3104) is mounted on the second arm segment (3104). Two motors (3105) are used to drive the second connecting gear (3111) to rotate. The end of the second arm (3104) is rotatably provided with a third connecting gear (3103). The third connecting gear (3103) is provided with a third arm (3101). The third arm (3101) is provided with a third motor (3102) for driving the third connecting gear (3103) to rotate. The third arm (3101) is provided with an RGB camera (2000). The third arm (3101) is provided with a telescopic push rod (3112). The telescopic push rod (3112) is provided with a telescopic arm (3113). The data acquisition platform (1000) is provided on the telescopic arm (3113).

7. The intelligent inspection device for natural rubber plantations in hilly and mountainous areas according to claim 3, characterized in that: The insect trapping and killing assembly includes a trapping frame (4110), a trapping lamp (4112) on the trapping frame (4110), a dosing box (4111) above the trapping lamp (4112), an insect collecting dish (4109) below the trapping frame (4110), the insect collecting dish (4109) being funnel-shaped, an upper section pipe (4108) below the insect collecting dish (4109), an upper baffle (4113) slidingly mounted on a support frame (4116) below the upper section pipe (4108), a pipe push rod (4114) on the support frame (4116), and a corresponding position below the upper baffle (4113) corresponding to the upper section pipe (4108). A middle section pipe (4107) is provided with a negative pressure motor (4115). A collection box (4104) is provided on the support frame (4116). A collection camera (4106) is provided on the top of the collection box (4104). The outlet of the middle section pipe (4107) leads to the collection box (4104). A lower section pipe (4103) is provided below the collection box (4104). A lower baffle first motor (3107) (4117) is provided on the lower section pipe (4103). A lower baffle is provided at the output end of the lower baffle first motor (3107) (4117). An insect storage box (4102) is provided at the outlet of the lower section pipe (4103).

8. The intelligent inspection device for natural rubber plantations in hilly and mountainous areas according to claim 3, characterized in that: The navigation module (6000) includes a GNSS RTK receiver, an inertial measurement unit and a visual odometry unit, and the electronic control unit (7000) performs multi-sensor fusion positioning. The navigation module (6000) also includes a visual positioning unit that recognizes artificial markers to achieve accurate alignment for repeated visits in areas where RTK signals are blocked. The quadrupedal walking platform (5000) can walk continuously on terrain with a slope of 25° to 35° and pass through terrain with a slope of 40° for a short period of time.

9. The intelligent inspection device for natural rubber plantations in hilly and mountainous areas according to claim 3, characterized in that: The electronic control unit (7000) includes an industrial computer and a wireless communication module, which are used to perform fusion analysis on the above data and upload it to the cloud for management. The industrial computer has edge computing capabilities.

Images