Intelligent image recognition disinfection robot and intelligent control system

Abstract

The invention relates to the technical field of image recognition data processing, in particular to an intelligent image recognition disinfecting and killing robot and an intelligent control system.The intelligent image recognition disinfecting and killing robot comprises a robot head, a camera, a first robot hand, a central processing unit, the intelligent control system of the intelligent image recognition disinfecting and killing robot, a robot shell, a robot walking device and a chemical spraying device; according to the robot, pesticide spraying is accurately controlled through an intelligent control system of the intelligent image recognition disinfection and killing robot according to the humidity of the mountain environment, the soil fertility degree and the harm degree of pest behaviors to crops; meanwhile, the spraying angle of the pesticide is controlled according to the growth state of the crops, uneven pesticide spraying on the crops is avoided, and the purpose of accurately controlling pesticide spraying according to the pest behaviors and the growth state of the crops is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition data processing, and in particular to an intelligent image recognition and disinfection robot and an intelligent control system. Background Art

[0002] In mountainous areas, pest control is crucial for ensuring crop yield and quality. However, existing pest control methods have numerous drawbacks and are unable to meet the demands of complex mountainous terrain and diverse crop cultivation. Traditional pest control methods rely heavily on manual labor, which presents significant challenges in mountainous areas. Mountainous terrain is complex, with large undulating terrain and scattered, vast farmlands. While widely used in plain areas, large agricultural machinery for pest control is not suitable for mountainous areas. These devices are bulky and difficult to navigate through narrow, rugged mountain roads and small plots of farmland. Their large turning radius makes them inflexible and unable to adapt to the complex shapes and layouts of mountain farmland. Manually carrying pesticide sprayers for pest control is not only labor-intensive and inefficient, but also requires workers to navigate rugged mountain roads and fields, exposing them to numerous safety risks.

[0003] China Publication No. CN113361429B discloses a method for analyzing insect movement behavior, specifically a method for analyzing the movement behavior of stored-grain pests. This method includes the following steps: 1) generating a target block diagram; 2) determining the number of pests; 3) detecting their behavior; and 4) generating a statistical table and control strategy. However, this invention fails to integrate crop growth status with pest behavior, making it difficult to achieve precise pest control. Summary of the Invention

[0004] To this end, the present invention provides an intelligent image recognition pest control robot and an intelligent control system to overcome the problem in the prior art that it is difficult to accurately control pest control based on pest behavior and crop growth status, resulting in low rationality, efficiency and accuracy of pest control.

[0005] To achieve the above objectives, the present invention provides an intelligent image recognition and disinfection robot and an intelligent control system, comprising:

[0006] The robot head is connected to the robot shell, central processing unit, and camera respectively, and is used for 360° rotation, so that the camera can obtain a 360° working environment;

[0007] The camera is connected to the central processing unit and the robot head respectively to obtain information about the surrounding environment;

[0008] The first robot hand is connected to the robot housing, the chemical spray device, and the central processing unit, and is used to fix the chemical spray device;

[0009] The central processing unit is connected to each component of the robot and is used to process all relevant data obtained by the robot;

[0010] The intelligent control system of the intelligent image recognition disinfection robot is connected to the central processing unit and is used to control the robot based on the data obtained by the robot;

[0011] The robot housing is connected to the first robot hand, the second robot hand, and the robot walking device, respectively, and is used to protect the internal components;

[0012] A robot walking device, connected to the robot housing and the central processing unit, respectively, for making the robot walk;

[0013] a chemical spray device, connected to the central processing unit and the second robot arm, respectively, for killing pests with pesticides;

[0014] A physical pest control board is connected to the central processing unit and the first robot arm respectively, and is used for physically killing pests;

[0015] The second robot hand is connected to the robot shell, the physical insect-killing board and the central processing unit respectively, and is used for fixing the physical insect-killing board.

[0016] The present invention also provides an intelligent control system for an intelligent image recognition and disinfection robot, the system comprising:

[0017] Data acquisition module, used to acquire mountain environment data, pesticide spraying data, crop growth data and pest data;

[0018] A terrain simulation module is used to construct a mountain terrain model based on mountain environment data and to calibrate the construction process of the mountain terrain model;

[0019] A crop growth monitoring module is used to construct a crop growth status model based on crop growth data, obtain crop growth status based on the crop growth status model, obtain the planting area type based on the mountainous terrain model, and adjust the construction process of the crop growth status model based on the planting area type;

[0020] A pest behavior recognition module is used to input pest data into a pest behavior classification probability model to obtain pest behavior, obtain the degree of damage to crops based on the pest behavior, and calibrate the adjustment of the construction process of the crop growth status model based on the degree of damage to crops;

[0021] The intelligent pest control module is used to input pesticide spraying data into the pesticide pest simulation model to obtain the pesticide spraying amount and to correct the pesticide pest simulation model. It is also used to adjust the correction of the pesticide pest simulation model according to the degree of damage to the crops. It is also used to adjust the training process of the crop growth status model according to the pesticide spraying data and to adjust the correction process of the mountainous terrain model.

[0022] Furthermore, the terrain simulation module constructs a mountain terrain model by inputting mountain environment data into a mountain environment simulation model.

[0023] Furthermore, the terrain simulation module generates a terrain simulation result according to the first coordinate of the i-th control point of the three-dimensional point cloud data in the mountain environment data. The second coordinate of the i-th control point in the image data The registration accuracy error RMSE is calculated based on the number of control points ek, and the registration accuracy error RMSE is compared with the preset registration accuracy error RMSE0. The registration accuracy is judged according to the comparison results, and the construction process of the mountain environment simulation model is corrected according to the judgment results.

[0024] Furthermore, the crop growth monitoring module generates a crop growth signal according to each data point x of the independent variable X in the crop growth data. t , standard deviation S of the independent variable X X , the mean of the independent variable X Each data point y at crop growth stage Y t , the mean value of Y in the crop growth stage Standard deviation S of crop growth stage Y Y The Pearson correlation coefficient r is calculated based on the number of data n, and the Pearson correlation coefficient r is compared with the preset Pearson correlation coefficient r0. The correlation between the independent variable and the crop growth stage is judged according to the comparison result, and the independent variable and the crop growth stage are divided into feature learning data according to the judgment result, and the crop growth status model is constructed using the feature learning data.

[0025] Furthermore, the crop growth monitoring module divides the crop planting areas according to the mountain environment data to obtain the planting area types, which include general flat areas, remote slope areas and highly sensitive areas. The crop growth data acquisition situation is judged according to the planting area type, and the construction process of the crop growth status model is adjusted according to the judgment result.

[0026] Furthermore, the pest behavior recognition module inputs the pest video data and crop growth data in the pest data into the LSTM model, and obtains the probability P of the occurrence of the pest behavior corresponding to each node output by the LSTM model. LSTM, P LSTM =[P 1 LSTM ,P 2 LSTM ,...P f LSTM ], f is the number of categories of pest behavior;

[0027] The pest behavior recognition module inputs pest data into the STGCN model and obtains the probability distribution P of pest behavior in the spatial dimension output by the STGCN model. STGCN , P STGCN =[P 1 STGCN ,P 2 STGCN ,...P f STGCN ];

[0028] The pest behavior recognition module is based on the probability P of the pest behavior corresponding to each node. LSTM and the probability distribution P of pest behavior in spatial dimension STGCN For the fused probability vector P fusion Calculate and get the fused probability vector P fusion , w1 is the weight coefficient of the probability of pest behavior corresponding to each node, w2 is the weight coefficient of the probability distribution of pest behavior in the spatial dimension, and the fused probability vector P fusion Input the pest behavior classification probability model, obtain the pest behavior z output by the pest behavior classification probability model, compare the pest behavior with the preset behavior in the pest behavior expert database, and obtain the damage degree value C of the crop based on the comparison result.

[0029] Furthermore, the pest behavior identification module compares the crop damage level value C with the preset crop damage level value C0, 0.26≤C0≤0.31, and judges the pest behavior damage situation based on the comparison result, and calibrates the adjustment of the construction process of the crop growth status model based on the judgment result.

[0030] Furthermore, the intelligent pest control module inputs the required pest mortality rate into the pesticide pest simulation model to obtain the pesticide spraying amount J output by the pesticide pest simulation model.

[0031] Furthermore, the intelligent pest control module compares the crop damage level value C obtained in the pest behavior recognition module with the preset crop damage level value C0, judges the pest behavior damage situation based on the comparison result, and adjusts the pest density judgment process based on the judgment result.

[0032] Compared with the prior art, the beneficial effect of the present invention is that the system acquires mountain environment data, pesticide spraying data, crop growth data and pest data through the data acquisition module, so that the system can timely analyze the pest situation and control the spraying of pesticides. The system also constructs a model of the mountain environment through the terrain simulation module, which facilitates the robot to divide the mountain environment into regions and improves the accuracy of pesticide spraying in each region. The system also monitors the growth of crops through the crop growth monitoring module, so as to control the amount of pesticide spraying according to the growth of crops, avoid excessive pesticide spraying causing damage to crop growth, and improve the control accuracy of pesticide spraying amount. The system also uses the pest behavior recognition module to identify pests. The system can identify insect behavior and judge the degree of harm caused by pest behavior to crops in order to control the amount of pesticide spraying and realize precise control of pesticide spraying. The system also monitors the humidity and soil fertility of the mountain environment through the intelligent pest control module, and accurately controls the spraying of pesticides according to the humidity, soil fertility and the degree of harm caused by pest behavior to crops in the mountain environment. At the same time, the spraying angle of pesticides is controlled according to the growth status of crops to avoid uneven spraying of pesticides on crops, so as to achieve the purpose of precise control of pesticide spraying according to pest behavior and crop growth status, thereby realizing precise control of pest disinfecting according to pest behavior and crop growth status, thereby improving the rationality, efficiency and accuracy of pest disinfecting. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a schematic diagram of the structure of the intelligent image recognition and disinfection robot in this embodiment;

[0034] Figure 2 This is a schematic diagram of the structure of the intelligent control system of the intelligent image recognition and disinfection robot in this embodiment. DETAILED DESCRIPTION

[0035] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0036] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0037] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0038] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0039] See also Figure 1 As shown in FIG, it is a schematic diagram of the structure of the intelligent image recognition and disinfection robot of this embodiment, and the structure includes:

[0040] The robot head 1 is connected to the robot housing 5, the central processing unit 4, and the camera 2, and is used for 360-degree rotation, so that the camera 2 can obtain a 360-degree working environment;

[0041] The camera 2 is connected to the central processing unit 4 and the robot head 1 respectively, and is used to obtain the surrounding environment information;

[0042] The first robot hand 3 is connected to the robot housing 5, the chemical spray device 7, and the central processing unit 4, respectively, and is used to fix the chemical spray device;

[0043] The central processing unit 4 is connected to each component of the robot and is used to process all relevant data acquired by the robot;

[0044] The intelligent control system of the intelligent image recognition disinfection robot (not shown in the figure) is connected to the central processor 4 and is used to control the robot based on the data obtained by the robot;

[0045] The robot housing 5 is connected to the first robot hand 3, the second robot hand 9, and the robot walking device 6, respectively, and is used to protect the internal components;

[0046] The robot walking device 6 is connected to the robot housing 5 and the central processing unit 4 respectively, and is used to make the robot walk;

[0047] A chemical spray device 7, connected to the central processing unit 4 and the second robot arm 9, respectively, for killing pests with pesticides;

[0048] The physical pest killing board 8 is connected to the central processing unit 4 and the first robot hand 3 respectively, and is used for physically killing pests;

[0049] The second robot hand 9 is connected to the robot housing 5 , the physical insect-killing plate 8 and the central processing unit 4 respectively, and is used to fix the physical insect-killing plate 8 .

[0050] Specifically, the intelligent image recognition and pest control robot is used to control pests of mountain crops. The intelligent image recognition and pest control robot is equipped with a multi-sensor fusion perception system such as visual bionic detection technology, which can accurately identify the types, quantity and distribution of pests, achieve precise positioning and targeted application of pesticides, improve pest control effects, and reduce the use of pesticides; unmanned pest control humanoid robots can replace manual labor to perform long-term, high-intensity pest control operations, avoiding the dangers of workers slipping and getting injured, encountering wild animals, etc. in mountainous areas, reducing labor intensity and safety risks, and improving pest control efficiency; the robot can automatically adjust the spraying angle and flow rate according to the growth conditions of farmland crops and the distribution of pests through precise motion control and intelligent spraying system, ensuring uniform spraying of pesticides, improving pest control quality, and ensuring the healthy growth of crops.

[0051] See also Figure 2 As shown in FIG, it is a schematic diagram of the structure of the intelligent control system of the intelligent image recognition and disinfection robot of this embodiment, and the system includes:

[0052] Data acquisition module, used to acquire mountain environment data, pesticide spraying data, crop growth data and pest data;

[0053] A terrain simulation module, used to construct a mountain terrain model based on mountain environment data and to calibrate the construction process of the mountain terrain model, the terrain simulation module being connected to the data acquisition module;

[0054] A crop growth monitoring module, configured to construct a crop growth status model based on crop growth data, obtain crop growth status based on the crop growth status model, obtain planting area types based on a mountainous terrain model, and adjust the construction process of the crop growth status model based on the planting area types. The crop growth monitoring module is connected to the terrain simulation module;

[0055] A pest behavior recognition module, configured to input pest data into a pest behavior classification probability model to obtain pest behavior, determine the degree of damage to crops based on the pest behavior, and calibrate the construction process of the crop growth status model based on the degree of damage to crops. The pest behavior recognition module is connected to the crop growth monitoring module;

[0056] The intelligent pest control module is used to input pesticide spraying data into the pesticide pest simulation model to obtain the pesticide spraying amount, and to correct the pesticide pest simulation model. It is also used to adjust the correction of the pesticide pest simulation model according to the degree of damage to the crops. It is also used to adjust the training process of the crop growth status model according to the pesticide spraying data and adjust the correction process of the mountain terrain model. The intelligent pest control module is connected to the pest behavior recognition module.

[0057] Specifically, the system is applied to intelligent image recognition pest control robots. The system acquires mountain environment data, pesticide spraying data, crop growth data and pest data through a data acquisition module, so that the system can timely analyze the pest situation and control the spraying of pesticides. The system also constructs a model of the mountain environment through a terrain simulation module, which facilitates the robot to divide the mountain environment into regions and improve the accuracy of pesticide spraying in each region. The system also monitors the growth of crops through a crop growth monitoring module, so as to control the amount of pesticide spraying according to the growth of crops, avoid excessive spraying of pesticides that damages crop growth, and improve the control accuracy of pesticide spraying. The system also uses a pest behavior recognition module to The module identifies pest behavior and determines the degree of harm caused by pest behavior to crops, so as to control the amount of pesticide spraying and realize precise control of pesticide spraying. The system also monitors the humidity and soil fertility of the mountain environment through the intelligent pest control module, and accurately controls the spraying of pesticides according to the humidity, soil fertility and the degree of harm caused by pest behavior to crops in the mountain environment. At the same time, the spraying angle of pesticides is controlled according to the growth status of crops to avoid uneven spraying of pesticides on crops, so as to achieve the purpose of precise control of pesticide spraying according to pest behavior and crop growth status, thereby realizing precise control of pest disinfecting according to pest behavior and crop growth status, thereby improving the rationality, efficiency and accuracy of pest disinfecting.

[0058] Specifically, the mountain environment data includes UAV aerial survey images, three-dimensional point cloud data and ground control point data. The data acquisition module obtains UAV aerial survey images and three-dimensional point cloud data through UAV aerial survey. The data acquisition module obtains ground control point data through a total station. The pesticide spraying data includes the required pest mortality rate. The data acquisition module obtains the required pest mortality rate through user input control system requirements. The crop growth data includes meteorological data, soil data and maximum pesticide spraying amount. The data acquisition module obtains meteorological data through meteorological sensors, the data acquisition module obtains soil data through soil monitoring instruments, and the data acquisition module obtains the maximum pesticide spraying amount through pesticide instructions. The pest data includes regional pest density and pest video data. The data acquisition module obtains regional pest density and pest video data through a camera.

[0059] Specifically, the terrain simulation module constructs a mountain terrain model by inputting mountain environment data into a mountain environment simulation model.

[0060] Specifically, the mountain environment simulation model refers to a terrain analysis model that takes mountain environment data as input and a mountain terrain model as output. When constructing the mountain environment simulation model, this embodiment sets the preprocessing of the three-dimensional point cloud data and ground control point data in the mountain environment data to obtain preprocessed three-dimensional point cloud data and preprocessed control point data, and uses the nearest neighbor search algorithm to find the point closest to the preprocessed control point data in the preprocessed three-dimensional point cloud data and record it as a matching point to obtain matched three-dimensional point cloud data. The Delaunay triangulation algorithm is used to construct a triangulated network for the matched three-dimensional point cloud data to obtain a triangulated network basis, and the least squares surface fitting method is used to obtain the sum of the squares of the distance of the fitting surface. The sum of the squares of the distance of the fitting surface is used for surface fitting on the basis of the triangulated network to obtain an initial mountain terrain model, and the drone aerial survey image is used as texture information, and it is mapped to the surface of the initial mountain terrain model by bilinear interpolation to obtain a mountain terrain model.

[0061] Specifically, the Delaunay triangulation algorithm refers to an algorithm for converting a point set on a plane into a triangular mesh. This embodiment does not limit the specific implementation of the Delaunay triangulation algorithm, and only needs to satisfy the output of the mountainous terrain model. The nearest neighbor search algorithm refers to an algorithm for finding the data point closest to the query point in a given data set. This embodiment does not limit the specific implementation of the nearest neighbor search algorithm, and only needs to satisfy the output of the mountainous terrain model. The least squares surface fitting method refers to a mathematical method for finding a surface that best fits the given data points. Its core is to determine the parameters of the surface by minimizing the sum of squared errors between the observed data points and the fitted surface. This embodiment does not limit the specific implementation of the least squares surface fitting method, and only needs to satisfy the output of the mountainous terrain model. The bilinear interpolation method refers to an interpolation algorithm for estimating the value of an unknown point when the discrete data points are known. This embodiment does not limit the specific implementation of the bilinear interpolation method, and only needs to satisfy the output of the mountainous terrain model.

[0062] Specifically, the terrain simulation module is based on the first coordinate of the i-th control point of the three-dimensional point cloud data in the mountain environment data. The second coordinate of the i-th control point in the image data The registration accuracy error RMSE is calculated by the number of control points ek, i = 1, 2, 3...se, se is the order of the control points, and the setting The registration accuracy error RMSE is compared with the preset registration accuracy error RMSE0, 0.23≤RMSE0≤0.31, and the registration accuracy is judged according to the comparison results. The construction process of the mountain environment simulation model is corrected according to the judgment results, where:

[0063] When RMSE≤RMSE0, the terrain simulation module determines that the registration accuracy is qualified and does not correct the construction process of the mountain environment simulation model;

[0064] When RMSE>RMSE0, the terrain simulation module determines that the registration accuracy is unqualified and corrects the construction process of the mountain environment simulation model. The correction scheme is to oversample the UAV aerial survey image to obtain the oversampled UAV aerial survey image, and replace the UAV aerial survey image with the oversampled UAV aerial survey image, and use the oversampled UAV aerial survey image as texture information until RMSE≤RMSE0.

[0065] Specifically, the three-dimensional point cloud data of mountainous terrain refers to a spatial data form used to represent mountainous terrain, consisting of a large number of discrete three-dimensional coordinate points, each of which contains its position information in three-dimensional space. The first coordinate of the i-th control point of the three-dimensional point cloud data refers to a reference coordinate in a coordinate system used to determine the positions of the remaining points in space. The second coordinate corresponding to the i-th control point in the image data refers to the coordinate in the image data corresponding to the first coordinate of the i-th control point in the three-dimensional point cloud data. The preset registration accuracy error refers to a preset value used to judge the registration accuracy. The registration accuracy refers to the registration accuracy of the first coordinate of the i-th control point in the three-dimensional point cloud data with the second coordinate of the i-th control point in the image data. The registration accuracy includes a registration accuracy of acceptable accuracy and an unacceptable accuracy. Supersampling refers to an image acquisition technology that adds additional sampling points to the original sampling points to obtain more image information. This embodiment does not limit the specific implementation of supersampling. Those skilled in the art can configure it according to actual needs, as long as it meets the requirements for calibrating the mountainous environment simulation model.

[0066] Specifically, the crop growth monitoring module is based on each data point x of the independent variable X in the crop growth data. t , standard deviation S of the independent variable X X , the mean of the independent variable X Each data point y at crop growth stage Y t , the mean value of Y in the crop growth stage Standard deviation S of crop growth stage Y YCalculate the Pearson correlation coefficient r with the number of data n, t = 1, 2, 3...m, m is the order of the data points, set The Pearson correlation coefficient r is compared with the preset Pearson correlation coefficient r0, 0.5≤r0≤7. The correlation between the independent variable and the crop growth stage is judged based on the comparison result, and the independent variable and the crop growth stage are divided into feature learning data based on the judgment result, where:

[0067] When r<r0, the crop growth monitoring module determines that the correlation between the independent variable and the crop growth stage is insufficient, and does not classify the independent variable and the crop growth stage as feature learning data;

[0068] When r≥r0, the crop growth monitoring module determines that the correlation between the independent variable and the crop growth stage is sufficient, divides the independent variable and the crop growth stage into feature learning data, and constructs a crop growth status model using the feature learning data.

[0069] Specifically, the independent variable refers to an influencing factor variable that is actively measured and controlled and has an impact on the growth stage of crops. The independent variables include meteorological data and soil data. The meteorological data refers to various types of data collected through scientific instruments, equipment and observation means during the growth of crops, reflecting the physical state of the atmosphere, weather phenomena and their changing processes. The meteorological data includes temperature, humidity, precipitation, light, wind speed and air pressure. The soil data refers to quantitative and qualitative information obtained through scientific means during the growth of crops, reflecting the physical, chemical and biological properties of the soil and their spatial distribution. The soil data includes pH value, nutrient content, conductivity, soil temperature and soil moisture content. The preset Pearson correlation coefficient refers to a preset value used to judge the correlation between the independent variable and the growth stage of crops. The correlation between the independent variable and the growth stage of crops refers to the expression of the correlation between the independent variable and the growth stage of crops. The correlation between the independent variable and the growth stage of crops includes insufficient correlation between the independent variable and the growth stage of crops and sufficient correlation between the independent variable and the growth stage of crops.

[0070] Specifically, the crop growth state model refers to a machine learning model that takes independent variables as input and uses the crop growth stage as output. The crop growth monitoring module constructs the crop growth state model using a crop growth state model construction method, and the crop growth state model construction method includes:

[0071] The feature learning data is divided into a 70% feature training set, a 20% feature validation set, and a 10% feature test set. The feature training set is input into the decision tree model to train the decision tree model, and the feature validation set is input into the trained decision tree model. The hyperparameters of the trained decision tree model are iteratively optimized, and the feature test set is input into the iteratively optimized decision tree model to perform feature testing on the iteratively optimized decision tree model to obtain the feature test results. The total number of samples in the feature test set is set to f0, the number of correct feature test samples is set to f, and the feature test accuracy is set to F, F=f / f0. The feature test accuracy F is compared with the preset feature test accuracy F0, F0=82%. The training compliance of the iteratively optimized decision tree model is judged based on the comparison results, and the judgment result is output, where:

[0072] When F≥F0, the crop growth monitoring module determines that the iteratively optimized decision tree model has met the training standards, outputs the iteratively optimized decision tree model as a crop growth status model, and provides the crop growth status model to the crop growth monitoring module;

[0073] When F<F0, the crop growth monitoring module determines that the iteratively optimized decision tree model training does not meet the standards, updates the feature learning data, obtains updated feature learning data, and trains the decision tree model according to the updated feature learning data, iteratively optimizes the hyperparameters, and performs analysis and testing until the decision tree model training meets the standards.

[0074] Specifically, the feature training set refers to a data set used to train the crop growth status model, the feature verification set refers to a data set used to verify the trained crop growth status model, the feature test set refers to a data set used to test the verified crop growth status model, the preset feature test accuracy refers to a preset value used to judge whether the training of the iteratively optimized decision tree model has met the standards, and the updated feature learning data refers to a data set consisting of newly acquired independent variables and the crop growth stages corresponding to the independent variables added to the feature learning data.

[0075] Specifically, the crop growth monitoring module determines the correlation between the independent variables and the crop growth stage, eliminates the independent variables with insufficient correlation, and retains the independent variables with sufficient correlation, thereby constructing a crop growth status model and improving the prediction accuracy of the crop growth status model.

[0076] Specifically, the crop growth monitoring module inputs real-time independent variables in the crop growth data into the crop growth state model and outputs the crop growth state, which includes the seedling stage, the mature stage and the fruiting stage.

[0077] Specifically, the crop growth monitoring module divides the crop planting areas according to the mountain environment data to obtain the planting area types, which include general flat areas, remote slope areas, and highly sensitive areas. The crop growth data acquisition situation is judged according to the planting area type, and the construction process of the crop growth status model is adjusted according to the judgment result, wherein:

[0078] When the crop planting area is a generally flat area, the crop growth monitoring module determines that the crop growth data acquisition is normal, and does not adjust the construction process of the crop growth status model;

[0079] When the crop planting area is a remote slope area, the crop growth monitoring module determines that the crop growth data acquisition situation is difficult to obtain, adjusts the construction process of the crop growth status model, and adjusts the preset feature test accuracy F0 by the difficulty acquisition adjustment coefficient kv=0.78 to obtain the first adjusted preset feature test accuracy F0v, sets F0v=F0×kv, and replaces the preset feature test accuracy F0 with the first adjusted preset feature test accuracy F0v, and re-compares the feature test accuracy F;

[0080] When the crop planting area is a highly sensitive area, the crop growth monitoring module determines that the crop growth data acquisition is accurately acquired, adjusts the construction process of the crop growth status model, and adjusts the preset feature test accuracy F0 by accurately acquiring the adjustment coefficient kj=1.01 to obtain the second adjusted preset feature test accuracy F0j, sets F0j=F0×kj, and replaces the preset feature test accuracy F0 with the second adjusted preset feature test accuracy F0j, and re-compares the feature test accuracy F.

[0081] Specifically, the planting area type refers to the classification of crop planting areas according to the regional terrain. The general flat area refers to the planting area with flat terrain, evenly distributed and easy-to-access resources, and a stable ecosystem. The remote slope area refers to the planting area with sloping terrain, with a fragile ecosystem that is prone to soil erosion. The highly sensitive area refers to the planting area with sunken terrain, with a fragile ecosystem and poor self-repairing ability, which is prone to chain reactions leading to ecosystem collapse. The crop growth data acquisition situation refers to the requirements for obtaining crop growth data, and the crop growth data acquisition situation includes whether the crop growth data acquisition situation is normal, agricultural, or The crop growth data acquisition situation is difficult to obtain and the crop growth data acquisition situation is precise acquisition. The crop growth data acquisition situation of normal acquisition means that the crop growth data in the planting area is simple to obtain, and its acquisition requirement is normal acquisition. The crop growth data acquisition situation of difficult acquisition means that the crop growth data in the planting area is difficult to obtain, and its acquisition requirement is high, which is difficult acquisition. The crop growth data acquisition situation of precise acquisition means that the crop growth data in the planting area is normally difficult to obtain, but the ecosystem in the area is changeable, and the crop growth data needs to be obtained carefully to judge the growth status of crops in the area, and its acquisition requirement is precise acquisition.

[0082] Specifically, the pest behavior recognition module inputs the pest video data and crop growth data in the pest data into the LSTM model, and obtains the probability P of the pest behavior corresponding to each node output by the LSTM model. LSTM , P LSTM =[P 1 LSTM ,P 2 LSTM ,...P f LSTM ], f is the number of pest behavior categories.

[0083] Specifically, the P LSTM It means that P 1 LSTM ,P 2 LSTM ,...P f LSTM The set of P 1 LSTM Refers to the probability of occurrence of the first type of pest behavior, the P 2 LSTM Refers to the probability of occurrence of the second type of pest behavior, the P f LSTMrefers to the probability of occurrence of the fth type of pest behavior. The LSTM model refers to a recurrent neural network model that takes pest video data and crop growth data in the pest data as input and outputs the probability of occurrence of pest behavior corresponding to each node. This embodiment does not limit the specific construction method of the LSTM model. Those skilled in the art can configure it according to actual needs, and only needs to output the probability of occurrence of pest behavior corresponding to each node. For example, three layers of LSTM units are configured, and the number of nodes in each layer of LSTM units is adjusted according to the complexity of the data and model performance. In the output layer of the LSTM network, the number of nodes is set according to the number of pest behavior categories, each node corresponds to a behavior category, and the probability of occurrence of the pest behavior corresponding to each node is output. The full name of the LSTM model is Long Short-Term Memory Network model, and its Chinese name is Long Short-Term Memory Network model. The probability of occurrence of pest behavior corresponding to each node refers to the probability of the pest behavior corresponding to the node occurring at each node, and the number of pest behavior categories refers to the total number of nodes corresponding to the pest behavior.

[0084] Specifically, the pest behavior recognition module inputs pest data into the STGCN model and obtains the probability distribution P of pest behavior in the spatial dimension output by the STGCN model. STGCN ,[P 1 STGCN ,P 2 STGCN ,...,P f STGCN ].

[0085] Specifically, the P STGCN It means that P 1 STGCN ,P 2 STGCN ,...,P f STGCN The set of P 1 STGC Refers to the probability distribution of the first type of pest behavior in the spatial dimension, the P 2 STGCN Refers to the probability distribution of the second type of pest behavior in the spatial dimension, the P f STGCNIt refers to the probability distribution of the f-th type of pest behavior in the spatial dimension. The STGCN model refers to a convolutional neural network model that takes the pest video data in the pest data as input and the probability distribution of the pest behavior in the spatial dimension as output. This embodiment does not limit the specific construction method of the STGCN model. Those skilled in the art can set it by themselves according to actual needs. It only needs to satisfy the output of the probability distribution of the pest behavior in the spatial dimension. For example, a corresponding graph structure is constructed for each frame image in the pest video data. The STGCN network is designed, including a graph convolution layer and a temporal convolution layer. The graph convolution layer is used to extract the spatial characteristics of the pests. By performing convolution operations on the graph structure, the spatial distribution patterns, aggregation patterns and interaction relationships between the pests are mined. The temporal convolution layer is used to capture the changes in the spatial characteristics of the pests in the temporal dimension, and the spatial features of adjacent frames are fused and dynamically modeled. In the output layer of the STGCN network, the number of nodes is set according to the number of classifications of the pest behavior, and the probability distribution of the pest behavior in the spatial dimension is output. The probability distribution of the pest behavior in the spatial dimension refers to the distribution pattern of the probability of the pest exhibiting a certain behavior within the spatial range in the spatial position.

[0086] Specifically, the pest behavior recognition module is based on the probability P of the pest behavior corresponding to each node. LSTM and the probability distribution P of pest behavior in spatial dimension STGCN For the fused probability vector P fusion Perform calculations and set P fusion =w1×P LSTM +w2×P STGCN , w1=0.5,w2=0.5, and get the fused probability vector P fusion , w1 is the weight coefficient of the probability of pest behavior corresponding to each node, w2 is the weight coefficient of the probability distribution of pest behavior in the spatial dimension, and the fused probability vector P fusion Input the pest behavior classification probability model, obtain the pest behavior z output by the pest behavior classification probability model, compare the pest behavior with the preset behavior in the pest behavior expert database, and obtain the crop damage degree value C based on the comparison result, where:

[0087] When the pest behavior is consistent with the preset behavior in the pest behavior expert database, the preset crop damage level value C0 corresponding to the preset behavior in the pest behavior expert database is used as the crop damage level value C;

[0088] When the pest behavior is inconsistent with the preset behavior in the pest behavior expert database, the pest behavior is pushed to the expert terminal, and the preset crop damage level value C0 input by the expert terminal is obtained and used as the crop damage level value C.

[0089] Specifically, the pest classification probability model refers to a machine learning model with a fused probability vector as output and pest behavior as output. This embodiment does not limit the construction method of the pest classification probability model. Those skilled in the art can set it up according to actual needs, and only need to meet the output of pest behavior. For example, 75% of the historical pest probability data set is divided into a probability training set to train the pest classification probability model, 15% of the historical pest probability data set is divided into a probability verification set to verify the trained pest classification probability model, and 10% of the historical pest probability data set is divided into a probability test set to test the verified pest classification probability model until the test accuracy of the pest classification probability model reaches 92%. The historical pest probability data set refers to a set used to train the pest classification probability model. The training data set includes the historical pest probability data set including the historical fused probability vector and the historical pest behavior corresponding to the historical fused probability vector, the pest behavior expert database refers to a database specifically storing pest behaviors and the damage degree values of crops corresponding to the pest behaviors, the preset behavior refers to the preset value of pest behavior corresponding to the damage degree value of crops in the pest behavior expert database, the pest behavior being consistent with the preset behavior in the pest behavior expert database means that the character comparison of the pest behavior and the preset behavior in the pest behavior expert database is consistent, the pest behavior being inconsistent with the preset behavior in the pest behavior expert database means that the character comparison of the pest behavior and the preset behavior in the pest behavior expert database is inconsistent, and the expert terminal refers to the human-computer interaction interface of the pest behavior expert database.

[0090] Specifically, the pest behavior recognition module compares the crop damage level value C with the preset crop damage level value C0, 0.26≤C0≤0.31, and judges the pest behavior damage situation based on the comparison result, and calibrates the adjustment of the construction process of the crop growth state model based on the judgment result, wherein:

[0091] When C≤C0, the pest behavior recognition module determines that the pest behavior hazard is small and does not calibrate the adjustment of the construction process of the crop growth state model;

[0092] When C>C0, the pest behavior recognition module determines that the pest behavior hazard is serious, calibrates the adjustment of the construction process of the crop growth status model, divides the area into a highly sensitive area, re-judges the acquisition of crop growth data in the area, and uses the pest behavior as supplementary feature learning data to retrain the crop growth status model.

[0093] Specifically, the damage degree value of crops refers to the numerical representation of the damage degree of pest behavior to crops. The damage degree value of crops refers to a preset value used to judge the damage situation of pest behavior. The damage situation of pest behavior refers to the damage situation of pest behavior to crops. The damage situation of pest behavior includes the damage situation of pest behavior being small and the damage situation of pest behavior being large.

[0094] Specifically, the intelligent pest control module constructs a pesticide pest simulation model based on the required pest mortality rate Ps, the first model parameter ac and the second model parameter bc in the pesticide spraying data, and sets the formula of the pesticide pest simulation model as follows: e is the base of the natural logarithm. The intelligent pest control module inputs the required pest mortality rate into the pesticide pest simulation model to obtain the pesticide spraying amount J output by the pesticide pest simulation model.

[0095] Specifically, the pesticide spraying amount refers to the amount of pesticide sprayed, the required pest mortality rate refers to the pest mortality rate required to kill pests according to actual needs, and the first model parameter and the second model parameter refer to the calculation parameters used to construct a pesticide pest simulation model. This embodiment does not limit the specific values of the first model parameter and the second model parameter. Those skilled in the art can set them according to actual needs, as long as the requirements for constructing a pesticide pest simulation model are met. For example, the first model parameter and the second model parameter can be estimated using the nonlinear least squares method. The pesticide pest simulation model refers to a machine learning model with the required pest mortality rate as input and the pesticide spraying amount as output.

[0096] Specifically, the intelligent pest control module compares the regional pest density ρ in the pest data with the preset pest density ρ0, 10 / m 2 ≤ρ0≤23 pieces / m 2 , judge the pest density situation according to the comparison results, and modify the construction process of the pesticide pest simulation model according to the judgment results, including:

[0097] When ρ≤ρ0, the intelligent pest control module determines that the pest density is not serious and does not modify the construction process of the pesticide pest simulation model;

[0098] When ρ>ρ0, the intelligent pest control module determines that the pest density is serious, modifies the construction process of the pesticide pest simulation model, modifies the required pest mortality rate Ps, obtains the corrected required pest mortality rate Ps`, sets Ps`=Ps+t×ρ / bc, where t is the pest distribution density parameter, replaces the required pest mortality rate Ps with the corrected required pest mortality rate Ps`, and reconstructs the pesticide pest simulation model.

[0099] Specifically, the regional pest density refers to the number of pests present per square meter in the area, the preset pest density refers to a preset value used to judge the pest density situation, the pest density situation refers to the severity of the pest density in the area, and the pest density situation includes a non-serious pest density situation and a serious pest density situation. The pest distribution density parameter refers to a coefficient used to calculate the corrected pesticide spraying amount. This embodiment does not limit the specific value of the pest distribution density parameter. Those skilled in the art can set it according to actual needs, as long as the calculation requirements for the corrected pesticide spraying amount are met. For example, the specific value of the pest distribution density parameter can be limited according to the type of pest.

[0100] Specifically, the intelligent pest control module compares the crop damage level value C obtained by the pest behavior recognition module with the preset crop damage level value C0, judges the pest behavior damage situation based on the comparison result, and adjusts the pest density judgment process based on the judgment result, wherein:

[0101] When C≤C0, the intelligent pest control module determines that the hazard caused by the pest behavior is small and does not adjust the judgment process of the pest density;

[0102] When C>C0, the intelligent pest control module determines that the pest behavior is harmful and adjusts the judgment process of the pest density. The harm coefficient Wh=0.7+0.3×e -0.7×(C-C0) The preset pest density ρ0 is adjusted to obtain the adjusted preset pest density ρ0t, ρ0t=Wh×ρ0 is set, the preset pest density ρ0 is replaced by the adjusted preset pest density ρ0t, and the regional pest density ρ is re-compared with the adjusted preset pest density ρ0t.

[0103] Specifically, the intelligent pest control module calculates the crop growth state sensitivity Mn based on the pest sensitivity M1, the pesticide sensitivity M2, the pest sensitivity coefficient m1, and the pesticide sensitivity coefficient m2, sets Mn = m1 × M1 + m2 × M2, compares the crop growth state sensitivity Mn with the preset crop growth state sensitivity M0, 0.66≤M0≤0.73, judges the crop sensitivity based on the comparison result, and revises the adjustment process of the pest density judgment process based on the judgment result, wherein:

[0104] When Mn<M0, the intelligent pest control module determines that the crop sensitivity is not sensitive and does not revise the adjustment process of the pest density judgment process;

[0105] When Mn≥M0, the intelligent pest control module determines that the crop sensitivity is sensitive, revises the adjustment process of the pest density judgment process, and revises the preset crop damage level value C0 through the sensitivity coefficient α, setting 0.54≤α≤0.87 to obtain the revised preset crop damage level value C0x, set C0x=C0×α, and replace the preset crop damage level value C0 with the revised preset crop damage level value C0x, and re-compare the crop damage level value C with the revised preset crop damage level value C0x.

[0106] Specifically, the sensitivity to pests refers to the sensitivity of the current crop growth state to pests. The sensitivity to pests is obtained by comparing the current crop growth state with historical expert data, and the sensitivity to pesticides is obtained by comparing the current crop growth state with historical expert data sets. This embodiment does not limit the comparison process of the current crop growth state with the historical expert data sets. Those skilled in the art can set it up according to actual needs. It only needs to meet the acquisition requirements of the sensitivity to pesticides and the sensitivity to pests. For example, the current crop growth state can be compared with the historical crop growth state in the historical expert data set. When the current crop growth state is compared with the historical crop growth state in the historical expert data set, When comparing the historical crop growth status, the historical pesticide sensitivity and historical pest sensitivity corresponding to the historical crop growth status in the historical expert dataset are obtained, and the historical pesticide sensitivity and historical pest sensitivity are used as the pesticide sensitivity and the pest sensitivity. The historical expert dataset refers to a dataset composed of the experience of experts analyzing the crop growth status and the sensitivity to pests and pesticides. The preset crop growth status sensitivity refers to a preset value used to judge the sensitivity of the crop. The crop sensitivity refers to the sensitivity of the current crop growth status, and the crop sensitivity includes the crop sensitivity being insensitive and the crop sensitivity being sensitive.

[0107] Specifically, the intelligent pest control module compares the pesticide spraying amount J output by the pesticide pest simulation model with the maximum pesticide spraying amount J0, 100g / mu≤J0≤150g / mu, and judges the excessive pesticide spraying according to the comparison result, and adjusts the pesticide spraying amount according to the judgment result, wherein:

[0108] When J≤J0, the intelligent disinfection module determines that the excessive spraying of pesticides is not excessive and does not adjust the amount of pesticide spraying;

[0109] When J>J0, the intelligent pest control module determines that the pesticide spraying is excessive, adjusts the pesticide spraying amount, replaces the pesticide spraying amount J with the maximum pesticide spraying amount J0, and uses physical insect killing boards to capture and kill pests while spraying pesticides on crops according to the maximum pesticide spraying amount J0.

[0110] Specifically, the maximum pesticide spraying amount refers to the maximum pesticide spraying amount that crops can withstand, the excessive pesticide spraying situation refers to whether the pesticide spraying amount exceeds the maximum pesticide spraying amount, and the excessive pesticide spraying situation includes the excessive pesticide spraying situation being not excessive and the excessive pesticide spraying situation being excessive.

[0111] Specifically, when the intelligent disinfection module determines that the pesticide over-spraying is excessive, the crop growth status model is adjusted, and the preset feature test accuracy F0 is adjusted by the spraying adjustment coefficient kp=1.16 to obtain the third adjusted preset feature test accuracy F0p, set F0p=F0×kp, and replace the preset feature test accuracy F0 with the third adjusted preset feature test accuracy F0p, and re-compare the feature test accuracy F.

[0112] Specifically, when the intelligent disinfection module determines that the excessive pesticide spraying is excessive, the judgment process of the alignment accuracy is adjusted, and the preset alignment accuracy error RMSE0 is adjusted through the accuracy judgment adjustment coefficient kjd=1.43 to obtain the adjusted preset alignment accuracy error RMSE0k, set RMSE0k=RMSE0×kjd, replace the preset alignment accuracy error RMSE0 with the adjusted preset alignment accuracy error RMSE0k, and re-compare the alignment accuracy error RMSE with the adjusted preset alignment accuracy error RMSE0k.

[0113] Specifically, the intelligent pest control module compares the soil organic matter content Fw in the crop growth data with the preset soil organic matter content Fw0, 2%≤Fw0≤3%, and judges the soil fertility based on the comparison results. Based on the judgment results, the judgment process of excessive pesticide spraying is corrected, wherein:

[0114] 当 FW ＜ At Fw0, the intelligent pest control module determines that the soil fertility is infertile and does not correct the judgment process of excessive pesticide spraying;

[0115] 当When Fw≥Fw0, the intelligent disinfection module determines that the soil fertility is fertile, and corrects the judgment process of excessive pesticide spraying. The maximum pesticide spraying amount J0 is corrected by the fertility correction coefficient Er=1.78 to obtain the corrected maximum pesticide spraying amount J0E, and J0E=J0×Er is set. The maximum pesticide spraying amount J0 is replaced with the corrected maximum pesticide spraying amount J0E, and the pesticide spraying amount J is re-compared with the corrected maximum pesticide spraying amount J0E.

[0116] Specifically, the soil organic matter content refers to the percentage of the total amount of all carbon-containing organic matter in the soil to the total mass of the soil. The preset soil organic matter content refers to a preset value used to judge the fertility of the soil. The soil fertility refers to whether the soil is fertile. The soil fertility includes the soil fertility being infertile and the soil fertility being fertile.

[0117] Specifically, the intelligent disinfection module compares the ambient relative humidity G in the meteorological data with the maximum ambient relative humidity Gmax and the minimum ambient relative humidity Gmin, Gmax = 60%, Gmin = 40%, and judges the impact of the ambient relative humidity on soil fertility based on the comparison results, and adjusts the judgment process of the soil fertility based on the judgment results, wherein:

[0118] When Gmin<G≤Gmax, the intelligent disinfection module determines that the influence of the relative humidity of the environment on the soil fertility is not affected, and does not adjust the judgment process of the soil fertility;

[0119] When G≤Gmin, the intelligent disinfection module determines that the influence of the relative humidity of the environment on the soil fertility is an influence, adjusts the judgment process of the soil fertility, adjusts the preset soil organic matter content Fw0 by the relative humidity coefficient ε, 1.21≤ε≤1.33, obtains the adjusted preset soil organic matter content Fw0t, sets Fw0t=Fw0, replaces the preset soil organic matter content Fw0 with the adjusted preset soil organic matter content Fw0t, and re-compares the soil organic matter content Fw with the adjusted preset soil organic matter content Fw0t;

[0120] When G>Gmax, the intelligent disinfection module determines that the impact of the relative humidity of the environment on the soil fertility is an impact, adjusts the judgment process of the soil fertility, adjusts the preset soil organic matter content Fw0 by the relative humidity coefficient ε, 1.21≤ε≤1.33, and obtains the adjusted preset soil organic matter content Fw0t. Set Fw0t=Fw0 to replace the preset soil organic matter content Fw0 with the adjusted preset soil organic matter content Fw0t, and re-compare the soil organic matter content Fw with the adjusted preset soil organic matter content Fw0t.

[0121] Specifically, the ambient relative humidity refers to the percentage of the actual water vapor content in the air at the current temperature to the maximum water vapor content that the air can hold at the current temperature. The maximum ambient relative humidity refers to the maximum preset value used to judge the impact of the ambient relative humidity on soil fertility. The minimum ambient relative humidity refers to the minimum preset value used to judge the impact of the ambient relative humidity on soil fertility. The impact of the ambient relative humidity on soil fertility refers to whether the ambient relative humidity affects soil fertility. The impact of the ambient relative humidity on soil fertility includes the impact of the ambient relative humidity on soil fertility being no impact and the impact of the ambient relative humidity on soil fertility being impact.

[0122] Specifically, the intelligent pest control module obtains the crop growth stage output by the crop growth state model and sets the pesticide spraying angle according to the crop growth stage, wherein:

[0123] When the crop growth stage output by the crop growth state model is the seedling stage, the intelligent pest control module sets the pesticide spraying angle to jd1, and sets 5°≤jd1≤15°;

[0124] When the crop growth stage output by the crop growth state model is the maturity stage, the intelligent pest control module sets the pesticide spraying angle to jd2, and sets 30°≤jd2≤60°;

[0125] When the crop growth stage output by the crop growth state model is the fruiting period, the intelligent pest control module sets the pesticide spraying angle to jd3, and sets 15°<jd3<30°.

[0126] Specifically, the pesticide spraying angle refers to the angle between the pesticide spraying path and the ground.

[0127] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. An intelligent image recognition and disinfection robot, characterized in that: include: The robot head is connected to the robot shell, central processing unit, and camera respectively, and is used for 360° rotation, so that the camera can obtain a 360° working environment; The camera is connected to the central processing unit and the robot head respectively to obtain information about the surrounding environment; The first robot hand is connected to the robot housing, the chemical spray device, and the central processing unit, and is used to fix the chemical spray device; The central processing unit is connected to each component of the robot and is used to process all relevant data obtained by the robot; The intelligent control system of the intelligent image recognition disinfection robot is connected to the central processing unit and is used to control the robot based on the data obtained by the robot; The robot housing is connected to the first robot hand, the second robot hand, and the robot walking device, respectively, and is used to protect the internal components; A robot walking device, connected to the robot housing and the central processing unit, respectively, for making the robot walk; a chemical spray device, connected to the central processing unit and the second robot arm, respectively, for killing pests with pesticides; A physical pest control board is connected to the central processing unit and the first robot arm respectively, and is used for physically killing pests; The second robot hand is connected to the robot shell, the physical insect-killing board and the central processing unit respectively, and is used for fixing the physical insect-killing board.

2. An intelligent control system applied to the intelligent image recognition and disinfection robot according to claim 1, characterized in that: The intelligent control system comprises: Data acquisition module, used to acquire mountain environment data, pesticide spraying data, crop growth data and pest data; A terrain simulation module is used to construct a mountain terrain model based on mountain environment data and to calibrate the construction process of the mountain terrain model; A crop growth monitoring module is used to construct a crop growth status model based on crop growth data, obtain crop growth status based on the crop growth status model, obtain the planting area type based on the mountainous terrain model, and adjust the construction process of the crop growth status model based on the planting area type; A pest behavior recognition module is used to input pest data into a pest behavior classification probability model to obtain pest behavior, obtain the degree of damage to crops based on the pest behavior, and calibrate the adjustment of the construction process of the crop growth status model based on the degree of damage to crops; The intelligent pest control module is used to input pesticide spraying data into the pesticide pest simulation model to obtain the pesticide spraying amount and to correct the pesticide pest simulation model. It is also used to adjust the correction of the pesticide pest simulation model according to the degree of damage to the crops. It is also used to adjust the training process of the crop growth status model according to the pesticide spraying data and to adjust the correction process of the mountainous terrain model.

3. The intelligent control system of the intelligent image recognition and disinfection robot according to claim 2 is characterized in that: The terrain simulation module constructs a mountain terrain model by inputting mountain environment data into a mountain environment simulation model.

4. The intelligent control system of the intelligent image recognition and disinfection robot according to claim 3 is characterized in that: The terrain simulation module is based on the first coordinate of the i-th control point of the three-dimensional point cloud data in the mountain environment data. The second coordinate of the i-th control point in the image data The registration accuracy error RMSE is calculated based on the number of control points ek, and the registration accuracy error RMSE is compared with the preset registration accuracy error RMSE0. The registration accuracy is judged according to the comparison results, and the construction process of the mountain environment simulation model is corrected according to the judgment results.

5. The intelligent control system of the intelligent image recognition and disinfection robot according to claim 4 is characterized in that: The crop growth monitoring module is based on each data point x of the independent variable X in the crop growth data. t , standard deviation S of the independent variable X X , the mean of the independent variable X Each data point y at crop growth stage Y t , the mean value of Y in the crop growth stage Standard deviation S of crop growth stage Y Y The Pearson correlation coefficient r is calculated based on the number of data n, and the Pearson correlation coefficient r is compared with the preset Pearson correlation coefficient r0. The correlation between the independent variable and the crop growth stage is judged according to the comparison result, and the independent variable and the crop growth stage are divided into feature learning data according to the judgment result. The crop growth status model is constructed using the feature learning data.

6. The intelligent control system of the intelligent image recognition and disinfection robot according to claim 5 is characterized in that: The crop growth monitoring module divides the crop planting areas according to the mountain environment data to obtain the planting area types, which include general flat areas, remote slope areas and highly sensitive areas. The crop growth data acquisition situation is judged according to the planting area type, and the construction process of the crop growth status model is adjusted according to the judgment result.

7. The intelligent control system of the intelligent image recognition and disinfection robot according to claim 6 is characterized in that: The pest behavior recognition module inputs the pest video data and crop growth data in the pest data into the LSTM model, and obtains the probability P of the occurrence of pest behavior corresponding to each node output by the LSTM model. LSTM , P LSTM =[P 1 LSTM ,P 2 LSTM ,...P f LSTM ], f is the number of categories of pest behavior; The pest behavior recognition module inputs pest data into the STGCN model and obtains the probability distribution P of pest behavior in the spatial dimension output by the STGCN model. STGCN , P STGCN =[P 1 STGCN ,P 2 STGCN ,...P f STGCN ]; The pest behavior recognition module is based on the probability P of the pest behavior corresponding to each node. LSTM and the probability distribution P of pest behavior in spatial dimension STGCN For the fused probability vector P fusion Calculate and get the fused probability vector P fusion , w1 is the weight coefficient of the probability of pest behavior corresponding to each node, w2 is the weight coefficient of the probability distribution of pest behavior in the spatial dimension, and the fused probability vector P fusion Input the pest behavior classification probability model, obtain the pest behavior z output by the pest behavior classification probability model, compare the pest behavior with the preset behavior in the pest behavior expert database, and obtain the damage degree value C of the crop based on the comparison result.

8. The intelligent control system of the intelligent image recognition and disinfection robot according to claim 7 is characterized in that: The pest behavior recognition module compares the crop damage level value C with the preset crop damage level value C0, 0.26≤C0≤0.31, and judges the pest behavior damage situation based on the comparison result, and calibrates the adjustment of the construction process of the crop growth status model based on the judgment result.

9. The intelligent control system of the intelligent image recognition disinfection robot according to claim 8, characterized in that: The intelligent pest control module inputs the required pest mortality rate into the pesticide pest simulation model to obtain the pesticide spraying amount J output by the pesticide pest simulation model.

10. The intelligent control system of the intelligent image recognition disinfection robot according to claim 9, characterized in that: The intelligent pest control module compares the crop damage level value C obtained from the pest behavior recognition module with the preset crop damage level value C0, judges the pest behavior damage situation based on the comparison result, and adjusts the pest density judgment process based on the judgment result.

Images