Intelligent monitoring and early warning robot for apple orchard diseases and insect pests

Abstract

The invention provides an apple orchard pest and disease damage intelligent monitoring and early warning robot, and relates to the technical field of artificial intelligence. The apple orchard pest and disease damage intelligent monitoring and early warning robot comprises a power supply module, a pest luring and capturing module, a data processing and analyzing module, a data transmission and communication module, a walking and navigation module, a user interface and control module and an image acquisition and processing module. Identification, counting, monitoring and early warning and knowledge questioning and answering of diseases and insect pests are realized. The apple orchard pest and disease damage intelligent monitoring and early warning robot provided by the invention realizes efficient, accurate and real-time pest and disease damage monitoring and early warning of the apple orchard, and improves the automation level and the intelligent degree of apple orchard pest and disease damage management.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an intelligent monitoring and early warning robot for diseases and insect pests in an apple orchard. Background Art

[0002] As an economic fruit, apple plays a huge role in rural economic development and farmers' income increase. Due to the long growth cycle and stable ecosystem of apple, there are many kinds of diseases and pests, such as apple rust, apple brown spot, borer, apple moth, cotton bollworm, etc., which seriously restrict the yield and quality of apples.

[0003] In the existing technology, the identification of pests and diseases in apple orchards is done manually, which is not only time-consuming and labor-intensive, but also inefficient; on the other hand, static or semi-static monitoring equipment is used, which not only makes it difficult to fully cover all areas, but also has low identification accuracy and efficiency, and cannot respond to sudden situations of pests and diseases in real time. Summary of the invention

[0004] The present invention provides an apple orchard disease and insect pest intelligent monitoring and early warning robot, which is used to solve the problems of low efficiency, poor accuracy and poor real-time performance in the prior art of apple orchard disease and insect pest identification, and realizes efficient, accurate and real-time disease and insect pest monitoring and early warning in apple orchards.

[0005] The present invention provides an apple orchard disease and insect pest intelligent monitoring and early warning robot, comprising the following modules: Pest attracting and capturing module, image acquisition and processing module, data processing and analysis module, power supply module and data transmission and communication module; The pest attracting and capturing module is used to attract and capture pests, and to inactivate the pests; The image acquisition and processing module at least includes a first camera, a second camera and a third camera; the first camera and the second camera are used to collect real-time video streams on both sides of the apple orchard; the third camera is used to collect images of inactivated pests; The data processing and analysis module includes an orchard environmental data collection system, a pest and disease identification and counting system, a pest and disease monitoring and early warning system, and a knowledge question-answering system based on a knowledge graph; The orchard environmental data collection system is used to collect real-time environmental data of the apple orchard; The pest identification and counting system includes a pest identification subsystem and a pest counting subsystem; the pest identification subsystem is used to output pest detection results through the MH-YOLO model according to the real-time video streams on both sides of the apple orchard and the inactivated pest images; the pest counting subsystem is used to output the number of pests through the DeepSORT counting model according to the pest detection results; The pest and disease monitoring and early warning system is used to output a pest and disease early warning signal based on the real-time environmental data of the apple orchard, the pest and disease detection results and the pest and disease quantity; The knowledge graph-based knowledge question-answering system is used to generate comprehensive information on pests and diseases based on the pest and disease detection results, the pest and disease quantity, and the pest and disease early warning signal; The power module is used to provide energy required for the operation of the apple orchard pest and disease intelligent monitoring and early warning robot; The data transmission and communication module is used for real-time transmission and storage of data.

[0006] According to the apple orchard pest intelligent monitoring and early warning robot provided by the present invention, the MH-YOLO model is optimized based on the YOLOv5s model; The optimization includes at least a first optimization, a second optimization, a third optimization and a fourth optimization; The first optimization is to introduce a CBAM attention mechanism into the backbone network of the YOLOv5s model, and the first optimization is used to enhance the perception ability of the YOLOv5s model to the characteristics of pests and diseases; The second optimization is to introduce Swin-Transformer blocks into the first CSP2_1 module of the neck network of the YOLOv5s model, and the second optimization is used to enhance the information integration capability of the YOLOv5s model; The third optimization is to introduce an ASFF adaptive feature fusion module at the end of the neck network of the YOLOv5s model, and the third optimization is used to improve the recognition ability of the YOLOv5s model for pest and disease characteristics; The fourth optimization is to perform structured pruning on the YOLOv5s model after the first optimization, the second optimization and the third optimization, and the fourth optimization is used to reduce the number of parameters and complexity of the YOLOv5s model.

[0007] According to the apple orchard pest intelligent monitoring and early warning robot provided by the present invention, the DeepSORT counting model includes a region recognition module, a feature extractor, a Kalman filter, a feature association module and a re-recognition module; The area recognition module is used to identify the pest and disease detection results output by the MH-YOLO model to obtain the pest and disease area; The feature extractor is used to extract features from the pest area and generate a feature vector describing the appearance features of the pest; The Kalman filter is used to predict the state of the pests and diseases based on the characteristic vector to obtain a new state estimate; The feature association module is used to generate a pest tracking trajectory based on the new state estimation; The re-identification module is used to determine the number of pests and diseases based on the feature vector and the pest and disease tracking trajectory.

[0008] According to the invention, an intelligent monitoring and early warning robot for apple orchard pests and diseases is provided, wherein the pest attracting and capturing module comprises a capturing unit, an insect killing unit, an insect receiving plate, an insect dropping plate, a controller, a conveying crawler and a cleaner; The capture unit is used to capture apple orchard pests; The pest control unit is used to perform a deactivation treatment on pests; the deactivation treatment includes a killing treatment and a drying treatment; The insect receiving plate is used to receive the pests after killing; The insect drop plate is used to receive the pests after drying and to make a predetermined number of pests fall onto the conveyor belt through a weight sensor; The controller is used to control the opening and closing mechanism so that the pests on the insect receiving board fall onto the insect dropping board after being dried; The conveyor crawler is used to convey the pests dropped from the insect dropping plate to the image acquisition and processing module; The cleaner is used for cleaning the pests adhered to or overlapped on the conveying crawler.

[0009] According to the apple orchard disease and insect pest intelligent monitoring and early warning robot provided by the present invention, the capture unit includes an impact plate and an insect attractant lamp; The impact plates are placed crosswise to capture the lured apple orchard pests; The insect-attracting lamp is used for attracting pests in apple orchards.

[0010] According to the intelligent monitoring and early warning robot for apple orchard pests provided by the present invention, a sex attractant core is placed at the center of the impact plate, and the sex attractant core is used to attract pests in the apple orchard.

[0011] According to the apple orchard disease and insect pest intelligent monitoring and early warning robot provided by the present invention, the insecticide unit includes a high-temperature insecticide and an oven; The high-temperature insect killer is used to kill pests by high heat; The oven is used for drying the pests after the killing process.

[0012] According to the present invention, an apple orchard pest intelligent monitoring and early warning robot further includes a walking and navigation module; the walking and navigation module includes a laser radar, a GPS module, a path planning unit, a walking mechanism and an obstacle avoidance sensor; The laser radar is used to generate a three-dimensional environment model based on the terrain, vegetation distribution and obstacle locations in the apple orchard; The GPS module is used to generate positioning information according to the real-time position of the apple orchard pest and disease intelligent monitoring and early warning robot; The path planning unit is used to generate a path planning map according to the three-dimensional environment model and the positioning information; The walking mechanism is used to control the apple orchard pest and disease intelligent monitoring and early warning robot to walk according to the path planning map; The obstacle avoidance sensor is used to detect obstacles ahead in real time when the apple orchard pest and disease intelligent monitoring and early warning robot walks according to the path planning map.

[0013] According to the apple orchard pest intelligent monitoring and early warning robot provided by the present invention, the walking mechanism includes a walking drive module, a suspension module and a steering module; The walking drive module is used to provide power for the apple orchard pest and disease intelligent monitoring and early warning robot to walk; The suspension module is used to buffer the impact of uneven ground on the walking of the apple orchard pest and disease intelligent monitoring and early warning robot; The steering module is used to control the steering of the apple orchard pest and disease intelligent monitoring and early warning robot.

[0014] An apple orchard disease and insect pest intelligent monitoring and early warning robot provided according to the present invention also includes a user interface and a control module, and the user interface and the control module are used to support information interaction between the user and the apple orchard disease and insect pest intelligent monitoring and early warning robot.

[0015] The present invention provides an apple orchard disease and insect pest intelligent monitoring and early warning robot, which traps apple orchard pests through a pest luring and capturing module, collects apple orchard data in real time through an image acquisition and processing module, obtains apple orchard disease and insect pest information, performs real-time data processing and analysis through a data processing and analysis module, realizes disease and insect pest identification, counting, monitoring and early warning, and knowledge question and answer, and provides energy and communication guarantees for the operation of the robot through the coordinated operation of a power module and a data transmission and communication module, thereby realizing efficient, accurate and real-time disease and insect pest monitoring and early warning in the apple orchard, and improving the automation level and intelligence level of disease and insect pest management in the apple orchard. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1The present invention provides a module schematic diagram of an apple orchard disease and insect pest intelligent monitoring and early warning robot.

[0018] Figure 2 The present invention is a schematic diagram of the appearance of an intelligent monitoring and early warning robot for apple orchard pests and diseases provided by the present invention.

[0019] Figure 3 The present invention is a schematic diagram of the main structure of an intelligent monitoring and early warning robot for apple orchard pests and diseases.

[0020] Figure 4 The present invention is a schematic diagram of the partial structure of an intelligent monitoring and early warning robot for apple orchard pests and diseases provided by the present invention.

[0021] Figure 5 The present invention provides a chassis structure schematic diagram of an apple orchard disease and insect pest intelligent monitoring and early warning robot.

[0022] Figure 6 It is a schematic diagram of the system embedded in the edge computing platform of the apple orchard pest and disease intelligent monitoring and early warning robot provided by the present invention.

[0023] Figure 7 It is a schematic diagram of the YOLOv5s model structure and the MH-YOLO model structure provided by the present invention.

[0024] Figure 8 It is a schematic diagram of the structured pruning of the MH-YOLO model provided by the present invention.

[0025] Reference numerals: 1-solar panel, 2-insect attractant lamp, 3-impact plate, 4-sexual attractant core, 5-insect receiving funnel, 6-insect receiving plate, 7-insect killing unit, 8-opening and closing mechanism, 9-insect dropping plate, 10-lidar, 11-first camera, 12-fill light, 13-collecting device, 14-second and third cameras, 15-sensor module, 16-display screen, 17-edge computing platform, 18-data transmission and communication module, 19-sweeper, 20-track, 21-tire, 22-axle, 23-spring, 24-shock absorber, 25-motor, 26-transmission device, 27-power module. DETAILED DESCRIPTION

[0026] With the modernization of global agriculture, the demand for refined and intelligent management of apple orchards is growing. As an economic fruit, apples play a huge role in rural economic development and farmers' income. Due to the long growth cycle of apples and the stable ecosystem, there are many kinds of diseases and pests, such as apple rust, apple brown spot, peach borer, pear borer, citrus borer, apple roller moth and cotton bollworm, which seriously restrict the yield and quality of fruits.

[0027] At present, the identification of pests and diseases in apple orchards is mainly done manually or through monitoring systems. On the one hand, manual identification is not only time-consuming and labor-intensive, but also inefficient. On the other hand, the existing apple orchard pest and disease monitoring systems mostly use static or semi-static monitoring equipment. Although they can achieve a certain degree of automation, they have many limitations in terms of coverage, flexibility and real-time performance. Especially in large-scale apple orchards, traditional equipment is difficult to fully cover all areas and cannot respond to sudden pest and disease situations in real time. In addition, the data processing capacity of the existing system is limited, and it is difficult to fully utilize multi-source sensor data for accurate pest and disease identification and dynamic early warning.

[0028] Therefore, existing pest and disease monitoring equipment often cannot adapt to the complex and changeable terrain of apple orchards. The equipment needs to be monitored in a fixed or semi-fixed form, resulting in a limited monitoring range and inability to achieve comprehensive coverage. At the same time, these devices lack flexible navigation and mobility capabilities, making it difficult to conduct efficient autonomous navigation in large-scale orchards. The data processing of existing monitoring systems also mainly relies on centralized computing, which has certain limitations in real-time and response speed, making it impossible to quickly process large-scale environmental data and pest and disease information.

[0029] In addition, traditional pest and disease identification methods still rely on manual experience and simple image processing technology, with low recognition accuracy and efficiency, and it is difficult to cope with the needs of identifying multiple types and forms of pests and diseases. In terms of pest and disease early warning, the existing system lacks an effective dynamic early warning mechanism, and the accuracy and timeliness of early warning information are insufficient, which cannot provide fruit farmers with real-time and scientific decision-making support.

[0030] In order to achieve efficient, accurate and real-time identification and early warning of diseases and pests in apple orchards, the present invention proposes an apple orchard disease and pest intelligent monitoring and early warning robot, which aims to achieve automatic identification, monitoring and early warning of diseases and pests in apple orchards.

[0031] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Combine the following Figures 1 to 8 The invention describes an intelligent monitoring and early warning robot for diseases and insect pests in an apple orchard.

[0033] Figure 1 Schematic diagram of a module of an intelligent monitoring and early warning robot for apple orchard pests provided by the present invention. Figure 1As shown, the apple orchard pest and disease intelligent monitoring and early warning robot includes at least the following modules: a pest attracting and capturing module, an image acquisition and processing module, a data processing and analysis module, a power supply module, and a data transmission and communication module.

[0034] The pest attracting and capturing module is used to attract and capture pests, and to inactivate the pests; The image acquisition and processing module at least includes a first camera, a second camera and a third camera; the first camera and the second camera are used to collect real-time video streams on both sides of the apple orchard; the third camera is used to collect images of inactivated pests; The data processing and analysis module includes an orchard environmental data collection system, a pest and disease identification and counting system, a pest and disease monitoring and early warning system, and a knowledge question-answering system based on a knowledge graph; The orchard environmental data collection system is used to collect real-time environmental data of the apple orchard; The pest identification and counting system includes a pest identification subsystem and a pest counting subsystem; the pest identification subsystem is used to output pest detection results through the MH-YOLO model according to the real-time video streams on both sides of the apple orchard and the inactivated pest images; the pest counting subsystem is used to output the number of pests through the DeepSORT counting model according to the pest detection results; The pest and disease monitoring and early warning system is used to output a pest and disease early warning signal based on the real-time environmental data of the apple orchard, the pest and disease detection results and the pest and disease quantity; The knowledge graph-based knowledge question-answering system is used to generate comprehensive information on pests and diseases based on the pest and disease detection results, the pest and disease quantity, and the pest and disease early warning signal; The power module is used to provide energy required for the operation of the apple orchard pest and disease intelligent monitoring and early warning robot; The data transmission and communication module is used for real-time transmission and storage of data.

[0035] Figure 2 FIG. 1 is a schematic diagram of the appearance of an intelligent monitoring and early warning robot for diseases and insect pests in an apple orchard provided by the present invention. Figure 2 As shown, in order to further demonstrate Figure 2 The specific structure of the intelligent monitoring and early warning robot for apple orchard pests and diseases is shown in the figure. The present invention also provides Figures 3 to 5 , the specific structure of the intelligent monitoring and early warning robot for apple orchard pests and diseases is introduced.

[0036] Figure 3 The main structure diagram of an apple orchard disease and insect pest intelligent monitoring and early warning robot provided by the present invention is as follows: Figure 3As shown, the main structure of the apple orchard pest intelligent monitoring and early warning robot provided by the embodiment of the present invention includes at least 1-solar panel, 2-insect attractant lamp, 3-impact plate, 4-sex attractant core, 5-insect receiving funnel, 6-insect receiving plate, 7-insect killing unit, 8-opening and closing mechanism, 9-insect falling plate, 10-laser radar, 11-first camera, 12-fill light, 13-collecting device, 14-second and third cameras, 15-sensor module, 16-display screen, 17-edge computing platform, 18-data transmission and communication module and 19-sweeper.

[0037] Figure 4 FIG. 1 is a partial structural diagram of an apple orchard disease and insect pest intelligent monitoring and early warning robot provided by the present invention, such as Figure 4 As shown, the local structure of the intelligent monitoring and early warning robot for apple orchard pests and diseases provided by the embodiment of the present invention includes at least: 20-tracks, 21-tire, 22-axle, 23-spring and 24-shock absorber.

[0038] Figure 5 Schematic diagram of the chassis structure of an apple orchard pest intelligent monitoring and early warning robot provided by the present invention, such as Figure 5 As shown, the chassis structure of the apple orchard disease and insect pest intelligent monitoring and early warning robot provided by the embodiment of the present invention at least includes: 25-motor, 26-transmission device and 27-power module.

[0039] In an embodiment of the present invention, the power module in the apple orchard pest and disease intelligent monitoring and early warning robot (hereinafter referred to as the robot) includes a built-in power supply and a solar panel, which are used to provide the energy required for the operation of the robot and ensure that the robot can operate autonomously for a long time.

[0040] The robot is equipped with a solar panel on the top of the shell, which can effectively utilize solar energy resources to provide power for various control units, track systems and overall operation inside the robot. At the same time, the robot is also designed with a built-in power supply (such as a battery, etc.), which can provide reliable power support through the built-in power supply in insufficient light or at night, and power the various modules inside the robot to ensure that the robot can work normally under insufficient light or at night. The power supply method combined with the built-in power supply and the solar panel ensures the continuous and stable operation of the robot under various environmental conditions.

[0041] The pest attracting and capturing module in the robot is used to attract and capture pests and to inactivate them.

[0042] Optionally, the pest attracting and capturing module includes a capturing unit, an insect killing unit, an insect receiving plate, an insect dropping plate, a controller, a conveying crawler and a cleaner; The capture unit is used to capture apple orchard pests; The pest control unit is used to perform a deactivation treatment on pests; the deactivation treatment includes a killing treatment and a drying treatment; The insect receiving plate is used to receive the pests after killing; The insect drop plate is used to receive the pests after drying and to make a predetermined number of pests fall onto the conveyor belt through a weight sensor; The controller is used to control the opening and closing mechanism so that the pests on the insect receiving board fall onto the insect dropping board after being dried; The conveyor crawler is used to convey the pests dropped from the insect dropping plate to the image acquisition and processing module; The cleaner is used for cleaning the pests adhered to or overlapped on the conveying crawler.

[0043] The capture unit includes an impact plate and an insect trap light; The impact plates are placed crosswise to capture the lured apple orchard pests; The insect-attracting lamp is used for attracting pests in apple orchards.

[0044] A sex attractant core is placed at the center of the impact plate, and the sex attractant core is used to attract apple orchard pests.

[0045] The insecticide unit includes a high-temperature insecticide and an oven; The high-temperature insect killer is used to kill pests by high heat; The oven is used for drying the pests after the killing process.

[0046] Specifically, the pest attracting and capturing module includes a capturing unit, an insect killing unit, a controller, an insect dropping plate and a cleaner.

[0047] The capturing unit comprises an insect attracting lamp and an impact plate, wherein the impact plate is made of transparent material, for example, glass or crystal, etc., to make a transparent impact plate.

[0048] Place a sex attractant under the impact plate, and release artificial sex pheromones to attract specific apple orchard pests. At the same time, take advantage of the phototaxis of some pests, and release a light source of a specific spectrum through the insect trap lamp to attract apple orchard pests. Then place the impact plates crosswise to better capture the attracted pests.

[0049] The pests are inactivated using an insecticidal unit, and the inactivation treatment includes a killing treatment and a drying treatment.

[0050] The insecticide unit is a high-pressure and high-heat unit, including a high-heat insecticide and an oven. The high-heat insecticide kills pests by high heat, making it impossible for the pests to escape; the oven dries out the moisture in the pests to prevent them from deteriorating and rotting.

[0051] The killed pests fall onto the insect receiving board, and then the controller controls the opening and closing mechanism. When the preset time (the specific time can be customized according to the actual situation) or the predetermined number of pests (the specific number can be customized according to the actual situation) accumulates on the insect receiving board, the controller opens the opening and closing mechanism, so that the killed pests fall into the drying oven for drying. After drying, the pests further fall onto the insect dropping board.

[0052] A weight sensor is provided at the bottom of the insect drop plate. When the pests accumulate to a predetermined number (the specific number can be customized according to actual conditions), the weight sensor triggers the switch to ensure that the pests fall evenly onto the conveyor track in the track system, thereby effectively controlling the distribution of pests, preventing pest accumulation, and improving the efficiency of conveying and processing.

[0053] In order to clean up stuck or overlapping pests and to subsequently take high-quality pest images, the embodiment of the present invention also provides a sweeper above the conveyor belt, so that the inactivated pests on the conveyor belt are cleaned by the sweeper and then transmitted to the image acquisition and processing module, so that the subsequent camera can take clear images of the inactivated pests, thereby improving the recognition accuracy.

[0054] In order to facilitate the smooth sliding of the pests, an insect receiving funnel is further provided under the impact plate in the embodiment of the present invention. The insect receiving plate and the insect receiving funnel are both made of smooth materials to prevent the pests from stagnating during the sliding process, thereby ensuring that the pests smoothly enter the insecticidal unit.

[0055] In the embodiment of the present invention, the image acquisition and processing module at least includes a first camera, a second camera and a third camera. In addition, in order to improve the shooting quality of the camera, a fill light may be added.

[0056] During the robot monitoring process, the first and second cameras are used to collect real-time video streams from both sides of the apple orchard, and the third camera is used to collect images of inactivated pests. In dimly lit scenes, fill lights can be used to improve shooting quality.

[0057] The videos and images captured by the above cameras are input into the data processing and analysis module of the robot. The data processing and analysis module is the core module of the robot, which integrates the orchard environmental data collection system, pest and disease identification and counting system, pest and disease monitoring and early warning system, and knowledge question and answer system based on knowledge graph through the edge computing platform.

[0058] Figure 6 Schematic diagram of the system embedded in the edge computing platform of the apple orchard pest intelligent monitoring and early warning robot provided by the present invention. Figure 6 shown.

[0059] The orchard environmental data collection system integrates a variety of high-precision sensors such as temperature sensors, humidity sensors, light sensors, wind speed sensors, and wind direction sensors. The sensors are connected to the edge computing platform and can collect the orchard's environmental data in real time. The orchard environmental data collection system uses a variety of sensors to monitor and analyze the orchard's microclimate conditions in real time, providing a scientific basis for subsequent judgments on the development patterns of pests and diseases.

[0060] Through the monitoring and data transmission of these sensors, the robot can obtain parameters such as temperature, humidity, and light in the orchard, realize real-time monitoring and accurate assessment of the orchard's climatic conditions, and provide high-precision data support for the identification and prevention of pests and diseases.

[0061] The pest identification and counting system integrates the pest identification subsystem and the pest counting subsystem. At the same time, the pest identification and counting system is also connected to the image acquisition and processing module, so that it can obtain the high-definition image and video data collected by the image acquisition and processing module in real time, and analyze and process them through the deep learning model (i.e., MH-YOLO model) to achieve real-time automatic identification and classification of pests. At the same time, the DeepSORT counting algorithm is used to track and count pests in real time based on the preliminary processing results of the photographed pests by the MH-YOLO model, accurately track the movement trajectory of individual pests, avoid repeated counting, and thus improve the recognition accuracy of pests.

[0062] In an embodiment of the present invention, the pest and disease identification and counting system can accurately determine the type and quantity of pests and diseases through efficient image processing, data analysis and quantity statistics. Among them, the pest and disease identification subsystem is used to detect and classify pests and diseases in video frames and pest images based on the real-time video streams on both sides of the apple orchard and the images of inactivated pests through the MH-YOLO model, and output the pest and disease detection results including bounding boxes, category labels and confidence scores. The pest and disease counting subsystem is used to output the number of pests and diseases through the DeepSORT counting model based on the pest and disease detection results.

[0063] Based on the above embodiments, the pest and disease monitoring and early warning system provided by the present invention outputs a pest and disease early warning signal based on the real-time environmental data of the apple orchard, the pest and disease detection results and the number of pests and diseases.

[0064] Specifically, the pest and disease monitoring and early warning system first integrates the environmental data collected by the orchard environmental data collection system and the pest and disease type and quantity information obtained by the pest and disease identification and counting system to construct a multidimensional data set, and uses the Bayesian data fusion algorithm to integrate multi-source data to improve the reliability and consistency of the data; then the AutoRegressiveIntegrated Moving Average (hereinafter referred to as ARIMA) model is used to perform time series prediction on the number of pests and diseases to identify seasonal and trend changes. The ARIMA model captures the temporal dynamic characteristics of the data through differential, stationary processing and autoregressive moving average methods; then, combined with environmental data, the vector autoregression (VAR) model is used to evaluate the impact of environmental factors on the number of pests and diseases. The VAR model predicts future changes in the number of pests and diseases by capturing the interdependence of multivariate time series.

[0065] In addition, the pest and disease monitoring and early warning system also uses a random forest model to classify pest and disease levels. By integrating multiple decision trees and using bagging technology, the model's stability and generalization ability are improved. During the model training process, grid search is used for hyperparameter optimization. The model performance is evaluated through cross-validation technology to ensure that the model can accurately classify pest and disease levels, thereby achieving accurate and timely early warnings.

[0066] In an embodiment of the present invention, the pest and disease monitoring and early warning system is connected to the pest and disease identification and counting system, and through real-time data analysis and processing, dynamic monitoring and early warning of pests and diseases are realized. According to the data provided by the pest and disease identification and counting system, the occurrence and development trend of pests and diseases are evaluated in real time, and early warning information is generated.

[0067] Through the rapid analysis and response of the pest monitoring and early warning system, the robot can detect early signs of pests and diseases in a timely manner and send early warning signals to fruit farmers. The early warning signals generated by the pest monitoring and early warning system are displayed in real time through the user interface and the display screen in the control module, helping fruit farmers to carry out scientific pest control.

[0068] The knowledge question answering system based on knowledge graph is used to generate comprehensive information on pests and diseases based on pest and disease detection results, pest and disease quantity, and pest and disease warning signals.

[0069] The knowledge question-answering system based on the knowledge graph integrates a large amount of agricultural knowledge and pest control information to form a complete knowledge graph. The knowledge question-answering system based on the knowledge graph uses natural language processing and intelligent reasoning technology to generate targeted answers to questions based on the data provided by the pest monitoring and early warning system and the pest identification system, including comprehensive pest information, pest quantity and type, etc. Users can query pest-related information through the system to obtain scientific prevention and control suggestions and solutions.

[0070] The robot also includes a user interface and a control module, including a display screen and a remote control interface, and the user interface and the control module are used to support information interaction between the user and the apple orchard pest and disease intelligent monitoring and early warning robot.

[0071] The knowledge question and answer system based on the knowledge graph interacts with the user through the display screen in the user interface and control module to realize intelligent question and answer.

[0072] The knowledge question-answering system based on the knowledge graph builds an agricultural knowledge graph and combines it with natural language processing technology to achieve intelligent answers to user queries and provide fruit farmers with real-time and accurate pest and disease management recommendations.

[0073] Among them, building the knowledge graph of apple orchard pests and diseases includes the following steps: (1) Collect data from a variety of data sources such as agricultural literature, expert manuals, and pest control records. Use Web crawler technology to automatically capture relevant information on the Internet and combine it with manually organized expert knowledge base data to form a rich original data set. (2) Clean, convert and normalize the collected data to ensure its accuracy and consistency; (3) Use natural language processing technology to preprocess text data and extract key information; (4) Use information extraction algorithms to extract entities, attributes, and relationships from the preprocessed data. The types of entities include pest and disease types, symptom descriptions, prevention and control methods, and drug names.

[0074] (5) Extract complex knowledge structures through the BERT model and build a knowledge graph; (6) Store the knowledge graph in a graph database, storing entities and relationships in a structured manner using nodes and edges.

[0075] When the knowledge graph-based knowledge question answering system is running, it first uses natural language processing technology to understand and analyze the questions input by users. For example, it identifies and extracts keywords in the questions and uses Word2Vec word embedding technology to convert keywords into vector representations to facilitate subsequent knowledge retrieval. Secondly, perform knowledge retrieval in the knowledge graph, use Cypher to write query language, and find relevant entities and relationships in the knowledge graph based on question types and keywords; Next, a graph-based similarity algorithm is used to calculate the knowledge nodes related to the question, filter out the most relevant knowledge items, and integrate and organize the retrieved knowledge items to generate a structured answer; Finally, natural language processing technology is used to convert structured data into natural language descriptions to ensure the readability and accuracy of the answers.

[0076] In the embodiment of the present invention, the knowledge question-answering system based on the knowledge graph synchronizes the knowledge graph and user interaction data to the central server in real time through the wireless communication module to ensure the unification of data, so that users can input query questions through the display screen on the robot and obtain real-time answers. The display screen interface is simple and easy to use, supports text input and voice input, and improves the user experience.

[0077] The intelligent answers provided by the knowledge question-and-answer system based on the knowledge graph are displayed in real time on the display screen to help fruit farmers carry out scientific pest control. In addition, the knowledge question-and-answer system based on the knowledge graph also has a user feedback mechanism, and users can evaluate and provide feedback on the answers through the display screen. The knowledge question-and-answer system based on the knowledge graph can continuously optimize the knowledge graph and question-and-answer model based on the feedback information, thereby continuously improving the quality of the service.

[0078] Based on any of the above embodiments, the robot also includes a data transmission and communication module, which is used for real-time transmission and storage of data. Real-time data transmission between various systems in the robot is achieved through wireless communication, including receiving and sending instructions, so that the robot can perform various tasks such as path adjustment, speed control and navigation correction in real time according to the instructions, thereby improving the efficiency of the robot's intelligent monitoring and early warning of diseases and pests in apple orchards.

[0079] The data transmission and communication module is connected to the edge computing platform, and the data transmission and communication module includes a communication unit and a data storage unit.

[0080] The communication unit uses wireless communication technology to achieve efficient transmission of data between the orchard environment collection system, pest and disease identification and counting system, pest and disease monitoring and early warning system, and the knowledge question and answer system based on the knowledge graph.

[0081] The data storage unit is used to store data, for example, to store video data and image data collected by the robot through a camera through a memory card, so that it can be combined with the communication unit to achieve remote monitoring through the camera.

[0082] The data transmission and communication module can ensure the stable transmission of real-time data, support high-speed transmission and processing of large amounts of data, ensure data synchronization and information sharing between systems, and provide reliable communication guarantee for the overall coordination and efficient operation of the robot.

[0083] The present invention provides an intelligent monitoring and early warning robot for apple orchard pests. The robot traps apple orchard pests through a pest luring and capturing module, collects apple orchard data in real time through an image acquisition and processing module, monitors apple orchard pest information, performs real-time data processing and analysis through a data processing and analysis module, and realizes identification, counting, monitoring and early warning of pests and diseases as well as knowledge question and answering in combination with environmental data collected by multiple sensors. The power module and the data transmission and communication module work in coordination to provide energy and communication guarantees for the operation of the robot, thereby realizing efficient, accurate and real-time monitoring and early warning of pests and diseases in the apple orchard, and improving the automation and intelligence level of pest and disease management in the apple orchard.

[0084] Optionally, the MH-YOLO model is optimized based on the YOLOv5s model; The optimization includes at least a first optimization, a second optimization, a third optimization and a fourth optimization; The first optimization is to introduce a CBAM attention mechanism into the backbone network of the YOLOv5s model, and the first optimization is used to enhance the perception ability of the YOLOv5s model to the characteristics of pests and diseases; The second optimization is to introduce Swin-Transformer blocks into the first CSP2_1 module of the neck network of the YOLOv5s model, and the second optimization is used to enhance the information integration capability of the YOLOv5s model; The third optimization is to introduce an ASFF adaptive feature fusion module at the end of the neck network of the YOLOv5s model, and the third optimization is used to improve the recognition ability of the YOLOv5s model for pest and disease characteristics; The fourth optimization is to perform structured pruning on the YOLOv5s model after the first optimization, the second optimization and the third optimization, and the fourth optimization is used to reduce the number of parameters and complexity of the YOLOv5s model.

[0085] Specifically, the pest identification and counting system includes a pest identification subsystem and a pest counting subsystem. The pest identification subsystem can output pest detection results through the MH-YOLO model based on the real-time video streams on both sides of the apple orchard and the images of inactivated pests.

[0086] Figure 7 Schematic diagram of the YOLOv5s model structure and the MH-YOLO model structure provided by the present invention, such as Figure 7 shown. Figure 7 (a) in the figure shows the schematic diagram of the YOLOv5s model structure. Figure 7(b) in the figure shows a schematic diagram of the MH-YOLO model structure. In order to facilitate the deployment of a model for identifying pests and diseases in apple orchards on the edge, the embodiment of the present invention adopts a YOLOv5s model with low training cost and parameter count as a baseline network, and improves it on this basis to obtain the MH-YOLO model.

[0087] The MH-YOLO model is obtained by performing multi-angle structural optimization based on the YOLOv5s model, including at least the first optimization, the second optimization, the third optimization and the fourth optimization.

[0088] The first optimization is to optimize the feature extraction part of the backbone network of the YOLOv5s model and add the Convolutional Block Attention Module (CBAM) attention mechanism to enhance the model's perception of the key features of pest and disease images.

[0089] CBAM can effectively enhance the model's perception of the key features of pest and disease images by combining channel attention and spatial attention.

[0090] Channel attention is to highlight the key channel features by learning the importance of different channels; spatial attention is to enhance the feature expression of salient areas by weighting the spatial dimensions of the feature map.

[0091] The second optimization is to integrate Swin-Transformer blocks in the first CSP2_1 module of the neck network of the YOLOv5s model. Swin-Transformer is a Transformer variant based on the window attention mechanism. It achieves efficient feature extraction and expression through local attention calculation. By introducing Swin-Transformer blocks in the CSP2_1 module and realizing the fusion of features through the Add operation, it not only improves the richness of feature extraction, but also enhances the information integration ability of the model.

[0092] The third optimization, namely multi-scale feature fusion, integrates the Adaptive Spatial Feature Fusion (ASFF) module at the end of the neck network of the YOLOv5s model. The ASFF module can effectively solve the problems of information loss and redundancy in the multi-scale feature fusion process by adaptively weighted fusion of feature maps from different scales, thereby improving the model's ability to identify pests and diseases at different scales. The ASFF module flexibly adjusts the feature fusion strategy by learning the weights of features at different scales, enabling the model to perform high-precision recognition in complex backgrounds.

[0093] The fourth optimization is structured pruning. After the first, second and third optimizations, the YOLOv5s model can efficiently detect pests and diseases in apple orchards, but the structure and parameters of the model have not yet reached the optimal level, occupying more resources. In order to further simplify the model, improve the detection efficiency of the model, and facilitate embedding on the edge platform, the present invention prunes the model through structured pruning to reduce model parameters and complexity, thereby achieving the best balance between model accuracy and speed.

[0094] Figure 8 : is a schematic diagram of the structured pruning of the MH-YOLO model provided by the present invention, such as Figure 8 shown. Figure 8 In i -thconv-layer represents the 𝑖th convolutional layer of the MH-YOLO model, ( i +1)= j The -th conv-layer represents the 𝑗th convolutional layer in the network, and its number is the same as the 𝑖+1th layer.

[0095] Specifically, the process of structured pruning is to first perform channel sparsity regularization training on the initial network, then prune the channels with smaller scaling factors, and finally obtain a compact network by fine-tuning the pruned network. The channels represented by factors close to 0 after sparsification have a low contribution rate to the model, so these channels and input and output aspects can be pruned, mainly channel pruning. The essence of channel pruning is to eliminate channels that contribute little to model performance. Through channel pruning, the redundant parameters and computational complexity of the network can be effectively reduced, and a compact network can be obtained after pruning, thereby compressing the model structure and improving the computing speed without affecting the model performance.

[0096] In order to improve the network's rapid convergence and generalization capabilities, convolutional neural networks often use batch normalization as a standard method. When pruning channels, the input and output of the BN layer are and , and at the same time, a scaling factor and bias are introduced in the batch normalization layer of each channel to normalize the channel data, as shown in the formula: Where: and Represent the mean and standard deviation of each batch, Represents a very small positive constant used to avoid a denominator of 0. represents the normalized result, represents the scaling factor in normalization, Represents the bias factor in normalization.

[0097] During training, L1 regularization is used for sparseness, and the total loss function in the channel pruning algorithm is The expression is as follows: Where: Represent the input and output of the training respectively, represents the weight parameter, Represents the scaling factor penalty item. Represents the regularization coefficient, which is used to control the weight of the regularization term in the loss function.

[0098] In order to improve the reasoning speed of the model, the embodiment of the present invention adopts the TensorRT framework for deployment. TensorRT is an efficient reasoning engine launched by NVIDIA, which can efficiently deploy the trained deep learning model to the edge platform for reasoning, and can provide low latency and high throughput for the deep learning model.

[0099] Specifically, the embodiment of the present invention uses TensorRT to optimize two aspects. The first aspect is inter-layer fusion or tensor fusion. Through vertical fusion, layers with the same structure but different weights can be merged into a wider layer, that is, the horizontal layer can merge the convolution, bias and activation layers into a CBR structure, thereby reducing the calculation steps and data transmission time; the second aspect is data accuracy calibration. Since the network accuracy of the PyTorch framework training is 32-bit floating point (FP32), there are forward propagation and back propagation in the model training process, which requires higher-precision network parameters, while there is only forward propagation in the model reasoning process, and the parameter requirements are not high, and the parameter performance can be appropriately reduced. Therefore, the data accuracy can be reduced to FP16 or INT8 accuracy, reducing the amount of calculation, thereby improving the reasoning speed of the model.

[0100] The embodiment of the present invention optimizes the YOLOv5s model in multiple aspects to obtain a lightweight MH-YOLO model. Compared with the original YOLOv5s model, the MH-YOLO model has stronger perception ability, information integration ability and recognition ability of pest and disease characteristics. Moreover, through structured pruning and TensorRT optimization, the model has lower parameter quantity and complexity, and reduces the amount of calculation, so that the MH-YOLO model can improve the reasoning speed while improving the recognition accuracy, thereby enabling the robot to accurately and efficiently identify pests and diseases.

[0101] Optionally, the DeepSORT counting model includes a region recognition module, a feature extractor, a Kalman filter, a feature association module and a re-recognition module; The area recognition module is used to identify the pest and disease detection results output by the MH-YOLO model to obtain the pest and disease area; The feature extractor is used to extract features from the pest area and generate a feature vector describing the appearance features of the pest; The Kalman filter is used to predict the state of the pests and diseases based on the characteristic vector to obtain a new state estimate; The feature association module is used to generate a pest tracking trajectory based on the new state estimation; The re-identification module is used to determine the number of pests and diseases based on the feature vector and the pest and disease tracking trajectory.

[0102] Specifically, the pest identification and counting system includes a pest identification subsystem and a pest counting subsystem. The pest counting subsystem can output the number of pests through the DeepSORT counting model according to the pest detection results.

[0103] The DeepSORT counting model works together with the MH-YOLO model. The first and second cameras collect real-time video streams from both sides of the apple orchard, and the third camera collects images of inactivated pests. The MH-YOLO model first detects and classifies the pests and diseases in the video frames and pest images, and then inputs the detection results into the DeepSORT algorithm for target tracking.

[0104] The DeepSORT model mainly includes region recognition module, feature extractor, Kalman filter, feature association module and re-recognition module. The operation process of each module of the DeepSORT model is as follows: The MH-YOLO model detects pests and diseases collected by the first, second, and third cameras, and generates detection results including bounding boxes, category labels, and confidence scores; The detection results are input into the regional recognition module in the DeepSORT model to detect and identify pest and disease areas; The feature extractor is used to extract features from each detected pest area to generate a feature vector describing the appearance features of the pest; The Kalman filter is used to predict the state of pests and diseases, including their position and speed, based on the feature vector. For example, the predicted position of pests and diseases is updated based on the state information of the previous frame and the detection results of the current frame to generate a new state estimate. Through the Hungarian algorithm in the feature association module, combined with the new state estimation, the detection frame and the tracking path are associated. By calculating the association score between the current detection result and the existing tracking path, the continuous tracking of pests and diseases in the video sequence is ensured. For successfully associated pests and diseases, the pest and disease tracking trajectory is updated, including position, speed and appearance features. For the detection results that are not successfully associated, a new pest and disease tracking trajectory is generated and its state information is initialized. That is, the state of the pest and disease trajectory that has not been detected in the current frame will be updated. If it is not detected for multiple consecutive frames, its tracking trajectory will be terminated; The re-identification module calculates the similarity between feature vectors based on the feature vectors that describe the appearance characteristics of pests and diseases, determines the identity of pests and diseases, and achieves accurate tracking across frames based on the pest and disease tracking trajectory, ensuring the consistency of the identity of the same pest and disease in different frames, thereby determining the accurate number of pests and diseases; Finally, all tracking trajectory data, including the number, location, speed, feature vector and other information of pests and diseases, are stored in the edge computing platform.

[0105] Optionally, the robot further includes a walking and navigation module; the walking and navigation module includes a laser radar, a GPS module, a path planning unit, a walking mechanism and an obstacle avoidance sensor; The laser radar is used to generate a three-dimensional environment model based on the terrain, vegetation distribution and obstacle locations in the apple orchard; The GPS module is used to generate positioning information according to the real-time position of the apple orchard pest and disease intelligent monitoring and early warning robot; The path planning unit is used to generate a path planning map according to the three-dimensional environment model and the positioning information; The walking mechanism is used to control the apple orchard pest and disease intelligent monitoring and early warning robot to walk according to the path planning map; The obstacle avoidance sensor is used to detect obstacles ahead in real time when the apple orchard pest and disease intelligent monitoring and early warning robot walks according to the path planning map.

[0106] Specifically, the laser radar generates high-resolution point cloud data by scanning the orchard environment in real time, and transmits the data to the edge computing platform for processing; the edge computing platform integrates high-performance processors and GPU units to denoise, filter and align the point cloud data generated by the laser radar to generate a high-quality three-dimensional environmental model; then the SLAM technology is used, and the GPS module is used for real-time positioning and map construction to ensure the dynamic update and precise positioning of the environmental model; the path planning unit uses the Dijkstra algorithm to calculate the optimal cruising path according to the three-dimensional environmental model, comprehensively considers the terrain undulations, vegetation density and obstacle positions, obtains the optimal path information, generates a path planning map, and stores it in the edge computing platform; the walking mechanism controls the robot to cruise in the apple orchard according to the path planning map; during the cruising process, the obstacle avoidance sensor is used in combination with the dynamic obstacle avoidance algorithm to detect and avoid obstacles in front in real time to ensure driving safety, and the robot's position and posture information is updated in real time through the GPS module.

[0107] The embodiment of the present invention generates a high-precision three-dimensional environmental model by scanning and collecting data on the overall environment of the apple orchard, thereby enabling the robot to accurately identify the terrain, vegetation distribution and obstacle locations within the apple orchard, providing reliable data support for the robot's path planning and automatic navigation, and improving the robot's efficiency and safety in intelligent monitoring of pests and diseases in the apple orchard.

[0108] Optionally, the walking mechanism includes a walking drive module, a suspension module and a steering module; The walking drive module is used to provide power for the apple orchard pest and disease intelligent monitoring and early warning robot to walk; The suspension module is used to buffer the impact of uneven ground on the walking of the apple orchard pest and disease intelligent monitoring and early warning robot; The steering module is used to control the steering of the apple orchard pest and disease intelligent monitoring and early warning robot.

[0109] Specifically, the robot's walking mechanism (see Figure 4 ), including a walking drive module, a suspension module and a steering module. The walking drive module provides power for the robot, the suspension module provides shock absorption, and the steering module is used to control the robot's flexible steering in complex terrain.

[0110] The travel drive module includes four tires, axles, motors and transmissions (see Figure 4 and Figure 5 ). The motor drives the wheel axle to rotate through the transmission device, thereby driving the tire to move. Each tire is driven independently to achieve all-terrain adaptability.

[0111] The suspension module consists of springs, shock absorbers and suspension arms, which can effectively cushion the impact of uneven ground on the robot and improve the stability and comfort of the robot during walking.

[0112] The steering module includes a steering shaft, a steering motor and a steering controller. The steering motor drives the front wheels to steer through the steering shaft, thereby realizing the robot's flexible turning.

[0113] The embodiment of the present invention provides power for the robot to walk through the walking drive module, the suspension module cushions the impact of the uneven ground on the robot, and the steering module controls the flexible steering of the robot in complex terrain, thereby controlling the robot to walk smoothly in the apple orchard, allowing the robot to adapt to various complex terrains in the apple orchard, increasing the scope of intelligent monitoring of diseases and pests in the apple orchard, and realizing effective monitoring of large-scale apple orchards.

[0114] In order to further enable the robot to adapt to complex terrain such as wet, muddy or potholes when working in the apple orchard, the embodiment of the present invention also uses high-strength lightweight materials to make a chassis for the robot (see Figure 5 ), which makes it have strong load-bearing capacity and corrosion resistance, thus providing better support for the robot.

[0115] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0116] It should also be noted that the terms "target", "first", "second", etc. in the present invention are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more.

[0117] "Determine B based on A" in the embodiments of the present application means that the factor A should be considered when determining B. It is not limited to "B can be determined based on A alone", but should also include: "Determine B based on A and C", "Determine B based on A, C and E", "Determine C based on A, and further determine B based on C", etc. In addition, it can also include taking A as a condition for determining B, for example, "When A meets the first condition, use the first method to determine B"; for another example, "When A meets the second condition, determine B", etc.; for another example, "When A meets the third condition, determine B based on the first parameter", etc. Of course, it can also be a condition that takes A as a factor for determining B, for example, "When A meets the first condition, use the first method to determine C, and further determine B based on C", etc.

[0118] In the present invention, the term "plurality" refers to two or more than two, and other quantifiers are similar thereto.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent monitoring and early warning robot for apple orchard pests and diseases, characterized in that: include: Pest attracting and capturing module, image acquisition and processing module, data processing and analysis module, power supply module and data transmission and communication module; The pest attracting and capturing module is used to attract and capture pests, and to inactivate the pests; The image acquisition and processing module at least includes a first camera, a second camera and a third camera; the first camera and the second camera are used to collect real-time video streams on both sides of the apple orchard; the third camera is used to collect images of inactivated pests; The data processing and analysis module includes an orchard environmental data collection system, a pest and disease identification and counting system, a pest and disease monitoring and early warning system, and a knowledge question-answering system based on a knowledge graph; The orchard environmental data collection system is used to collect real-time environmental data of the apple orchard; The pest identification and counting system includes a pest identification subsystem and a pest counting subsystem; the pest identification subsystem is used to output pest detection results through the MH-YOLO model according to the real-time video streams on both sides of the apple orchard and the inactivated pest images; the pest counting subsystem is used to output the number of pests through the DeepSORT counting model according to the pest detection results; The pest and disease monitoring and early warning system is used to output a pest and disease early warning signal based on the real-time environmental data of the apple orchard, the pest and disease detection results and the pest and disease quantity; The knowledge graph-based knowledge question-answering system is used to generate comprehensive information on pests and diseases based on the pest and disease detection results, the pest and disease quantity, and the pest and disease early warning signal; The power module is used to provide energy required for the operation of the apple orchard pest and disease intelligent monitoring and early warning robot; The data transmission and communication module is used for real-time transmission and storage of data.

2. The apple orchard pest intelligent monitoring and early warning robot according to claim 1 is characterized in that: The MH-YOLO model is optimized based on the YOLOv5s model; The optimization includes at least a first optimization, a second optimization, a third optimization and a fourth optimization; The first optimization is to introduce a CBAM attention mechanism into the backbone network of the YOLOv5s model, and the first optimization is used to enhance the perception ability of the YOLOv5s model to the characteristics of pests and diseases; The second optimization is to introduce Swin-Transformer blocks into the first CSP2_1 module of the neck network of the YOLOv5s model, and the second optimization is used to enhance the information integration capability of the YOLOv5s model; The third optimization is to introduce an ASFF adaptive feature fusion module at the end of the neck network of the YOLOv5s model, and the third optimization is used to improve the recognition ability of the YOLOv5s model for pest and disease characteristics; The fourth optimization is to perform structured pruning on the YOLOv5s model after the first optimization, the second optimization and the third optimization, and the fourth optimization is used to reduce the number of parameters and complexity of the YOLOv5s model.

3. The apple orchard pest intelligent monitoring and early warning robot according to claim 1 is characterized in that: The DeepSORT counting model includes a region recognition module, a feature extractor, a Kalman filter, a feature association module and a re-recognition module; The area recognition module is used to identify the pest and disease detection results output by the MH-YOLO model to obtain the pest and disease area; The feature extractor is used to extract features from the pest area and generate a feature vector describing the appearance features of the pest; The Kalman filter is used to predict the state of the pests and diseases based on the characteristic vector to obtain a new state estimate; The feature association module is used to generate a pest tracking trajectory based on the new state estimation; The re-identification module is used to determine the number of pests and diseases based on the feature vector and the pest and disease tracking trajectory.

4. The apple orchard pest intelligent monitoring and early warning robot according to claim 1, characterized in that: The pest attracting and capturing module includes a capturing unit, an insect killing unit, an insect receiving plate, an insect dropping plate, a controller, a conveying crawler and a sweeper; The capture unit is used to capture apple orchard pests; The pest control unit is used to perform a deactivation treatment on pests; the deactivation treatment includes a killing treatment and a drying treatment; The insect receiving plate is used to receive the pests after killing; The insect drop plate is used to receive the pests after drying and to make a predetermined number of pests fall onto the conveyor belt through a weight sensor; The controller is used to control the opening and closing mechanism so that the pests on the insect receiving board fall onto the insect dropping board after being dried; The conveyor crawler is used to convey the pests dropped from the insect dropping plate to the image acquisition and processing module; The cleaner is used for cleaning the pests adhered to or overlapped on the conveying crawler.

5. The apple orchard pest intelligent monitoring and early warning robot according to claim 4 is characterized in that: The capture unit includes an impact plate and an insect trap light; The impact plates are placed crosswise to capture the lured apple orchard pests; The insect-attracting lamp is used for attracting pests in apple orchards.

6. The apple orchard pest intelligent monitoring and early warning robot according to claim 5, characterized in that: A sex attractant core is placed at the center of the impact plate, and the sex attractant core is used to attract apple orchard pests.

7. The apple orchard pest intelligent monitoring and early warning robot according to claim 4, characterized in that: The insecticide unit includes a high-temperature insecticide and an oven; The high-temperature insect killer is used to kill pests by high heat; The oven is used for drying the pests after the killing process.

8. The apple orchard pest intelligent monitoring and early warning robot according to claim 1, characterized in that: It also includes a walking and navigation module; the walking and navigation module includes a laser radar, a GPS module, a path planning unit, a walking mechanism and an obstacle avoidance sensor; The laser radar is used to generate a three-dimensional environment model based on the terrain, vegetation distribution and obstacle locations in the apple orchard; The GPS module is used to generate positioning information according to the real-time position of the apple orchard pest and disease intelligent monitoring and early warning robot; The path planning unit is used to generate a path planning map according to the three-dimensional environment model and the positioning information; The walking mechanism is used to control the apple orchard pest and disease intelligent monitoring and early warning robot to walk according to the path planning map; The obstacle avoidance sensor is used to detect obstacles ahead in real time when the apple orchard pest and disease intelligent monitoring and early warning robot walks according to the path planning map.

9. The apple orchard pest intelligent monitoring and early warning robot according to claim 8, characterized in that: The walking mechanism includes a walking drive module, a suspension module and a steering module; The walking drive module is used to provide power for the apple orchard pest and disease intelligent monitoring and early warning robot to walk; The suspension module is used to buffer the impact of uneven ground on the walking of the apple orchard pest and disease intelligent monitoring and early warning robot; The steering module is used to control the steering of the apple orchard pest and disease intelligent monitoring and early warning robot.

10. The apple orchard pest intelligent monitoring and early warning robot according to claim 1, characterized in that: It also includes a user interface and a control module, which are used to support information interaction between the user and the apple orchard pest and disease intelligent monitoring and early warning robot.

Images