Research And Application Of Navigation Technology For Tomato Disease And Pest Identification Robot

Agricultural robots have become an important part of modern agriculture,and its application fields are expanding.With the development of technology and cost reduction,agricultural robots have gradually become popular and have been applied to agricultural production,bringing great convenience and benefits to agricultural production.Due to the features of automation,refined operation and completing a huge number of agricultural production tasks in a short period of time,robots are widely used in agriculture.In practice,robots can be applied to cultivation,harvesting,irrigating,pest and disease identification and weeding,thus improving the efficiency and quality of agricultural production.Among the many agricultural fields where robots are used,tomato cultivation is one of the important one.Since tomatoes are grown in a large area in China and are susceptible to various diseases and pests,it is important to study tomato pests and diseases to improve the efficiency and quality of agricultural production.To address this issue,this paper combined artificial intelligence and robotics in order to address the challenges of pest control and monitoring in tomato cultivation and provide more options and support for the development of modern agriculture in China.The main research contents of this paper are as follows:(1)An agricultural robot system was designed based on the need for intelligent tomato pest identification in the field,and the system as a whole consists of hardware architecture and robot software development.(2)In terms of navigation,a multi-sensor fusion approach was applied to solve the robot positioning problem,using the GY-85 nine-axis degree of freedom IMU sensor and encoder to feed two data into the ROS in one data frame through the serial port to achieve accurate robot positioning.The map was constructed by sensing the environmental information in real time through the laser radar,and the advantages and disadvantages of the Extended Kalman Filter algorithm and the particle filter algorithm were analyzed to determine the most suitable algorithm for the required environment in this paper to improve the accuracy and thus build a more accurate map.(3)In terms of path planning,this paper optimized the A~* algorithm based on the analysis of its advantages and disadvantages,improved the problem of excessive inflection points of the A~* algorithm by increasing the inflection penalty,and improved the problem of large search volume and large computation of the A~* algorithm by integrating the idea of bounded suboptimality.In addition,the effectiveness of this paper on the optimization of the A~* algorithm was demonstrated by conducting comparative tests on a raster map.The effectiveness of the DWA algorithm was demonstrated through matlab to test the DWA algorithm,which is able to readjust the path until it reaches the target point when it suddenly encounters an obstacle.In the overall path planning,the improved A~* algorithm was utilized as the global path planning algorithm and the DWA algorithm as the local path planning algorithm.(4)Pest detection was accomplished with YOLOv3.The required weights were obtained through training dataset,and the weight files in the darknet＿ros functional package were replaced with pest identification weight files.The study showed that this pest detection system can capture the images detected by the camera in real time and classify the detected tomato leaves.The system can detect healthy leaves as well as seven common tomato pests and diseases such as early blight,late blight,and leaf mold.(5)To meet the functional requirements,we integrated the developed navigation system and conducted a series of experimental tests on the agricultural robot navigation system.Through continuous adjustment and optimization,the final test results show that the agricultural robot is able to accomplish safe and collision-free autonomous navigation functions in complex and unknown space environments.At the same time,we also achieved the real-time monitoring function of the robot on the mobile terminal using Websocket technology.