Research And Development Of Fertilization And Pesticide Application Robot Based On Machine Vision

The intelligent agricultural robot using autonomous navigation technology will play an important role in improving the efficiency and quality of agricultural production,reducing the labor intensity,alleviating the shortage of agricultural production labor and promoting the construction of agricultural modernization.The machine vision-based fertilization and pesticide application robot developed in this paper can realize self-walking fertilization and pesticide application in the field,and remotely control the working status of the robot through wechat mini program.The main research contents of this paper are as follows:（1）Hardware system design and implementation.Hardware design provides the basis for functional implementation.According to the actual operation situation of farmland,the crawler field management machine is selected as the robot chassis of this paper,and the fertilization system and the application system are designed and developed.According to the required functions,the selection of the main control chip of the upper and lower computer of the control system is determined,and the relevant power supply circuit,relay control circuit and other hardware circuits are designed and installed.（2）Research and development of image preprocessing algorithm.In view of the influence of uneven illumination and shadow,the MSRCR algorithm is selected to eliminate the interference and enhance the image quality.The gray extraction results of each component in multiple color Spaces were analyzed,and the I₂ components in I₁I₂I₃orthogonal space were selected for image gray feature extraction,which completed image preprocessing and provided clear and non-interference preprocessing images for subsequent algorithms.（3）Research and development of visual navigation algorithm.On the basis of preprocessing images,navigation lines are extracted,navigation parameters are calculated and autonomous navigation is finally realized.The median filtering method is selected to eliminate the noise.OTSU method was used for threshold segmentation,and morphological processing was performed to further eliminate spots and holes.The maximum connected domain algorithm is used to extract the road region.On the basis of edge detection and road center point extraction,the least square method is used to fit the road center line with high precision.Finally,the navigation parameters are calculated based on the navigation line data,and the robot is controlled according to the navigation parameters,and the autonomous navigation is realized.（4）Development of control system based on wechat mini program.In order to realize the remote control of robot,a remote control system based on wechat mini program is developed.This paper describes the basic principle of MQTT protocol,selects it as the communication protocol of wechat mini program control system,and builds MQTT server based on Ali Cloud.According to the functional requirements,the function module of wechat mini program is developed,and the corresponding operation interface is made,and the remote control based on wechat mini program is realized.（5）Experimental results and analysis.In order to prove the actual effect of the study and facilitate the subsequent system optimization,experiments were carried out on wechat mini program control system,navigation parameter acquisition effect and robot navigation control system.The experimental results show that the wechat mini program control system can achieve remote control of driving,fertilization and pesticide application according to the requirements.Navigation parameters can accurately represent the current offset Angle and offset distance,and the error is small.The error of the robot navigation control system can meet the expected accuracy requirements and is suitable for the actual farmland working environment.The error analysis results show that the average error of the offset Angle is3.42°,the maximum error is 7.71°,the standard deviation is 3.39°,the system error is 5.19°,and the random error is 1.70°.The average error of offset distance is 4.69cm,the maximum error is 9.76cm,the standard deviation is 4.67cm,the systematic error is 6.15cm,and the random error is 2.30cm.The machine vision-based fertilization and pesticide application robot developed in this paper has a certain promoting significance for reducing labor intensity and improving agricultural automation and intelligence level.