Research On Local Path Planning And Tracking Of Field-road Carrier Based On Machine Vision

The transportation of agricultural materials and agricultural products on the fieldroad is an important part of agricultural production in hilly and mountainous areas.In the present,motorcycles,trolleys and other tools are often used for field transportation,which causes low intelligence,high operating cost and low operating efficiency.With the development of navigation technology,navigation methods like visual navigation are more and more used in agricultural production,which promotes the level of agricultural production mechanization,automation,intelligence.According to the requirements of field road transportation in hilly and mountainous areas,using autopilot field-road carrier through machine vision technology,the paper did research on image identification,local path planning and tracking on the field road.Field-road carrier uses monocular visual navigation system,which is equipped with CMOS color camera as image acquisition equipment,Raspberry Pi as processor,servo as the implementing agencies to complete the steering operation.Base on this platform,the main research contents are:1.Most of field roads in the hills and mountains are unstructured roads.In the field of feature extraction,the preprocessing algorithm and image segmentation algorithm are studied.In the image preprocessing,the homomorphic filtering illumination compensation technique is used to eliminate the influence of the illumination on the road recognition.The weighted average method performs the gray scale of the image.The median filter technique eliminates noise on the image recognition.Based on the maximum interclass variance method and the minimum error rate Bayesian decision method,the image is divided into two parts: target and background.And we uses the morphological filtering method to remove the noise in the target.And aiming at the road shadow problem,researchers make a study on the shadows' color characteristics and removal methods.2.In the extraction of navigation parameters,Hough algorithm has high recognition accuracy and real-time difference,the least squares method is high in real time and is susceptible to noise interference.This paper proposes a new route selection algorithm based on image target centroid.The algorithm is used to extract the centroid of the partitioned image.Then fit the centroid as quadratic curve by the least squares method.The curve which can be called as virtual raod-centerline is the real-time navigation path for the carrier.3.The controller of the visual navigation is designed to achieve local path planning and path tracking.The controller adopts the preview strategy and the fuzzy control method.That is,two points are selected on the navigation line in front of the carrier,the lateral deviation is calculated as the input of the controller.And the fuzzy control rules are compiled,so the servo of carrier can be turned to complete the path tracking.We do the visual navigation experiment on the field road.Results of experiment shows that the path planning and tracking algorithm is suitable for field road.In the road tracking experiment,when the road width is 1.2m,the maximum deviation rate is 22.95% and the average deviation rate is 6.36%.In the shadow,water stains,the maximum deviation rate of 30%,the average deviation rate of 8.82%.The carrier doesn't be off the road in the experiment.It is said that the carrier can complete local path planning on the field road.