Research On Feasible Area Detection And Path Planning Of Construction Vehicles Based On Machine Vision

Construction vehicles are widely used in mines,railways,ports and other fields,and they also play an important role in disaster relief.In addition,construction vehicles generally work in the field or in construction areas with harsh environments,and there will be unpredictable obstacles such as gravel,mounds,and construction waste on the ground.Its safe and efficient operation has very high requirements for the driver,and it is not even suitable for the driver to operate in some dangerous areas,so unmanned construction vehicles are imperative.Aiming at this situation,The thesis is combined with the National Natural Science Foundation of China project "Electromechanical coupling dynamics and adaptive control of multi-track walking device"(No.51775225),this paper studies the feasible area detection and path planning of construction vehicles based on machine vision to improve the intelligent level of construction vehicles.This article summarizes the research background and research significance of unmanned construction vehicles,and summarizes the current research status of feasible area detection and path planning at home and abroad,introduces the common methods and application fields of feasible area detection,and discusses the common use of local path planning algorithm.Based on the ZED binocular camera,the feasible area was detected.First,calibrate the ZED binocular camera and preprocess the image before feasible area detection,and then perform stereo matching on the preprocessed image to obtain the disparity map.For the obtained disparity map,the depth information of the image is obtained by triangulation.Then we processed the 3D point cloud by extracting the normal vector,removing the outliers and smoothing the point cloud based on bilateral filtering.Finally,a random sampling consensus algorithm is used to extract the base level,and the feasible region is judged based on the threshold.The processed feasible area detection image has a good detection effect.Use the loader in the construction vehicle as a carrier to conduct a feasible area detection test,and conduct real-time detection of the feasible area in front of the loader in the campus.Detecting feasible areas in environments with large obstacles(such as pedestrians,vehicles,etc.)and small obstacles(such as snowdrifts,etc.),the detection effect of feasible areas in the external environment containing large and small obstacles is good.Using some strategies to improve a series of problems existing in the traditional dynamic window method of local path planning,so that the speed,stability and smoothing effect of construction vehicles path planning have been improved.The simulation verification of local path planning was carried out on the MATLAB simulation platform.The experimental results of the simulation show that the improved dynamic window method has a good obstacle avoidance effect for both static obstacles and dynamic obstacles.