With the progress of technology and the improvement of the quality of life,the public's demand for mobile robots is constantly improving.Vision-based dynamic obstacle avoidance and path planning is an important branch of the intelligent mobile robot.It determines whether the robot can complete the given task and reach the specified destination.The research focus of this paper is mainly in the operating environment where the environment is known and the obstacles are unknown.Based on the information collected by the visual sensors,the robot plans the driving path,detects the obstacles and avoids the dynamic obstacles to complete the given task.According to the subject,this paper studies from the following aspects:Firstly,studying the path planning algorithms of mobile robot.Several traditional path planning algorithms are compared and analyzed.According to the results of each algorithm in the simulation,the algorithms are compared and summarized one by one.Secondly,due to the constraints of large displacement,motion occlusion,light intensity change and other factors,the traditional optical flow detection algorithm will produce large errors in dynamic obstacle detection.Based on the research of traditional optical flow detection algorithm,this paper proposes a new optical flow detection algorithm based on multi frame fusion.On the basis of two frame detection,multi frame image fusion reduces the optical flow error caused by motion occlusion,increases the reliability of effective optical flow,and improves the accuracy of large displacement optical flow detection algorithm combined with pyramid algorithm.Thirdly,the optical flow detection algorithm can only get the location information of obstacles,and cannot achieve obstacle avoidance.The information(Time To Collision,TTC)is introduced to calculate the distance between the robot and the obstacle.By comparing with the pre-set threshold,we can judge whether the robot needs to design the obstacle avoidance scheme.This paper mainly uses the A* algorithm for path planning.During the path planning,obstacle detection is carried out according to the proposed optical flow method based on multi frame fusion.Through comparing the TTC,it is determined whether the robot needs to avoid obstacles.According to the results of optical flow detection and TTC information,a robot obstacle avoidance strategy is proposed.When the robot does not encounter obstacles,it can drive to the target point according to the path planning;when the robot encounters dynamic obstacles,it can analyze the obstacle avoidance strategy of the robot.According to the two schemes of general obstacle avoidance and emergency obstacle avoidance,the robot can avoid moving obstacles until it reaches the target point.In this paper,through simulation experiments,when the multi frame fusion optical flow detection algorithm is used to detect obstacles,the detection of obstacles at large displacement and moving occlusion is more accurate,and the average end-point error of obstacles detected is smaller EPE.In MPI data set,compared with two frame optical flow detection algorithm and a single pyramid algorithm,the error value is smaller and more closer for real optical flow estimation. |