Research On Robot Binocular Navigation Algorithm

With the development of modern computer technology and machine intelligence,Research on robotics has received extensive attention.Navigation technology is a key technology in the field of robotics,It is the key to achieving robotic automation and intelligence.So the progress of navigation technology is directly related to the development of robotics technology.Binocular vision navigation is the research trend of robot navigation.It has the advantages of rich information and high level of intelligence.Binocular can sense the three-dimensional information of the robot's movement environment.According to the environmental information returned by the camera,the robot can avoid obstacles and eventually arrive at the target position.Binocular visual navigation is mainly three aspects,including binocular vision obstacle positioning,path planning and motion control.The research content of this paper is the binocular obstacle location algorithm and path planning algorithm.The work done by this article is as follows:(1)The binocular obstacle location algorithm was studied.Binocular positioning algorithm includes camera calibration,image feature extraction,stereo matching and optimization,depth information recovery and other content.Studied the internal and external parameters of the camera using Zhang Dingyou calibration method,as well as the image distortion correction and stereo correction.The performance of SIFT algorithm and ORB algorithm in terms of scale transformation and algorithm efficiency are compared.Finally,ORB algorithm is selected to extract image features.In stereo matching,matching is the key to binocular positioning.It is also the most difficult content.In the application system,it needs both the accuracy of matching and the speed of matching.The paper uses the ratio of the Hamming distance of the nearest neighbor and the Hamming distance of the second nearest neighbor as the matching criterion,which can reduce the mismatching point to some extent.Using RANSAC algorithm to remove the wrong matching point,thereby improving the accuracy of the matching point.Besides,this paper improves the ORB matching point search method,and proposes a data structure using K-D tree and BBF search strategy.In depth information recovery,the principle of depth information recovery was studied.(2)The path planning algorithm was studied.In the research of path planning algorithm,the artificial potential field method is used to implement the robot path planning after analyzing and comparing several path planning algorithms.Learned the principle,advantages and disadvantages of the artificial potential field algorithm.The artificial potential field method includes two problems,namely,the target inaccessibility and the local minimum problem.For the not reaching the target location.,this paper proposes a modified repulsive force function,adding the distance factor from the robot to the target point.For the local minimum problem,this paper proposes the method of using escape force.(3)The binocular vision obstacle location algorithm was implemented and analyzed.Through the ORB,BBF search strategy,nearest neighbor/second nearest neighbor matching criterion,and RANSAC matching optimization algorithm,the three-dimensional coordinates of the matching point in space are restored.This gives the distance between the camera and the nearest obstacle.And the positioning error is within an acceptable range.(4)Artificial potential field algorithms are implemented and analyzed.The Matlab and Visual Studio platform analyzes the target inaccessibility and local minima of the traditional artificial potential field method,but the improved artificial potential field method solves them.The paper realizes the path planning of the improved artificial potential field algorithm in different complicated environments,and the planned path is smooth and safe.