Research On Vector Location And Map Construction Based On Vision Motion In Regular Environment

Mobile robots slowly into people's daily lives,which is a long-standing dream of mankind.With the development of the times,science and technology also developed rapidly,in complex and harsh environment,people also expect mobile robots to safe and normal work.It is particularly important to quickly create a map in a complex and unknown environment and to map itself to the map.This achieves the dream of the autonomous mobile robot.In an unknown environment,if you want to achieve mobile robot autonomous movement,you can use synchronous positioning and map building method(SLAM).In general,the map is already a priori knowledge,and the structure of the environment is the traditional mobile robot research sites.Positioning and mapping are two aspects of SLAM,the basis of which is positioning,but the accuracy of the mapping is affected by the positioning,the two aspects are complementary and interdependent.In the twenty-first century,SLAM is the research hotspot of the world's scientific research workers,for which a lot of solutions.This thesis mainly discusses the rules under the environment of mobile robot localization and map building problem.Paper first chapter introduces the meaning of this topic background,the development history and research status,and also simply introduces the monocular visual SLAM the research,the construction of the basic theory,common map method;The second chapter USES the SIFT feature point extraction and matching algorithm,detailed analysis of the principle and deduction process of SIFT algorithm,inorder to verify the validity of the SIFT algorithm and real-time,SIFT algorithm is verified by experiment in this paper the scale zoom,different perspectives,and light intensity has invariance.The third chapter discusses the monocular vision mobile robot SLAM system model,established the SLAM based on extended kalman filter as the main algorithm,and build the experimental platform for mobile robot based on monocular vision is used to SLAM at the back of the test.Thesis mainly study in detail the monocular visual SLAM the coordinates of the model,state model and observation model and noise model,and construct the inertial navigation of mobile robot motion model,and also verify the feasibility of each model.The fourth chapter chooses ilf camera calibration method of camera calibration,calibration of the camera's internal and external parameters.Experimental analysis and odometer error,road signs,using MATLAB to extend kalman filter SLAM for 2 d and 3 d simulation,verify the feasibility of the algorithm and model.