Research On Indoor Positioning And Navigation Technology Of Mobile Robot Based On Global Vision

In recent years,science and technology is developing faster and faster,with the vigorous development of our country's artificial intelligence and intelligent manufacturing industries,whether in the ascendant industrial production or in people's seemingly ordinary daily life,the existence of mobile robots has indispensable significance,especially in the fields of military,medicine,industrial production,intelligent warehousing and logistics,the application of mobile robots is much too common.Positioning navigation plays a vital role in the research field of mobile robots.It mainly involves three issues,namely,to clarify the initial position coordinates and posture of the robot,the target position coordinates and posture,and the planned path between the two coordinate positions.However,the traditional positioning and navigation technology has a small field of view,insufficient adaptive ability to dynamic unstructured environments,and many shortcomings in stability and accuracy,which greatly hinders the development of intelligent mobile robots in industrial production.Therefore,this paper proposes and designs a positioning and navigation system based on global vision and using a CCD monocular camera as a sensor to collect and process surrounding environmental information.The visual navigation system mainly includes three parts: image stitching,map construction,and path planning of mobile robot.The research work of this paper is detailed as follows:First of all,in view of the small field of view and limited visual range of the monocular camera,this paper uses image stitching technology to provide a large-scale scene map for the visual navigation of mobile robots.This paper creatively improves the feature points extraction part of image stitching algorithm,in which the Harris corners extraction is divided into two parts: rough extraction and fine extraction.By reducing the number of feature points and increasing the number of effective feature points,the accuracy of registration can be improved,the loss of stitching time can be reduced,and the stitching and fusion between images can be well realized.Secondly,in order to solve the problems of low positioning level and low accuracy of traditional grid map,this paper proposes a grid mapping method based on dynamic grid division.This method is based on fixed-size grid cells.Based on this method,the local grid is quartered and densified,and the boundary of the obstacle can be accurately extracted,so as to provide an accurate and stable prerequisite for the subsequent path planning.Finally,this paper compares the advantages and disadvantages of several common path planning algorithms such as A* algorithm,artificial potential field method and dynamic window approach,and conducts simulation experiments respectively.The experimental results show that the path curve planned by the dynamic window approach is smoother and the trajectory is optimal,and the path planning effect of the robot is the best.On this basis,the characteristics of the robot are identified and extracted,and the navigation experiments are carried out.The experimental data shows that the visual navigation system designed and proposed in this paper can not only meet the actual needs within the effective accuracy range,but also has practical significance to a certain extent.