Design And Implementation Of Environment Perception System In Visual Navigation

With the rapid development of China’s economy,car ownership has increased sharply,but the supporting infrastructure is not perfect enough.There has always been a phenomenon that car owners often get lost due to the lack of satellite signals in the indoor parking lot,resulting in difficulties in parking and getting out of the lot.However,the existing solutions are mainly guided by signage,which is of low intelligence.In addition,the positioning method of vehicles is mostly based on indoor wireless communication technology,which is easy to be interfered and the cost of transformation is too large,leading to its difficult application.Therefore,in order to solve the parking difficulties and low efficiency of indoor parking,this thesis designs a vehicle positioning and navigation system.The following research work is carried out.1.Build a parking lot navigation management system based on the visual positioning method.The system consists of two subsystems,the server terminal and the vehicle terminal,which realized the functions of visual positioning,navigation guidance,allocated parking space and quick positioning out of the lot.Among them,the positioning method is realized based on environment perception and multi-point pose estimation technology.The features of the surrounding environment(number)are collected by the onboard camera,and the P3 P algorithm is used to calculate the central position of the vehicle based on the coordinates of multiple feature points.Compared with the existing solutions,the system saves the cost of equipment installation and maintenance,reduces the computational complexity of the algorithm,which can solve the problems of inaccurate positioning and parking difficulties in the indoor parking lot at low costs,and provides convenience for car owners.2.Design and implement the environment perception system in visual navigation.Firstly,this thesis designs the keyframe extraction scheme that takes frames at a fixed distance according to the speed and extracts useful frames based on the histogram,and takes experiments to proves that this method can effectively extract image frames with different numbers.This scheme solves the problem of redundant frames and useless frames in the video acquisition and processing stage.Secondly,this thesis takes the numbers as the identification feature,compares the color differences between the numbered area and the background area,and realizes the number detection by the two-time threshold segmentation method,which effectively detects the numbers.The validity of the proposed method is verified by experiments on the collected data.Finally,this thesis uses the Tkinter to design the system,which realizes the functions including registration and login,image acquisition and identification feature detection.The environment perception system explores the feasibility of visual perception in the complex parking lot and lays a foundation for the development of visual navigation.