Research On Vehicle Stereo Positioning And Navigation Based On Computer Vision

Vehicle and vehicle queue control in traffic system has very important application value in real life. It is one of the hotspots in the research of domestic and international to use computer vision to locate and track the vehicle in traffic system. It is also a very challenging subject.This paper aims to build a real traffic simulation experiment environment to reduce the cost of carrying out experiments with real vehicles and Avoid unsafe factors. Use computer vision technology for vehicle positioning and tracking and navigation control to avoid the additional sensor consumption, thereby reducing the cost of testing. Compared with other sensors, the computer vision has stronger anti-interference ability and better robustness.Traffic system is a complex three-dimensional environment, vehicle location tracking and navigation in the traffic system has the following challenges:(1) It is very complicated in real traffic environment, in vehicle detection there are many interference factors, how to target vehicle tracking and positioning is a challenge.(2) In non open interior space, most of the existing vehicle position detection methods use the relative position location, and the vehicle can not get the global position information in the environment. GPS signal in the indoor environment is interferenced, use which vehicle detection method becomes a problem.(3) There are a lot of interfering objects in the background, which have a lot of impact on the target tracking and positioning. How to use computer vision technology to separate foreground object and background is a challenge.(4) Most of the existing technical solutions can only get the two-dimensional plane position of the vehicle. Today, the three-dimensional transport system, such as the overpass and the viaduct are gradually increasing. How to get the depth information of the vehicle to distinguish the position of the vehicle in the three-dimensional traffic system is a problem to be solved.(5) In the case of vehicle sensor assisted, the vehicle can do operations such as carring out with the queue and running along the lane line. But how to use the global position information to plan the shortest path and reach the destination has more practical application value. Navigation in the interior of the vehicle has become a challenge.In view of the above challenges and aiming at the indoor traffic simulation test environment, this paper mainly focuses on the research of moving target tracking and navigation control in the three-dimensional space:(1) Two kinds of plane localization methods are proposed, which are based on the class of two-dimensional code feature mark detection method. Another method is based on template matching and CAMSHIFT combination. Two kinds of methods can be used to detect the two-dimensional plane position of the feature mark in real time. Positioning and tracking experiments are carried out by using two kinds of detection methods. Second kinds of characteristic marker detection method has higher efficiency, the system uses second methods to obtain the two-dimensional plane position of vehicle.(2) Based on the theory of binocular stereo vision system, two cameras are calibrated and the binocular stereo vision system is established. The internal parameters and external parameters of the system are obtained by experiment, and the 3D position of the vehicle is obtained by using the weight projection matrix and disparity information derived from the calculation.(3) In this paper, a control strategy of double layer navigation for mobile vehicles is proposed. In the first layer, the global path planning is carried out, and the second layer is based on the sub target point for navigation control. The experimental results show that the control strategy is practical and scalable, and can be used in a variety of indoor environment. When the vehicle is deviated from the original formulation of the optimal route, the new optimal route can be re established according to the current position, and the vehicle is accurately guided to the destination.