Research On Vehicular Binocular Vision Measurement System Based On Dynamic Baseline Adjustment

With the continuous development of computer vision technology,binocular stereo vision has been widely used with its advantages of reliable precision,simple structure and low cost.The mobile car equipped with stereo vision system can realize the perception of the surrounding environment with high precision and fast,and has a wide range of applications in the fields of patrol and security,corridor inspection,exploration of unknown environment and so on.However,in today's more and more complex working environment,the traditional baseline fixed on-board binocular measuring device is increasingly limited in the use of the scene: in the close measurement(such as alignment inspection),the object is too close or biased to one side,resulting in the object is not in the binocular vision system of the public field of view,unable to image matching;If the object is too far away from the lens,it is not in the depth of field,which leads to the image blurring and pixel extraction error.At different distances,the baseline will produce fluctuating measurement errors,etc.The car equipped with fixed baseline binocular vision needs to adjust its position until it meets the measurement conditions.The process is complex and sometimes restricted by the environment.For example,the car cannot be adjusted to an appropriate posture in a narrow space,so the measurement cannot be carried out.In addition,the measurement results obtained by selecting different baseline values are also different,indicating that the size of the baseline will have an impact on the measurement accuracy of the binocular stereo vision system.In order to make the car can adapt to a variety of working conditions of information measure and improve the accuracy of measurement,this paper studies the baseline distance can dynamically adjust the on-board binocular stereo vision measurement system,the system by adjusting the on-board the baseline distance of binocular vision,makes the car without moving or rotating,efficient and convenient to make the system satisfy the measurement conditions,at the same time corresponding to different working distance to adjust the right baseline,reduce the baseline measurement error.Specific work contents and innovations are as follows:(1)Aiming at the lens distortion problem caused by adjusting the baseline distance of binocular camera,this paper analyzes the camera imaging model,obtains internal and external parameters of the camera through calibration,corrects the lens distortion,completes the stereo correction of image pair,and analyzes the error.The accuracy model of binocular stereo vision was established,and the effects of camera focal length and system baseline on the measurement accuracy were analyzed.The relationship curves of the measurement errors of the baseline in X and Z directions are given,and the baseline adjustment strategy to achieve the highest accuracy in different scenarios is obtained.The relationship between baseline and field of view and depth of field of the system is discussed,and the reference range of baseline adjustment is given.By adjusting the baseline,the system can reach the highest accuracy in the X and Z directions,and the object can be guaranteed in the public field of view and depth of field without moving the car,thus improving the pixel extraction accuracy.(2)The key techniques and some algorithms of stereo matching are studied,their advantages and disadvantages are analyzed,and parallax graph is generated to compare the effects.On the basis of local stereo matching algorithm,aiming at its low matching accuracy near image edge and image discontinuous region,through image edge extraction,feature points are screened in advance,and then an adaptive image matching window is established to screen the selected pixel points twice.The matching cost of all the pixels in the window is calculated by using the algorithm of gray value and difference,and the optimal parallax value is obtained,which is compared with the parallax map generated by the traditional fixed matching pane,so as to improve the matching situation of image edge and image discontinuous area,and reduce the parallax map void.(3)Physical development and experimental platform construction: This paper designed and developed the physical object of vehicle-mounted binocular vision system with dynamic baseline adjustment,and carried the system on the car;At the same time,this paper developed an experimental platform based on the PC terminal,which can realize the car movement control,binocular vision baseline adjustment control,remote image acquisition and other functions.By displaying the collected images on the PC terminal in real time,the accuracy experiment of multiple baseline adjustment was carried ou.