Research On Camera Calibration And Stereo Matching In Binocular Vision System

People can estimate the depth of object by the difference between two images captured by the left and right eyes.Similar method is used in binocular stereo vision system,i.e.,imitating human eyes with two cameras.Binocular stereo vision is widely used in many fields,such as stereo film / TV,virtual reality,3D modeling,robot and navigation.In particular,binocular stereo vision can fill up the gap of the traditional measurement methods in the visual measurement system.Typically,binocular stereo vision consists of camera calibration,image rectification,stereo matching and 3D reconstruction,where camera calibration and stereo matching are the key technologies of binocular stereo vision.Monocular camera technology is not suitable for distance computation and 3D reconstruction anymore,because when only one camera is used in the scene to shoot at a fixed angle,it will lose some information in the actual scenario.In the case of monocular camera,the influence of optical illusion can be avoided by using binocular technique.In addition,binocular camera technology can be used to calculate the depth of field information of the scene,and restore the three-dimensional information of the reference image.Camera calibration is one of the most important part of binocular stereo system.Both the accuracy of the final 3D reconstruction and depth of field measurement are affected by the accuracy of calibration.The camera calibration is used to measure the internal and external parameters of the camera.The former is the geometric parameters of the camera,while the latter is the position of the camera relative to the space.However,how to solve the required parameters? Many methods have been used to solve the parameters,such as nonlinear calibration method,linear calibration method and two step method.Recently,the most widely used method is Zhang Zhengyou calibration method which is a kind of two-steps method proposed by Dr.Zhang Zhengyou.Shooting is the process of projection three-dimensional to two-dimensional.In contrast,stereo matching is a process which obtain the disparity map containing 3D spatial information by two 2D images.That is,stereo matching is the inverse process of shooting.Due to the influence of illumination,exposure time and the difference between cameras,it is difficult to ensure that the same points in the left and right images have the same gray level.Therefore,it has been proved that Census transform has its own advantages in stereo matching.However,the way that uses Census transform to weaken the luminance information is too rough,and it does not make full use of the information in the image.Therefore,there are some problems in the Census transform,such as the difference between the similar regions and the noise sensitivity.In this paper,a long coding strategy is proposed,called L-census transform.Taking full advantage of the local information in the window,the L-Census transform also makes use of the luminance information,and the noise is reduced to make the transformation process more robust.Further,a stereo matching method based on L-Census transform is proposed.The experimental results show that the performance of the proposed method is superior to some commonly used stereo matching methods.