Research On Binocular Ranging System Based On OpenCV And CNN

Humans get most of the information through their eyes,and the eyes obtain information of the world efficiently.Therefore,the machine that has a human-like vision can obtain external information more quickly and efficiently.The development of machine vision has important theoretical and practical significance.Machine vision is an important part of our implementation of artificial intelligence.Machine vision has the advantages of wide measurement range,high measurement accuracy and low measurement cost.Machine vision has a wide range of applications.Image processing is an important part of machine vision.Machine Vision can get distance information from the image taken by camera.It processes the information of the image by the technologies of image segmentation and feature extraction.Binocular vision is a branch of machine vision.Binocular vision can get the three-dimensional information of an object from two pictures.Binocular ranging is an important branch of binocular vision.The internal and external parameters of the camera are obtained by calibrating the camera.Then the images are processed.We propose training a convolutional neural network on pairs of small image patches where the true disparity is known.Stereo matching includes global stereo matching and semi-global stereo matching.Semi-global stereo matching is faster and more accurate than global stereo matching.We need to compute the stereo matching cost.We proceed with a number of post-processing steps to achieve good results.These steps use cross-based cost aggregation,and a left-right consistency check,and apply two filters to optimal disparity map.After performing a three-dimensional reconstruction of the object,the depth information and distance information of the object are all obtained.The convolutional neural network we examine is the GAO-CNN architecture.The disparity map and distance information are all obtained by using the model.The system is written in Python language.The parallel processing function of the Nvidia graphics card can increase the speed of the ranging system.Our disparity map is better than the disparity map obtained by the BM algorithm and the SGBM algorithm in the low-textured area.