Real-Time Stereo Matching System.

Getting depth information by stereo matching is one of the key steps in 3D reconstruct. In many practical applications, there are high requirements for the speed of processing and the accuracy of the results. Many algorithms have obtained good results in processing precision, like SGM. However, the processing speed often does not meet the real-time requirements. In this paper, we improve the traditional stereo matching method SGM so that it is able to meet the real-time requirements. We implement our improved algorithm in TX2 with parallel programming. And in the experiments, it shows that our algorithm obtains 21 fps for the video gained by ZED camera size of 640 * 360 pixels, 32 disparity levels and using 4 path directions for the traditional SGM method. To measure the accuracy of our method, we use Middlebury dataset as indoor scene and the video obtained by ZED camera as outdoor scene to exam our algorithm separately. The results show that we get great balance between the speed and the accuracy.