A Low-Cost Embedded Stereo Vision System for Real-Time applications using raspberry pi A Low-Cost Embedded Stereo Vision System for Real-Time applications using raspberry pi

Abstract

stereo vision is a widely applicable technology for extracting depth information.  Depth is obtained in the form of disparity map from a pair of stereo images of the observed scene, which is captured using a camera pair. The technique of inferring depth information is called stereo matching and it is the key element in stereo vision. There are different stereo matching algorithms  which are broadly classified into global algorithms and local algorithms. In this paper stereo vision system based on semi global block matching algorithm and its low cost implementation is proposed. It results in a standalone real_time stereo vision system for low cost and high performance embedded applications. Moreover, the efficiency of the algorithm is further improved by using additional pre and post processing steps. The proposed stereo vision system can be used in autonomous navigation, robot vision and object tracking etc.