Stereo vision for webcams

Abstract

This project describes the developmental process of creating and testing a webcam based image capture platform and library for determining the suitability of using webcams for stereo vision. Research and development has taken place in the field of stereo vision for many years, however the costs involved with capture devices and computer hardware has limited the access to the field. This has led to the requirement of a low-cost image capture system that is capable of operating multiple capture devices simultaneously for the application of stereo vision. The image capture software was developed in the Microsoft Windows operating system environment using the Microsoft DirectShow application programming interface to access and capture video frames. The Camera Calibration Toolbox was used to perform webcam calibration and image rectification to the captured images. Stereo processing was then applied with the Dense Stereo algorithm developed by Abhijit Ogale. The stereo disparity results were analysed for accuracy and precision to determine the suitability of the webcam for stereo vision applications. The outcome of the development and testing does confirm that webcams can be operated simultaneously and that they can provide a suitable platform for stereo vision. This outcome will increase the accessibility into the research of stereo vision, without the excessive costs that have previously been associated in this field.