A Fast Stereo Matching Algorithm using Interpolation for Sewer Inspection Robots

In this paper, we present a fast stereo matching algorithm for sewer inspection robots. Conventional algorithms such as the sum of squared difference (SSD) and cross-correlation based on segmentation, graph cuts and so on have been proposed. However, most of them are computationally expensive. In order to solve the problem, we propose a fast stereo matching algorithm using interpolation in this paper. In our algorithm, an image is initially divided into many blocks. Then, a matching measure combining SSD and the cross-correlation is calculated only at the corner points of each block. A downsampled sewer disparity image is composed of only the corner points. Because the disparities change continuously in the sewer environment, we use the cubic interpolation for extending the downsampled sewer disparity image into a regular-sized disparity image. The experimental results show that our algorithm outperforms the conventional algorithms in both accuracy and speed in the sewer environment