A data processing method to improve the accuracy of depth measurement by binocular stereo vision system

Abstract

After coarse calibration of the binocular 3D vision system, there still are large systematic errors existed. In order to improve the measurement accuracy, we directly calibrate the existed systematic errors after simple calibration steps of the binocular vision system. The depths of the targets located at different distances are measured at three different field of views(FOVs). After got original results, the relationship between the depth errors, the field of views and the distances is analyzed. The error curves for correction are obtained. The largest error is about 26% and after the correction the error is below 5%. Also a simple matching program is designed to automatically match the feature points at left and right images, and to calculate the depth of the feature points.