Design and Implementation of Binocular Vision System with an Adjustable Baseline and High Synchronization

Abstract

In this paper, we present a baseline-adjustable and highly synchronous real-time binocular image acquisition system which is easy to realize. Binocular stereo vision is gradually applied to the field of robot navigation, industrial detection and virtual reality etc. However, most designs are not flexible in practice due to fixed baseline and they have longer production cycle and higher costs. We propose a novel solution by using an adjustable baseline to acquire binocular image with robust hardware part and parallel software part. Cameras can capture binocular images triggered by external trigger pulse generated via Raspberry Pi in 20 Hz with synchronization error in microsecond level. Since our system is highly synchronized and portable, easy to deploy and low-cost. It is more convenient to generate binocular datasets for SLAM experiments. Finally, we testify the design via ORB-SLAM2 system.