A Mirror-based Compact Monocular Depth Estimation System for Environment Sensing

Environment sensing is one of the most challenging problem in robot sensing. Among the sensing systems, stereo camera systems provide dense information and usually is less expensive for environment sensing. Based on this, a monocular stereo system for robot environment sensing is proposed. The system combines a saccade mirror with an industrial camera and captures images of environment when the mirror rotates. With the prior of relative posture of cameras corresponding to each image, the stereo images can be easily rectified and the depth information of surroundings is estimated. The estimation accuracy is validated by experiments on small resin model and real-life indoor scene.