CamMirror: Single-Camera-based Distance Estimation for Physical Analytics Applications

Distance estimation is key to many physical analytics applications in settings such as driving, shopping, and more. The goal is to tell where an object or person is. While specialized sensors such LIDAR and stereoscopic cameras can solve the problem, these tend to be expensive. In this paper, we present CamMirror, which performs distance estimation with a single camera. The key idea is to use a pair of carefully-positioned mirrors to provide a second view of the scene akin to what a second camera would have provided, which then enables disparity-based ranging. We present the design of CamMirror and two applications, one on vehicle ranging and the other on smart shelf.