Object Detection and Localization Using Stereo Cameras

Camera systems have become increasingly popular because cameras are cheap and easy to deploy. Compared with other depth cameras, the stereo camera is small, and it is easily carried by subjects. Through a fixed baseline, the stereo camera is able to compute depth information. However, the traditional stereo matching algorithm can not compute the depth information on the edge of the image. Meanwhile, due to the large amount in the 3D point cloud, there is no specific numerical relationship between semantic information and depth information. In order to solve this problem, estimating depth and semantic information in an accurate way is required. A deep neural network model is used to predict semantic information and depth at the same time. Further, we propose a robust method to deal with variation brightness and improve performance under actual conditions.