Participants: Shuyang Sheng, B. Keith Jenkins
In this work our goal is to combine shadow and illumination detection in 3D point clouds.
Shadow and illumination information are among the most well known “hints” used in the process of human object recognition, and can also be extremely useful in various computer vision tasks. With good initial estimates of shadow and object, we can estimate illumination information and thereby improve object recognition and pose estimation results; on the other hand, illumination and object information can guide shadow detection as well. We are working on better ways to extract and utilize shadow and illumination information, and expect the whole process to be an enabling preprocessing step of further scene understanding algorithms.
For current progress, please refer to the project Github: