Occlusion Estimation in 3D Point Clouds using Visual Data from Home Care Scenarios

Today, home care monitoring systems are implemented more frequently, which is mainly due to the increasing number of elderly people and the reduced number of medical staff. This makes it all the more important that these systems are reliable and safe for its users. In this paper, we propose a risk management feature in form of an Occlusion State Index (OSI) based on the overall occlusion within an observed scene that can be integrated into an in-house developed smart home care monitor prototype. To implement and evaluate the proposed feature, visual data is acquired within a home care test scenario using two Kinect 2.0 depth cameras. After preprocessing, the recorded depth information is merged into a point cloud from which a scene occlusion map is computed based on the ray box intersection algorithm and the fast voxel traversal algorithm. Finally an OSI is computed depending on the amount of occluded voxels in a 3D point cloud.