Simulated Lidar Images of Human Pose using a 3DS Max Virtual Laboratory.

摘要

Large sets of 3D Simulated LIDAR (Light Detection and Ranging) images in the form of partial point clouds of human poses shapes were created to support the investigation of activity recognition using machine learning techniques. To produce these large sets of point clouds in an efficient and repeatable manner, a virtual laboratory environment was developed in Autodesk 3DS Max, with an animated, biofidelic 3D human mesh biped character (avatar) as the subject. The biped animation modifies the digital human model through a time sequence of motion capture data representing an activity cycle. A point cloud is captured by ray tracing the sensor-facing side of the body model at each time frame. Currently, the entire collection of point clouds consists of 68 human subjects performing five actions - digging, jogging, picking up, throwing, and limping.