Human motion tracking using 3D camera
Publish date: 2009-12-31
Report number: FOI-R--2828--SE
Pages: 87
Written in: English
Keywords:
- human tracking
- detection
- 3D camera
- LIDAR
- LADAR
- RIM
Abstract
The interest for techniques for people surveillance has increased in the recent years. Video surveillance is a technique that can be used to monitor people at big events such as football games or protest meetings, shopping centers and prisons. The videos are then analyzed in the search for potential abnormalities such as fights, steeling or escapes. When tracking a person in a crowd there is a risk of loosing the person or associating him/her with another person. If a 3D imaging sensor, so-called 3D camera, is used, the risk of confusion might be reduced as all distances are known. 3D cameras have been on the market for only some years and there is a rapid development going on in this technology. The sensor is capable of delivering measurements for all three spatial coordinates of the camera scene. These cameras can perform 3D imaging at video rate. Since they have been on the market for the last few years only, their possibilities and capabilities have not been fully discovered. One of the areas that need to be investigated is tracking of humans. This master thesis has been performed at the Informatics Unit at the division of Information Systems, FOI, Linköping. The work aimed at investigating a 3D camera and evaluating its performance when used in applications of tracking of one person. One goal with the thesis is to determine whether the tracking algorithm can run more seldom when range measurements are available compared to when range measurements are missing. The tracking performance in 2D and 3D is evaluated on both simulated data and data collected with a 3D imaging ladar camera. Comparisons are made in the sensor coordinates. It is concluded that there is no large difference in tracking error when the estimated image coordinates are compared. The small angle of view of the camera, relatively large target range and small target velocity are considered to facilitate the tracking with the 2D camera. However, the 3D tracker demonstrates very good range estimates which makes it well suited for tracking in world coordinates.