FPGA based mixture Gaussian background modeling and motion detection
Xuejiao Li, Xiaojun Jing · 2011
Motion detection is a technology that can extract moving objects from a sequence of frames. This is the enabling component for many important applications, such as security monitor, vehicle detection and human activity analysis. Out of many common motion detection algorithms, mixture Gaussian background modeling can perform more accurate results and requires relatively less computation when processing static background. However, as the resolution of the frames increases and real-time processing requirement is proposed, sequential processor can't finish the computation in time. In this paper, a full pipelined and parallel Gaussian background modeling and the whole motion detection system are proposed on Altera Stratix IV FPGA. Due to the parallel architecture, the system can process real world 1024*1280 video at more than 30 frames per second, which is the real-time requirement, and the system can achieve the same accuracy as the software version on experimental datasets.