Efficient Multi-Layer Background Model on Complex Environment for Foreground Object Detection

This paper proposes an establishment of multi-layer background model, which can be used in a complex environment scene. In general, the surveillance system focuses on detecting the moving object, but in the real scenes there are many moving background, such as dynamic leaves, falling rain etc. In order to detect the object in the moving background environment, we use exponential distribution function to update background model and combine background subtraction with homogeneous region analysis to find out foreground object. The system uses the TI TMS320DM6446 Davinci development platform, and it can achieve 20 frames per second for benchmark images of size 160 × 120. From the experimental results, our approach has better performance in terms of detection accuracy and similarity measure, when comparing with other modeling techniques methods.