Kernel Density Estimation for Dynamic Scene Modeling

A kernel density estimation (KDE) based on a multimodal model is presented for dynamic scene reference frame maintenance and update problems. A diversity sampling schem is proposed to choose a new sample set from the image sequence including moving objects. Using more popular and diversiform intensity samples, a Gaussian KDE is built to estimate the background model and to detect moving objects by background subtraction. The diversity sampling samples describe the key information of the original whole sample set and avoid the repetition computation in the evaluation phase. Compared with the whole samples based on algorithm, the proposed approach is proved to be efficientive in traffic surveillance systems.