Extraction from surveillance video based on face detection

A fast method for locating face-contained subsequences from surveillance video was proposed in this text. In the algorithm, only half of the video frames were chosen to be processed by dividing them. First, the video was divided into several sections according to time; then, the video frames were filtered by random algorithm for the second time; third, detect faces from the filtered frames using Adaboost algorithm; At last, it returned a flag frame sequence according to each detection result. The experimental results have impressively indicated the effectiveness of the method.