Face Detection Based on BPNN and Wavelet Invariant Moment in Video Surveillance

A multi-view face detection method for Video surveillance is proposed in this paper. It provides the ability to extract high-level features in terms of human activities rather than low-level features like color, texture and shape. The method is capable of locating human faces over a broad range of views in color image sequences or videos with complex scenes. Firstly, an improved frame difference is used to acquire promising regions of the image. Then it uses the presence of skin-tone pixels to locate faces. Finally, an improved method based on wavelet invariant moment and BPNN is used to verify the candidate face regions. The experimental results show that the proposed algorithm has high speed and low error-detection rate, so it can be used in the real-time video surveillance system. The main distinguishing contribution of this work is being able to detect faces irrespective of their poses by using the wavelet invariant moments as input of the BPNN, whereas contemporary systems deal with frontal-view faces only. The other novel aspect of the work lies in its accuracy of acquiring the candidate area to segment objects from background with the help of motion information and skin information.