Study on wavelet transformation-based low illumination & high dirt face detection algorithm

The paper¹, above all, establishes a face database conforming to environmental features of low illumination and high dirt. Wavelet transformation is utilized in recognition algorithm so as to establish a weight-based cascade classifier by Harr features extracted from the picture. Law 8 serves to adjust weight between levels of cascade classifier. Pictures of actual coal miners are used as samples for training in the paper, to establish an initial classifier. The method is applied to face and eye recognition to obtain xml-based documents of classification features for face and eye recognition with good experimental effects. It has a high face detection rate. Besides, prototype system design based on face and eye recognition of video flow and picture is accomplished.