LPP-Adaboost Based Face Detection in Complex Backgrounds

摘要

For the difficulties of the high dimensions and the complex environment in the video surveillance, a face detection method based on Locality Preserving Projections (LPP) incorporated with Adaboost is proposed. The LPP based features are extracted to reveal the local manifold structure of the faces using linear method. Then, the weak classifiers are constructed on LPP features according to the minimum error rate. Finally, the Adaboost algorithm combines all trained weak classifiers into a strong classifier in order to improve the recognition rate. The experiments are accomplished on the face databases and the video surveillance. On the CAS-PEAL database, the experiment results show that our method has good performance under the different expressions, the different illuminations, and the different pose samples. At the same time, the LPP-Adaboost based face detection algorithm is also satisfied with the requirements of the detection rate and the detection speed in the real complex backgrounds.