Face Recognition Method Based on Adaptive Feature Fusion
Pengfei Yan, Zhongmin Zhang · 2022
The rapid development of computer software and hardware technology has laid a very solid foundation for computer vision tasks such as target recognition that use images as input carriers. The purpose of this paper is to study face recognition methods based on adaptive feature fusion. The principle of Adaboost algorithm and the training process of strong classifier and the cascade process of strong classifier of Adaboost algorithm are studied. The algorithm of face detection is improved through Haar-like feature extraction and weight update during algorithm iteration. At the same time, in order to detect the face more accurately and quickly, the method of skin color detection is introduced to extract the candidate face region, and an initialization model that can adapt to the change of the target face is proposed based on the previous invariant average initialization model. Then, the Adaboost algorithm is experimentally tested through the face database, network pictures, and video images combined with the previous image preprocessing method. The detection rate of the face database can reach 97.56%. The test results show that the algorithm can meet the design requirements.