| Face recognition technology has already reached a new height after several decades of growth,and has been applied to various fields.Face recognition technology is mainly through the extraction of facial features,which will be compared to the identity authentication,due to the advantages of safe and convenient,not easy to lose and so on,people pay more attention to it.Because of the interference of some factors(illumination,occlusion,etc.)in the feature extraction process,the extracted features are not all good,so the effect is not ideal in the classification recognition stage,and the recognition rate is low.In order to improve the recognition rate,proposing two algorithms on the basis of the wavelet domain.The main contents are as follows:The first,a face recognition algorithm based on sparse representation of wavelet domain is proposed to solve this problem when the face image is affected by illumination or small occlusion in this paper.The wavelet transform is used to decompose the human face,the dictionary containing the multi-frequency information of the human face is established,and the multi-frequency subband is sparse.The result is calculated by calculating the effect of the multi-frequency subband in the multi-frequency dictionary.Experiments were performed on the extended Yale B and AR face database to improve the recognition rate of the image.Second,when the feature extraction of the image is extracted,the extracted feature is different,but the smaller feature not only increases the computational complexity but also affects the recognition.Therefore,this paper proposes an improved particle swarm and selects the excellent features.Firstly,the image is pre-processed,and the discrete noise is used to remove the image noise and reduce the dimension.Then,the MPSO algorithm is used to select the image which is more conducive to recognition.Finally,the nearest neighbor classifier is used to classify the face image.Experiments show that the recognition rate under MPSO algorithm is improved.Finally,the application of face recognition in real life is discussed in combination with the method proposed in this paper,which reflects its engineering application value. |
