Research on Automatic Recognition of Color Multi Dimensional Face Images Under Variable Illumination

The illumination changes and environmental noise cause facial recognition accuracy will be reduced, in order to effectively solve this problem, need to be multidimensional variable light color face image automatic identification. But with the current method for automatic face image recognition, slow convergence speed, easy to fall into local minimum, the existence question of error identification. For this, put forward a based on particle swarm optimization neural network of variable light color multidimensional automatic face image recognition method. The method using wavelet transform to obtain first face image of low frequency component, using two-dimensional differential analysis (2 dlda) algorithm to extract the face image linear differential characteristics of low frequency component, using particle swarm optimization and BP neural network for classification and recognition and in ORL face database verify the feasibility of the proposed method. Experimental simulation show that the proposed method the identification accuracy is higher, effectively reduces the influence of uneven illumination in face recognition.