56results about How to "Improve face recognition rate" patented technology

The invention relates to a sparse representation face identification method based on constrained sampling and shape features, belonging to the field of image processing. The method comprises the following steps: extracting the textural features and shape features of face images in a training set based on constrained regions to obtain a textural feature matrix and a shape feature matrix, wherein each category of the textural feature matrix and the shape feature matrix respectively corresponds to the multiple face images of one person in the training set; and comparing each category in the training set with the obtained textural feature vector and shape feature vector of the face image to be identified, and taking the category in the training set corresponding to the maximum comprehensive similarity value as the identification result of the person to be identified. The method has higher face identification rate.

The invention provides a face recognition method based on a convolutional neural network, comprising: performing necessary preprocessing on the face image to obtain an ideal face image; selecting the ideal face image as the input of the convolutional neural network Enter U0, the output of U0 enters UG, and the output of UG is used as the input of US1; the S neuron of US1 extracts the edge components in different directions in the input image through supervised training as the first feature extraction and outputs to the input of UC1 ;The output of UC1 is used as the input of US2, and US2 completes the second feature extraction and is used as the input of UC2; the output of UC2 is used as the input of US3, and US3 completes the third feature extraction and is used as the input of UC3; the output of UC3 is used as US4 US4 obtains the weights, thresholds and the number of neuron cell planes of each layer through supervised competitive learning as the input of UC4; UC4 is used as the output layer of the network, and outputs the final network determined by the maximum output result of US4 Pattern recognition results. The invention can improve the face recognition rate in complex scenes.

The invention discloses a face recognition method based on structuralized factor analysis. The problem that the face recognition accuracy rate is low due to the fact that local clustering features and a global distribution structure of face image data cannot be kept in the prior art is mainly solved. The face recognition method comprises the achieving steps of 1 dividing a face image data set; 2 performing clustering analysis on all training sample data; 3 calculating an optimal feature projection matrix through Gibbs sampling; 4 extracting low-dimension features of all test sample data and the training sample data; 5 recognizing a face image. The face recognition method synthesizes the local clustering features and the global distribution structure of the face image data and improves the face recognition accuracy rate.

The invention relates to a removing method for a glasses frame in a face image. The method comprises the steps of: detecting and positioning eyes in the input face image with a pair of glasses, geometrically correcting the face image on the basis of the positioned result to form a first image; extracting an eye estimation area from the first image to obtain a second image, and conducting mathematical morphology processing on the second image so as to determine the position of a glasses frame; based on the first image, establishing a face skin color model; in the second image, interpolating each pixel point of the glasses frame the position of which is determined, correcting the interpolation result in a part of pixel points of the glasses frame according to the skin color model, and conducting de-noising processing to form a ninth image; and replacing the eye estimation area in the first image by the ninth image, synthesizing and outputting the face image the glasses frame of which is removed. According to the technical scheme, the glasses frame in the face image with a pair of glasses can be conveniently and effectively removed, and the face recognition rate can be further improved.

The invention discloses a face recognition method for face images with lower occlusions. For face images to be recognized with lower occlusions and several face image samples without occlusions, feature images are extracted respectively. The extraction method is as follows: extracting and obtaining The image of the face area is input into the convolutional neural network, and the feature map output by the last layer of convolution layer is used as the initial feature map of the face image, and each initial feature map is multiplied by the weight matrix to obtain the feature image. The value of the weight element corresponding to the upper part of the initial feature map in the matrix is ​​larger, and the value of the weight element corresponding to the lower part of the initial feature map is smaller, and then calculate the feature image of the face image to be recognized and the weight of each face image sample The similarity of feature images completes face recognition. In the present invention, the area weight processing is performed on the feature map of the convolutional neural network, so as to strengthen the unoccluded feature and improve the face recognition rate of the lower occluded face image.

The invention discloses a selective integrated face recognition method based on genetic algorithm fusion differential evolution. Firstly, the HOG feature of the face image is extracted, and then the PCA algorithm is used to reduce the dimension of the face image, so as to reduce the computational complexity. Finally, use After dimensionality reduction, the GADESEN algorithm is used to classify and identify the data. This method takes the support vector machine as the base classifier, extracts N samples from the original training set with replacement, iterates T times according to this method, and uses the sample set generated each time to train the base classifier model. The classifier encodes real numbers to generate an initial population. In the mutation operation, the difference vector is used to guide the mutation and then produce high-quality individuals. The crossover operation uses the parent individual and the mutant individual to generate crossover individuals, which increases the diversity of individuals. Retention strategies for genetic evolution.