Research On Heterogeneous Face Recognition Of Sketch/Photo And Across Age Image

Face recognition has been one of the most concerned studies in machine vision, pattern recognition and image processing. The application fields of this research is very wide, such as security authentication, identification, monitoring and control system, the credit card validation, the human-computer interaction control, certificate verification, access control and so on. In recent years, heterogeneous face recognition as a recognition of the special type also got some attention, and its research mainly in the photograph and sketch images, near/thermal infrared and visible images, photographs and video images, different ages of face images research, etc. This paper focuses on photograph and sketch images of heterogeneous face recognition, and different ages of face images research. Photo/sketch recognition mainly used in the detection of cases, safety and security, especially in the detection of cases that can help investigators locate the suspect or narrow range. Different ages of face images research mainly applies in public security forensic, missing population survey, etc.In photo and sketch recognition, the paper studies the two kinds identifying methods. One is transformation of photos and sketch before recognition. Namely, first, transform the photos into pseudo sketch based on the feature extraction method, and then use the traditional method for sketch/pseudo-sketch recognition. When transforming a photo image into sketch, the algorithm is more rational and effective use of the characteristics of the image, also it reduces the difference between the sketch and photo image, improves the matching effect, and reduces the recognition time. The other is not transformed but recognition directly with their photos and sketches. To identify forensic sketches, our paper uses Histogram of Oriented Gradient (HOG) feature descriptors to represent both sketches and photos. Then, we get this feature space with Null space-based LDA (NLDA) after making "slices" of feature patches. Finally, the sketch image is recognized in feature space. HOG feature descriptors can not only diminish the intrapersonal variations between the sketch and photo modality but also maintain sufficient information for interclass discrimination.Different age (namely across age) heterogeneity face recognition, the direction of the research mainly has three aspects, the face age simulation, face age estimation and face verification. This paper mainly studies across face verification. First, the algorithm extracts the relatively stable characteristics HOG feature in the process of aging, difference space operation is used in feature vector of images to get difference within the class and difference between the classes; Difference within classes as a positive sample, difference between the classes as a negative sample, then the samples are sent to support vector machine classifier to classify, and then judge whether the human face is the same person.In this paper, the main work is some exploration and attempt based on the existing heterogeneous face recognition algorithm. Experimental results show that the algorithm has a better stability and accuracy.