Face Recognition Based On Improved PCA&Fisher

As an important branch of Biometrics, face recognition has been widely used in identityrecognition, public safety, machine vision and other areas. In the sixties of last century,researchers began to focus on the face recognition,because it has the property of non-contact,friendly and easy to get. As an interdisciplinary technology, face recognition includs imageprocessing, pattern recognition and computer vision, it is a cutting-edge topics in thesefields. Face images are typically high-dimensional nonlinear datas, it is susceptible tointerference of light, facial expressions, posture, age, and other effects. In order to improvethe recognition accuracy, researchers still have many challenges to face.Firstly, paper introduces the research background and the main method of facerecognition, it summarizes the technical difficulties that exist in face recognition,and prophecethe development trend of face recognition technology.To solve the problems of PCA, thispaper combine it with Fisher.It achieve the target of data’s dimensionality reduction twice. Onthe basis of PCA+Fisher, the paper has improved the algorithm to further improve theaccuracy of face recognition.In this paper, the study is based on the ORL face images, this database consists of40individual400facial images. In these photos, people’s gestures, facial expressions have somechanges, some people are also wearing jewelry. Because of the scale changes in20%, it fullyreflects the different manifestations of a person, so the results can be well explained thisalgorithm’s practicality. Paper implements the face recognition based on improved PCA&fisher to the400images,the high recognition rate prove that this algorithm is effective. In thepast papers, about the face recognition, they often concerned with the overall recognitionresults, but ignore the recognition results of each photo.To this problem，the paper establish arepository of recognition results, so we will make a specific analysis about every photo, evenwe will provide the basis for algorithm’s improve or the image preprocessing.