Recovering Information Of Worn Campus Intelligent-cards Based On Image Recognition

With the wide application of campus Intelligent-Cards, as a function to identify personal identification of certificate, it has been applied for several aspects in campus life. During the use of campus Intelligent-Cards, however, whose surfaces are usually inevitably affected with more or less wear, thus resulting in the application of personal identification that might not function. To solve this problem, this dissertation studies ways to recover the information of worn campus Intelligent-Cards by some correlation techniques of image recognition.Image recognition is a research area that covers a wide range of subjects and still has broad prospects. Combining with the characteristics of surface details of campus Intelligent-Cards, this dissertation presents separately from face recognition and digit recognition aspects. In this dissertation, firstly, we introduce the technique for face pre-processing, worn images processing to the same extend by average filters and median filters. The experimental results prove that the median filter is more effective in denoise processing. Then this dissertation introduces some common ways in face recognition, especially principal component analysis. During the system implementation, firstly, the dissertation makes denoise processing for face images of worn campus Intelligent-Cards by using median filters, and processes the original images which are too dark or too light through histogram equalization. Then we obtain the Eigenface space which consists of a set of features extracted by principal component analysis (PCA). Finally, Euclidian distance classifier was utilized in face recognition. To reduce the recognition scope and focus on student numbers of campus Intelligent-Cards, the dissertation introduces the technique of number preprocessing and several common methods. In the binarization stage, an efficient algorithm is applied to the same image. Through comparison analysis of the experimental results, we present a better way that combines the Otsu algorithm and the Bernsen algorithm to deal with the original student number with noises. On this basis, Thus student numbers and recognized using the algorithm based on feature template matching. For reducing the experimental results, the system recognizes the student number in the images which obtained from the face recognition.The information recovery system for worn campus Intelligent-Cards presented in this dissertation, realizes the combination the optimal technique of pre-processing and recognition, and pays more attention to the whole system design in order to obtain best recognition ability.