Digital Image Classification System Based On BP Neural Network

With the development of computer science and artificial intelligence and so on,computers have taken on human beings’ work in more and more fields, and led to manyhotspot science subjects, including computer vision. Computer vision uses machines tosimulate human being’s visual system, capturing images with cameras, by processing andanalyzing the images, machines can get the pattern information that images hold, in thisway,they can make their own decisions. Computer vision can be applied to industrialproducing, medical diagnosis and supervisory control and many other fields, which saveshuman beings from a lot of hard workings.Computer vision is a very comprehensive subject, which involves knowledge aboutimage processing、 pattern recognition and so on. By using knowledge of imageprocessing, we can apply operations like noisy-filtering and image segmenting to images,in this way, we can amend image defects and simplify images. By using patternrecgonition knowledge, we can make machines simulate the way human beings think,analyze and understand images intelligently, extract their characteristic values tosynthesize and compare,at last,make decisions for human beings.In this procedure,patternrecognition is a very important part, and also very hard part too. Because things varyvery much, their characteristic values can be very different from each other.Characteristic values of images can be extracted after careful examination, and then canbe used to train machines, so that machines can judge and classify images correctly andefficiently later.Pattern recognition procedures include information extraction、 preparationoperations、characteristic values selection and classification. In this paper, I adapt BPneural network, which is a supervised intelligent neural network, by learning some cases,it can keep some kind of “memory”, and use it to classify. There are many kinds of characteristic values, like euclidean distance and probability distance and so on. Later inthis paper, I will take hand gesture classification and stem cells classification as examplesto show how to examine different patterns of images and choose correspondingcharacteristic values to classify them. The result of the classification shows that thetheory works very well.