The purpose and emphasis of studying this research topic is to solve the problem in the field of image recognition utilizing the preponderance of RBF neural network at the aspects of information processing and pattern recognition.The technology of image recognition is a kind of new-style technology, which is developed due to the theory of present computer technology、image processing、artificial intelligence and pattern recognition. Its primary research content is to study the classification and description of some objects or process. To develop the system which could process some information automatically for achieving image classification and identification is image recognition\’s purpose. The field of investigation in the image recognition is far-ranging, for example: the license plate recognition in the system of traffic monitor; identify the spare parts in the machine processing; find out the ailing cells via medical image; identify the forest、lake and some other specific facility through the remote sensing images; sorting the letters from post system; the recognition of face and fingerprint; signature identification and so on. The image recognition\’s basic task is identifying one or more objects from the images via image analysis. Around this core topic, this paper discusses the image pre-processing、feature extraction and pattern recognition based on neural network.By many ways, image pre-processing eliminates the noise the images takes, cuts down the features which are unrelated to recognition goal and enhances the feature information the research needs. The routine methods consist of the image enhancement、image edge detection.The image features are diverse. In the phase of feature extraction, the thesis talks about the image feature category、image feature\’s representation and description and emphatically the extraction and quantification of moment feature, preparing for the pattern recognition.As a kind of pattern recognition based on the image information, image recognition almost follows all the characters the pattern recognition possess. Due to some special characters as follows, artificial neural network takes more precedence than the traditional pattern recognition: highly parallelisnu distribute storage、good fault freedom、adaptability and associable memory、robustness、highly nonlinear processing ability. Among those ANN, the feed-forward networks\’ ability of classification and pattern recognition is better than the other ANN.The radial basis function neural network, which is a new-style、available feed-forward network, is based on the local reaction of brain\’s neuron cells looks on the outside world. The RBF network possesses the advantages, such as the simplify of structure、strong global approximating、fast-speed and convenient training methods. All these advantages make the RBF network used widely in the field of pattern recognition.Using the RBF network to carry out the pattern recognition of image features, first of all, find some solution to conform the most important parameters: activation function centre, control parameter, weight value. And adjust these parameters constantly via network training. The recognition result proofs the availability of the ways in the field of image recognition based on ANN. Besides, the paper summarizes this research、deeply analyzes this kind of image recognition technology, finally , makes a outlook for the research in the future.

Related posts:

1. Research and Application of OMR Based on Image Recognition Technology
2. Studies on Classification and Recognition of Leukocyte Microscope-Image
3. Research of Recognition on Inverse Synthetic Aperture Radar Image
4. Research and Application of the Semantic Based Image Retrieval Technology in Medical Image System
5. Research on Content Based Image Retrieval Technology