Image Recognition And Classification Based On Artificial Neural Networks

The final goal of digital image processing is to know an image and target the part that people are interested in by the computer, and this is the main function of computer matrix recognition. The matrix recognition technology is to simulate human's competence of recognition which mainly covers visual and audial competence. The image matrix recognition is to build direct communication between the computer and the outside world in recognizing and processing information like words, images, pictures, and scenes.The image recognition consists of three steps: data recognition, data processing, and data classification. The four ways of image recognition are statistical matrix recognition, structural matrix recognition, hazy image recognition and intelligent matrix recognition. As a generalized mode of intelligent matrix recognition, the artificial neural networks which has become popular since the 1980s had the feature of parallelism, distributional memory, good fault tolerance, auto-adaptation, associative memory, and good non-linear processing, has achieved many achievements that could not be reached by many other conventional ways in the domain of matrix recognition. And the image recognition technology of the neural network is a newly developed image recognition technology together with the modern computer technology, the image processing, the artificial intelligence, and the matrix recognition theory. It is a new way of image recognition which combines the neural network algorithm and the traditional matrix recognition.This paper first introduced several commonly-used ways of image recognition based on neural network; then it proposed the adoption of the BP network and the radial basis function to be the image recognition models of neural network. This paper mainly covers the following four aspects:1. briefly introduced the primary principles of image preprocessing technology and image recognition;2. introduced several popular neural network models and matrix recognition methods;3. conducted a systematic analysis of the structure of BP neural network and the BP algorithm on the basis of which proposed the application of BP neural network in image recognition;4. finished the recognition and the classification of remote-sensing images by using neural network models of the radial basis function (RBF), and contrasted this method with other ones.The paper finally indicated the actual application of the image recognition in real life: the recognition and classification of a medical image based on the digital recognition method of BP neural network as well as the neural network model of RBF. The result of the test proved that the image recognition and classification method based on artificial neural network the writer proposed here has much value in practical application. At last, the writer summarized his research work and further analysed this image recognition technology as well as its future development.