Research Of Flat Image Recognition Based On Artificial Neural Network

With the popularization of technical application of electronic computer, digital image technology is widely used in all kinds of fields of industrial production, agricultural production and daily life day by day. Image storing & transmission technology has already been very ripe in digital image processing field, and has been used extensively. While compared with the former, the technology of image analysis and understanding is far behind neither in theory, nor in application. Image recognition as the key problem of image analysis and understanding area has being the studying focus and difficult point both at home and abroad all the time. After years of hard working of the researchers, considerable progress has been made in this filed, but there are many kinds of problems such as insufficient flexibility, too narrow scope of application, bad stability and so on. Further more, there are many kinds of theories and each has large difference from the others and all kinds of theories have not a solid or unified basis.The idea about this paper comes from observation to flexibility and stabilization vision behavior of human. In this paper, the advantages of human vision is introduced in the image recognition field and combined with general digital image processing technology to make the best of their advantages and make up their disadvantages. In this way, we can find a flexible, universal and stable algorithm of flat image recognition.This paper is divided into six chapters. In the first chapter, the background and significance of our research is introduced. In the second chapter, the history and researching trend of image recognition are introduced briefly. In the third chapter, the principal of image recognition and the structure of a basic image recognition system are expatiated. The subsequent chapters are ranged by the orders of the phases which compose of the whole processing course of image recognition system. At first, in the image preprocessing chapter, the subjects such as image noise eliminating edge detection, image binarization are discussed. In the same time every kinds of algorithmare tested by their experiments and each experiment result is compared with the others carefully. Secondly, in the chapter about image feature extraction, the necessary characteristics of the image feature used in image recognition are presented and the concept of general moment and geometrical moment are defined exactly and invariable moments have been construct subsequently and their characteristics have been proved by experiment. Following this is the chapter of pattern recognition based on BP network, it discusses the theory of neuron network using in pattern recognition and the studying and training of BP network. Then, using BP network, it performs the pattern recognition experiment of the invariable moments feature of the image extracted last stage. And the results prove the validity of the algorithm of flat image recognition. Finally it summarizes the whole research work as well as the prospect of the project, and points out some feature that need improved in future.Artificial Neural Network; Pattern Recognition; Image Processing; Moment Feature; Perceptron...