Automatic Recognition Technology Of Cannonball's Defect

Automatic recognition technology of cannonball's defect is mainly studied in the paper. What in the paper mainly is use method of digital image processing and image analysis to process and analysis the X-Ray image of cannonball getting under the X-Ray . Wavelet transform, edge detection of image processing and analysis, feature extraction is involved in the paper. Edge detection is an important step in digital image analysis. The wavelet transform is an important tool in the domain of mathematic and signal processing. The wavelet transform has excellenttime and frequency feature, which has a progressive application in image edge detection. Feature extraction is the first step in digital image analysis.Chapter 1 is introduction, represented the advance of study task, automatic recognition technology of cannonball's defection include edge detection, feature extraction and pattern recognition, and solving project of the system. The basic theory of wavelet transform, resolution analysis of wavelet transform, the fast Mallat algorithm of dyadic discrete wavelet transform and the theory of using the wavelet maxima representation detected the edge of image is given in chapter 2. Finally is examination, use the wavelet maxima edge detection algorithm to detect the edge of several X-Ray image, and contrast with the result of using traditional image segmentation. A conclusion is drawn that the result of wavelet maxima edge detection is better than the result of traditional image edge detection obviously. The descriptionof digital image feature, theory of invariable moment and the format of Hu invariable moment are mainly studied in chapter 3. And it is explained that the Hu invariable moment has not scale variation in discrete situation. So it derivate a new invariable moment based on Hu invariable moment. The new invariable moment ha translation, rotation, scale-variation in discrete situation. Finally is emulation used computer. We use Hu invariable moment and the new invariable moment exact several images' feature. A conclusion is drawn that the result of the new invariable moment is better than the result of Hu invariable moment. The basic theory of neural network and the back propagation neural network are represented in chapter 4.