Based On Image Feature Extraction Algorithm Design And Application

Along with industrial intensive development, surface production continued to grow substantially. In production, to quality of surface processing of high levels of demanding, especially the consistency requirements, on the image recognition technology has put forward higher requirements.This paper studies the surface image preprocessing, feature extraction, classification and recognition technology, pattern recognition problem, traditional filtering algorithm in the suppression of noise at the same time, also on the image edges and details are relatively large damage, so that the edge of image and blur detail problem, put forward the weighted directional smoothing algorithm. And in the image segmentation on the fusion of several segmentation methods, put forward a kind of improvement based on biorthogonal wavelet transform multiresolution image fusion method based on fusion and wavelet transform and morphological edge detection algorithm, optimization of the segmentation effect, for the follow-up feature extraction is to lay a good foundation.Finally, using the" license plate character recognition experiment" to image classification to achieve. This article from the color features are described in terms of. According to the parameters of the distribution, selection of less standard deviation parameters as classifier input feature vector; and using the principal component analysis method for feature extraction, texture feature dimension reduction, elimination patterns of correlation among features, highlighting their differences, meet the needs of the recognition layer.The experimental results show that: according to the image texture features and color features, using digital image processing technology, to solve the problem of image segmentation and recognition, is effective way.