The Key Techniques Research Of Feature Extraction Based On Complex Background

Feature extraction of the target under complex background is an important developingdirection in the field of digital image processing, In many areas digital image processing andthe machine vision system are using widely and requiring strictly, which increaseddifficulties on image processing.In complex background, how to find the target isparticularly important which brings the visual system great impact or be needed to payspecial attention. And in many aspects of image processing, the precise contour extraction isthe most important and the most basic march. People can only do further research on thebasic of getting precise outline image with textured target.This paper is researching the key techniques research of feature extraction based oncomplex background, the specific work are as follows:1) The physical construction、stress and deformation are analyzed in the new type offlexible four roll; According to the field investigation and system analysis on the rollingprocessing principle and system analysis, it is concluded that extract the characteristicparameters is important during the process of the machine vision system in the planktreatment shaping.2) According to the target for processing image, the image processing modules in thismachine vision system include image preprocessing, image recognition on complexbackground and feature extraction. Image preprocessing module is using double linearinterpolation method to do image scale, then using mixed noise filtering method processingimage noise, as follows with threshold segmentation method or based on watershedtransform or c-means clustering method to Segmentation images, finally using morestructure for edge detection algorithm to target image edge detection. Image recognitionmodule is using RHT algorithm for round detection. Feature extraction module is puttingforward rotating scanning contour extraction of multi-objective based on contour mark andcontour filling to extract rapidly the target image outline, Then using cluster of line ofthinking with the principle of three extraction to extract the characteristic parameters ofround target.