Dissertation > Excellent graduate degree dissertation topics show

Textile appearance based on machine vision defect detection and quality evaluation studies

Author: WangZuo
Tutor: ZhaoDaXing
School: Hubei University of Technology
Course: Mechanical Design and Theory
Keywords: Machine Vision Textile Appearance Defect Detection Defect classification
CLC: TP391.41
Type: Master's thesis
Year: 2009
Downloads: 126
Quote: 3
Read: Download Dissertation

Abstract

Product quality based on machine vision technology automatically detects more and more attention for the realization of product quality high-speed, high-precision automatic detection provides an important way. This paper focuses on the appearance of textiles based on machine vision defect detection technology launched, focused on the system architecture and hardware for the system to achieve the main impact of the textile image processing algorithms, on this basis, developed and developed a textile appearance defect detection prototype. First, different widths for textiles, high-speed high-precision testing requirements, considering the textiles in the image acquisition, processing, defect classification and other characteristics. Especially for the detection of wide fabric surface, distributed system structure is conducive to the wide surface of the textile fully detected and the system can effectively reduce the computational load. By analyzing the presence of noise and image textiles texture of digital image processing in linear filtering method, is designed based on Gaussian kernel adaptive filters. The filter application is better than traditional linear smoothing filter, not only to achieve an effective fabric image noise, but also reduce the target area on the image blur and facilitate the subsequent processing. To achieve an image in a periodic textile texture to find defect area, a kind of adaptive first-order differential operator and morphological gradient image edge enhancement method based on statistical characteristics of high-precision adaptive threshold segmentation algorithm. After, we propose a defect based on variable scale scanning Fast clustering algorithm for image preprocessing to reduce the problem of loss after the defect. Then, according to the common types of fabric defects, we propose a combination of defects typical spatial characteristic quantities based on spectral characteristics of textile defect detection methods and textile appearance defects eigenvectors as a classification criterion, based on improved BP neural input to network classifier textile appearance defect classification, defect classes to get the target area, and finally according to China National Standard GB / T 17759-1999, International Standard ISO / DIS 8498-1990 such as the establishment of a system based on four evaluation criteria textile appearance defects system for the inspection of textiles scoring grading. Finally, in the above algorithms and programs based on the design developed a textile appearance defect detection devices. Experiments show that the equipment online detection rate 2 ~ 70 m / min, off-line testing speed 80 ~ 200 m / min, the highest detection accuracy of 0.5 mm, can identify defects 8 categories with high accuracy.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile