基于机器视觉的毛杆缺陷检测方法  

Defect detection of feathers base on computer vision

作　　者：刘洪江[1] 汪仁煌[1] 

机构地区：[1]广东工业大学自动化学院,广东广州510090

出　　处：《大连工业大学学报》2012年第5期382-385,共4页Journal of Dalian Polytechnic University

基　　金：国家自然科学基金资助项目(11072063)

摘　　要：建立了一套羽毛毛杆缺陷在线检测的完备方案,该方案运用机器视觉方法对羽毛进行无损检测。其重点是把图像检测过程分为三步,首先区别羽毛和背景,其次从羽毛中分割出毛杆,最后检测毛杆上是否存在缺陷。本文对每步都提出有效算法:用阈值法快速简便地区分羽毛和背景;用气球力的主动轮廓模型来分割毛杆;用小波分解和统计分析法来判断缺陷的存在。在小波分解前,根据毛杆缺陷特征,把图像的二维信号变换到一维信号,把缺陷的检测变为信号中奇异点的检测。在分析缺陷大小和特征后提出小波基的选择方法和分解层次的限制。该方案经过试验验证行之快速有效。A set of scheme for on-line detection of feather using machine vision is developed,including distinguishing feather from background,segmenting quill from feather,defect detection on quill.Effective algorithms on each step are mentioned using threshold value to quickly and easily distinguish between feather and background,using the balloon force of active contour model to segment quill,using wavelet decomposition and statistical analysis to detect defects.The image′s two-dimensional signals are transformed into one-dimension signals according to the defect features of quill before the wavelet decomposition,and thus the defect detection is turned into signal′s singularity detection.Methods for the selection of wavelet base and determining the restrictions of decomposed layers are also raised after analyzing the size and characteristics of quill′s defects.

关 键 词：机器视觉 主动轮廓模型 小波变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP274[自动化与计算机技术—计算机科学与技术]