摘要
可疑目标区域的确定是大背景中微小目标的机器视觉检测的一个关键问题。以棉花中污染物的检测为背景,根据人类的视觉注意机制,提出了一种可疑目标区域的机器视觉检测算法:首先采用主成分分析法(PCA)和离散余弦变换(DCT)算法对分块图像进行数据压缩,然后采用BP神经网络、支持向量积(SVM)模拟人脑对检测目标区域的识别。结果表明:分块图像大小、数据压缩算法和模式识别方法对识别能力有较大的影响;本文提出的检测算法可以确定可疑目标区域。在实验分析的基础上,提出了提高精度和速度需要进一步解决的几个问题。
An algorithm for identification of small target suspicious region in a large background in machine vision inspections was developed, which was based on the simulation of human visual attention mechanism for detecting contaminants in cotton. Firstly, Principal Component Analysis （PCA） and discrete cosine transform （DCT） were adopted to compress the data in the divided image blocks and extract their main features for pattern recognition. Then BP neural networks and Support Vector Machine （SVM） were used to identify suspicous target area respectively. The results indicated that the algorithm can be used for identification of suspicious target region in machine vision inspection, the accuracy of which was mainly determined by the size of the divided image blocks, the algo- rithms for data compressing and methods for pattern recognition. Finally, based on analysis of the results, further research for improving the algorithm was suggested.
出处
《四川大学学报（工程科学版）》
EI
CAS
CSCD
北大核心
2010年第1期233-237,共5页
Journal of Sichuan University (Engineering Science Edition)
基金
国家自然科学基金资助项目(60875022/F030410)
关键词
注意机制
机器视觉系统
小目标检测
算法
visual attention
machine vision system
small target detection
algorithm
