不规则类圆形团块目标图像识别的新方法

来源期刊：中南大学学报(自然科学版)2004年第4期

论文作者：罗三定 肖飞

文章页码：632 - 637

关键词：图像处理;边缘检测;聚类;类圆形;模式识别

Key words：image process; edge detection; clustering; quasi-circular object;pattern recognition

摘    要：针对类圆形目标图像识别方法的不足,以轧钢厂成捆棒材图像为实例,提出了不规则类圆形团块目标模式识别新方法,设计了边缘检测、中心增强和重心聚合等一系列算法,以实现计算机对类圆形目标图像的自动识别与检测。研究结果表明:以新算法为核心技术的软件能够对直径为12~40 cm的棒材进行计数,点支准确的捆精度达到99%,且每捆计数时间少于3 s。

Abstract: Aimed at the disadvantages of the existing methods of identifying the round type images, a new method of identifying the round type particle target was put forward by an example of stick bundle images in factory. A series of algorithms, such as the border detection, the center enhancement and the convergence of center of gravity were designed to realize the automatic identification and detection of the images by computer. The software designed based on the above alorithms can count the sticks whose diameter is from 12 cm to 40 cm. The precision is 99% and the counting time for each bundle is less than 3 s.