灰度投影积分极值法的矩形检测  

Rectangle Detection of Gray Projection Integral Extreme Value Method

作　　者：任玲辉[1] 刘凯[1] 张海燕[2] 杨慧云[2] 

机构地区：[1]西安理工大学机械与精密仪器工程学院 [2]西安理工大学印刷包装工程学院,西安710048

出　　处：《计算机工程》2012年第8期159-160,163,共3页Computer Engineering

基　　金：陕西省教育厅科研计划基金资助项目(11JK0866);陕西省13115科技创新工程重大科技专项基金资助项目(2008ZDKG-28)

摘　　要：现有的矩形识别方法存在检测速度慢、识别率低和精度不足等问题，不能满足现代工业在线检测发展的需求。为此，提出一种灰度投影积分极值法的矩形检测方法，把待检测图像沿0-π方向进行投影得到灰度投影积分（GPI）矩阵，求得GPI矩阵中极值点的位置，将其保存为新的矩阵GPIEV，并在GPIEV矩阵中搜索平行线重构矩形。实验结果表明，该方法具有较好的抗干扰能力，且运算速度快，能够满足实时处理的要求。Most of the rectangle recognition methods due to the low speed of detection, low recognition rate and lack of precision, they can not meet the needs of modem industrial development. To solve the above problems, a gray projecting integral extreme value method to detect rectangular is presented. It gets the image gray value projected in all directions to form a Gray Projection Integral（GPI） matrix, obtains the GPI matrix extreme points saved as a new matrix GPIEV, and searches for parallel lines in the GPIEV matrix to reconstruction rectangle. Experimental results show that the method has better anti-jamming ability, quick operation speed, and it can meet the requirement of real-time processing.

关 键 词：灰度 极值 矩形检测 图像处理 HOUGH变换 在线检测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]