基于Radon变换与灰度投影积分极值方法的矩形检测

Rectangle detection based on Radon transform and gray projection integral extreme value method

针对现有的矩形检测方法存在计算量大、检测精度不足等问题,提出基于Radon变换(Radon Transform,RT)与灰度投影积分极值(Gray Projection Integral Extreme Value,GPIEV)方法相结合的矩形快速检测方法,该方法运用Radon变换在小角度搜索范围内对矩形图像进行快速准确的旋转校正,随后对校正后的图像的垂直和水平方向使用灰度投影积分极值方法确定矩形2对平行直线位置,完成矩形的角度快速校正与重构。该检测方法利用了Radon变换对直线倾斜角度检测的准确性和灰度投影积分极值方法检测直线的准确性,弥补了这两种方法各自的不足。实验结果表明,该检测方法可应用在矩形等规则平面的几何图形检测识别中,检测精度高,且检测速度快。

In view of most existing detection methods with various disadvantages,such as large calculating quantity and low detection precision,the present paper proposed a rapid detection method which integrates Radon Transformation（RT）and Gray Projection Integral Extreme Value method（GPIEV）.Firstly,Radon transformation is used to quickly and accurately correct a rectangular image by rotating it within a small angle range.Secondly,the positions of the parallel lines in the collected image are determined by applying Gray Projection Integral Extreme Value method along the vertical and horizontal directions of the image.In this way,the angle correction and reconstruction of the rectangular image can be rapidly achieved.This detection method makes full use of the accuracy of Radon transformation and Gray Projection Integral Extreme Value method in detecting line and its inclination angle,and makes up for their deficiencies.Experimental results show that this detection method can be used to detect and recognize rectangular and other regular plane geometric figures with high accuracy and fast speed.