摘要
不同于传统的Hough变换算法,文中提出一种基于曲线弧分割的椭圆检测方法。首先将图像转化为二值图像,基于细化算法由二值图像得到细化的轮廓图。根据轮廓点间相互的连接性对细化图像中的所有轮廓像素点进行跟踪。将轮廓从交点处分割成弧段,直到所有的轮廓被跟踪完毕。根据各个弧段长度的比例,确定在每段弧上采样的次数,在每一个连续的轮廓弧段中采样,每次采样随机取5个点。将每次采样得到的5个随机点带入椭圆一般方程,分别计算出一组椭圆参数。最后基于统计的思想,记录各组参数的出现次数。找出出现次数最多的一组参数,最终得到目标椭圆的参数。该方法从每一个连续的轮廓弧段中采样,使无效随机采样的概率大大降低。实验结果表明,该算法能快速检测出图中椭圆,运行时间远小于采用随机Hough变换算法,在具有噪声、椭圆残缺的情况下仍能有较好的检测结果。
Be different from the traditional Hough transform algorithm, put forward a kind of ellipse detection method based on curve arc segment in this paper. First,the original image is transformed to binary images. Then,thinning algorithm is used for converting the binary image to the thinning outline image. Based on the connectivity of outline pixels, all the pixels in the thinning outline image are tracked. The outline is divided into arc segments. The proportion of each segment length determines the number of samples in each section of the arc. Each sample randomly selects five points. Five random sampling points are pluged into the elliptic general equation. And a group of ellipse parameters are calculated respectively. Finally, based on the statistical ideas, find out a group of parameters that appear most, and the target ellipse parameters are finally get. In this method, pixels are sampled from every continuous outline arcs. The probability of inva- lid random sampling is reduced greatly. Experimental results show that the algorithm can quickly detect the ellipse. Running time is far less than the randomized Hough transform algorithm. In cases of noise and elliptical deficiency, this algorithm can still presents good re- sults.
出处
《计算机技术与发展》
2015年第10期19-23,28,共6页
Computer Technology and Development
基金
山东省科技发展计划(2014GSF118152
2012GSF12105)
山东省优秀中青年专家奖励基金(BS2011DX002)
关键词
椭圆检测
随机采样
无效采样
轮廓分割
边界跟踪
耗费时间
ellipse detection
random sampling
invalid sampling
outline segment
boundary tracing
consumption time
