一种改进的随机圆检测算法

Improved randomized algorithm for circle detection

提出了一种用于圆检测的改进随机算法。首先利用4-邻接优先的八方向跟踪方法对边缘图像中的边缘像素进行跟踪形成一系列边缘链,然后从每段边缘中随机选取3个边缘像素并通过定义一个距离准则来判定这三个点能否确定一个候选圆。确定候选圆后,利用证据积累的方法进一步判断该候选圆是否为真正的圆。在证据积累过程中,无需对候选圆外接正方形之外的像素进行平方运算,只需进行简单的比较运算。通过对真实图像进行了圆检测实验,结果表明,该算法具有准确率高、速度快的优点。

An improved randomized algorithm is presented for circle detection. Edge pixels are tracked by a 4-neighbor-priority eight-direction tracking method and series of edge chains are formed. Then three edge pixels are randomly selected from every edge chain and a distance criterion is defined to determine whether there is a possible circle, An evidence-collectlng process is applied to further determine if the possible circle is a true one when a possible circle is found. During the course, simple comparison operators, not complex square operators are used when edge pixels outside of the circum-rectangle of the possible circle are dealt with. Results obtained from experiments on real images demonstrate that the proposed algorithm performs well and fast.