针对随机Hough变换(RHT)及其改进算法在同一圆边缘点随机抽样时命中率低且投票处理过程复杂, 提出了基于有序搜索的广义Hough变换快速圆检测算法(SQHT)。该算法利用圆的几何性质和梯度方向信息来有序搜索圆边缘点集, 将Hough变换的基于投票判定基元参数转变为有效确定基元三点的讨论。执行算法时, 首先顺序搜寻第一边缘点并根据邻域边缘点集或图像灰度值计算其梯度值; 接着在第一点所在行匹配与其梯度信息相符的第二点, 根据前两点信息在第一点所在列匹配第三点; 最后按照RHT算法流程确定有效基元参数, 从而避免了随机抽样带来的时间不确定性。该算法具有计算速度快, 检测时间可控, 适用范围广泛, 抗干扰性强等特点。相比于RHT算法, 提出的算法在单目标圆检测情况下检测效率提高了2倍, 在多目标圆(5个及以上)且有非目标边缘点情况下检测效率提高了5倍, 在多圆检测方面能有效弥补了Hough变换算法的不足。

To solve the problem of low hit ratio of random sampling in the same circle and complexity of voting process when Randomized Hough Transform (RHT) or its improved algorithm were used to detect circles, an algorithm for fast circle detection using Sequenced Hough Transform (SQHT) was proposed. It used the geometric features and gradient direction information of a circle to sequentially search the circle edge points sets and turned the acquisition of circle parameters based on vote of RHT to a discussion of locating three points of circle effectively. By using the method, the first edge point was searched sequentially and its gradient was calculated based on the adjacent edge point set or its gray value. The second point was mapped according to gradient of the first point in the row. The third point was mapped according to the information of two points above. The true circle parameters were finally calculated by RHT method,which avoids timing uncertainty caused by randomness of sampling. This method has features of high speed, controllable detection time, wide range of application and strong anti-interference performance. Comparing with the RHTs, experimental results indicate that the proposed algorithm can improve 2 times or more in the image of single circle, and 5 times or more in the image of multi-circles (5 or more). It can efficiently remedy the shortage of RHT in multi-circle detection.