提出了一种新的参量式变形模板方法,用于检测带有缺陷和噪声的圆。根据圆的几何形状特性和梯度信息定义了模板的能量函数。然后通过动态调整模板的参量使能量函数和图像进行交互以把能量函数最大化。根据能量函数的定义,在其取得最大值时的参量即为所检测圆的位置参量。为了提高匹配运算速度,先求出被检测圆的圆心和半径的估计值,然后给出缩小了的搜索区域。在此搜索区域内采用贪婪优化的算法得到最终检测结果。实验结果表明算法可以在圆有较大缺陷和噪声的情况下,定位精度在1个像素以内,速度在0.5 s左右;有很好的抗干扰和抗噪声能力,并有较快的运算速度。

A new method for detecting circle with defects and noise using parametric deformable template is proposed. An energy function of the deformable template is defined according to geometric characteristics of the circle and gradient information of the image. The template then interacts dynamically with the image and energy function by altering its parameter values to maximize the energy function. The final parameter values (maximum) can give the location of the circle according to definition of the energy function. In order to raise the matching speed, a decreased search region is given by estimating values of the radius and the center of the circle. Then the energy function is optimized using greedy method in the search region. Experimental results show that the algorithm can localize the circle within a pixel and about 0.5 s; even the circle has defects and noise. This method has good resistence to interference and noise, and fast operation.