沥青混合料CT图像边缘检测算子模板的优选

Selection on Operator Model of CT Image Edge Detection for Asphalt Mixture

在沥青混合料微观材料和结构研究中,如何对沥青胶浆和集料进行有效的边缘检测是进行图像分割和二值化的前提,论文分析了沥青混合料试件CT图像边缘检测算子模板选择的原则,可作为进一步沥青混合料微观结构研究的基础。对常用的图像边缘检测算子模板及其特点分别进行了分析,并采用Sobel算子、Prewitt算子、LOG算子和Canny算子对沥青混合料马歇尔试件初始CT扫描图像和经受冻融劈裂后的同一扫描截面CT扫描图像实现边缘检测以进行算子模板的比选。多个不同算子模板的图像边缘检测结果表明,Canny算子运用高斯函数的一阶微分计算梯度,其算法不易受噪声干扰,能检测出真正的弱边缘,因此具有较好的边缘检测性能。Canny算子依据其"非极大值抑制"和形态学连接操作的特征,使提取的边缘十分完整而优于其它算子。

In the micro study of asphalt mixture＇ s material and structure, how to conduct a effective edge detection is a premise for the image segmentation and double threshold. The paper analyzes principles in the selection of operator models and can be a foundation of further study in asphalt mixture＇ s micro structure. The paper analyzes the characteristics of several image edge detection templates and contrasts the detect results of the original image and the＇image of the frozen-thaw specimen by using Sobel, Prewitt LOG and Canny models. The image edge detecting results of different operator templates in- dicate that Canny operator is anti-jamming, which compute grads using gauss function＇ s one order differential coefficient, and it can check out genuine weak edges. Canny operator obtains better results than other operators based on its characteristics of non-maximum restrain and its morphological connection.