基于脉冲耦合神经网络的路面裂缝提取  

Pavement cracks extraction based on pulse coupled neural network

作　　者：宋蓓蓓[1] 韦娜[1] 

机构地区：[1]长安大学信息工程学院,陕西西安710064

出　　处：《长安大学学报（自然科学版）》2011年第5期33-37,共5页Journal of Chang’an University(Natural Science Edition)

基　　金：国家自然科学基金项目(60902075);中央高校基本科研业务费专项资金项目(CHD2009JC014;CHD2010JC056);博士后科学基金特别资助项目(201003660)

摘　　要：考虑裂缝比路面背景更暗的特点,采用结合赋时矩阵的脉冲耦合神经网络模型,实现了路面图像分割和裂缝的粗提取;利用裂缝比杂质面积大的特点,提出一种基于数字形态学的连通区域提取算法,通过计算每个区域包含的像素数量,采用阈值方法剔除杂质,实现裂缝的精提取。研究结果表明:脉冲耦合神经网络裂缝粗提取方法的平均检测率和虚检率分别为92.43%和47.67%;综合方法平均检测率和虚检率分别为91.1%和7.68%,显著提高了路面裂缝检测的准确性。Based on the characteristic that pavement crack is much darker than its background,the image segment and crack coarse extraction were realized by adopting pulse coupled neural network model combined with time matrix.Based on the fact that the area of crack is much larger than that of impurities,a new connected region extraction algorithm based on mathematical morphology was proposed.It achieved the fine crack extraction by calculating the pixels number of each region and then adopted threshold method to remove impurities.The results show that the average hit rate and false alarm ratio of the coarse crack extraction method based on pulse coupled neural network are 92.43% and 47.67% respectively,and those of the finer method are 91.1% and 7.68% individually,the accuracy of crack detection is significantly improved.2 tabs,4 figs,8 refs.

关 键 词：道路工程 公路路面裂缝 脉冲耦合神经网络 数字形态学 图像分割 

分 类 号：U416.06[交通运输工程—道路与铁道工程]