基于改进GAC模型的钢丝绳图像检测方法  

Elevator Wire Rope Image Detection Method Based on Improved GAC Model

作　　者：刘剑[1] 赵悦[1] 韦松伯 

机构地区：[1]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168

出　　处：《沈阳建筑大学学报（自然科学版）》2017年第1期184-192,共9页Journal of Shenyang Jianzhu University：Natural Science

基　　金：国家自然科学基金项目(610272253);辽宁省自然科学基金项目(201602616);辽宁省教育厅科学研究项目(L2015443);住房和城乡建设部科技攻关项目(2015-K2-015)

摘　　要：目的针对测地线活动轮廓模型易受初始位置选择限制,处理弱边缘效果较差的问题,提出一种改进测地线活动轮廓模型.方法首先,将局部二元拟合的能量函数中引入图像的局部和全局方差信息,避免其在处理复杂图像时易陷入局部最小值;其次对该能量函数进行归一化处理,代替原来测地线活动轮廓模型的边缘停止函数;最后,驱使轮廓曲线运动到钢丝绳图像中的损坏区域的边缘处.结果改进后的模型能够有效地分割出钢丝绳图像中的轮廓和损坏边缘,以人机交互的形式提取轮廓部分和损坏部分,计算相应的面积,获得缺损度,实现钢丝绳的故障检测.结论笔者所提出的改进GAC模型可以简单、便捷的检测出钢丝绳故障缺损度,同时也证明了该模型在解决钢丝绳图像检测问题时具有较高的精度.The Geodesic Active Contour Model is easy to be restricted by the initial curve position and poor treatment of the weak edge effect. In this paper, we propose an improved Geodesic Active Contour Model. Firstly, the local information and the global variance information of the image is introduced into the energy functional of the Local Binary Fitting Model to avoid the local minimum value when dealing with the complex images. Secondly, the energy function is normalized, and replaces the edge stopping function of the Geodesic Active Contour Model. Lastly, it drives the contour curve moving to separate the edge of the wire rope in the image. The experiment results show that the improved model can effectively separate the edge of the contour of the wire rope and the damaged area. In the form of man-machine interaction to extract the contour of the wire rope and the damaged part, calculate the corresponding area, to obtain the defect degree, to achieve the fault detection of wire rope. The improved GAC model can simple and convenient detect wire rope fault, and it is proved that the model has better segmentation accuracy in solving wire rope image detection problems.

关 键 词：钢丝绳 图像分割 测地线活动轮廓模型 局部二元拟合 

分 类 号：TP277.3[自动化与计算机技术—检测技术与自动化装置]