桥梁病害检测与识别-武汉工程大学学报.DOCVIP

PAGE PAGE 7 收稿日期：2014-2-24 收稿日期：2014-2-24 基金项目：国家自然科学基金面上项目;湖北省自然科学基金创新群体项目( 2012FFA046 ) 作者简介：洪汉玉(1964-)，男，湖北阳新人，博士研究生导师，博士后。研究方向：图像识别与智能控制。 洪汉玉1,章秀华1,叶佳伦1,荆根强2 (1.武汉工程大学 图像处理与智能控制研究所，湖北 武汉 430205； 2.交通运输部公路科学研究所, 北京 100088） 摘要：桥梁裂痕图像检测过程中，采集的桥梁裂痕图像容易出现模糊，且桥梁裂痕本身具有裂痕特征不明显、杂质干扰大等特点，为了达到对桥梁裂痕准确、快速检测的目的，本文提出了一种桥梁混凝土结构裂痕病害的自动检测识别方法。首先对采集到的桥梁裂痕图像进行去模糊处理，在此基础上利用非负特征提取桥梁裂痕目标信息，然后利用方差特征去除特征结果图中的伪特征，并使用特征图像中目标特征象素的圆投影特征来增强目标裂痕信息，同时进一步去除虚假特征。分别对不同的桥梁裂痕图像进行了多种不同类型的处理实验，包括裂痕图像去模糊前后的目标裂缝检测结果对比实验，特征图像的方差特征去噪实验，以及圆投影进行特征目标特征增强同时进一步去噪的实验。大量实验结果表明，本文提出的算法对桥梁裂痕的提取与检测有效，有一定的实际意义。 关键词:桥梁病害; 模糊裂痕; 图像特征;自动识别 中图分类号： TP391 文献标志码： A Detection and recognition method of bridge fissure Hong Han-yu1,Zhang Xiu-hua1,Ye Jia-lun1,Jing Gen-qiang2 （1.Laboratory for Image Processing and Intelligent Control , Wuhan Institute of Technology, Wuhan,Hubei,430205, China； 2. Research Institute of Highway ,Ministry of Transport ,Beijing ,100088,China) Abstract: During the detection of bridge fissure, the acquired image of bridge fissure is easy to be blurred, and the bridge fissure has the characters of non-remarkable fissure feature and excessive noise. To detect the bridge fissure quickly and exactly, an automatic detection and recognition method for beton concrete girder bridge is proposed. First, the blur of the acquired bridge fissure image was removed, and the bridge fissure object information was extracted from the deblurred image by using the non-negative feature. Then false features were eliminated from the feature intensity image with the variance feature method. The circle projection method was used to enhance the object fissure information and further eliminate the false feature. Different type of experiments are implemented on bridge fissure images, which include the fissure detection result comparison experiment before and after deblurring, the feature image denoising experiment using variance feature, the object information enhancement experiment using circle p