混凝土桥梁裂缝图像自动匹配、拼接与融合方法  

Automatic matching,stitching and fusion method of concrete bridge crack images

作　　者：于利存 贺拴海[1] 蒋树庆[3] 项水英 YU Li-cun;HE Shuan-hai;JIANG Shu-qing;XIANG Shui-ying(School of Highway,Chang’an University,Xi’an 710064,Shaanxi,China;CCCC First Highway Consultants Co.,Ltd,Xi’an 710075,Shaanxi,China;State Key Laboratory of Integrated Service Networks,Xidian University,Xi’an 710071,Shaanxi,China)

机构地区：[1]长安大学公路学院,陕西西安710064 [2]中交第一公路勘察设计研究院有限公司,陕西西安710075 [3]西安电子科技大学综合业务网理论及关键技术国家重点实验室,陕西西安710071

出　　处：《长安大学学报（自然科学版）》2022年第6期33-41,共9页Journal of Chang’an University(Natural Science Edition)

基　　金：国家自然科学基金项目(61974177,61674119);交通运输行业重点科技项目(2020-MS1-060)。

摘　　要：针对目前使用自动化图像采集设备采集到的混凝土桥梁裂缝图像数量多、匹配准确度低、拼接及融合效果较差的问题,提出一种混凝土桥梁裂缝从匹配、拼接、融合到合并统计的端到端的解决方法。首先,在裂缝匹配阶段利用特征点匹配结合并查集来代替常规的特征点匹配方法。然后,利用移动直接线性变换算法和奇异值分解计算局部单应性矩阵和实现映射,解决了全局单应性矩阵不足以拟合所有特征点的问题;同时基于距离变换的Alpha加权融合处理算法实现拼接线的自然融合;以实际工程采集的裂缝图像数据集为研究对象,从裂缝拼接线的对齐性、裂缝可视化结果的准确性和完整性、非裂缝区域的纹理一致性等方面,综合对比分析了采用移动直接线性变换算法与传统OpenCV方法的拼接融合效果。此外,在图像拼接阶段,通过控制源图像和目标图像中的裂缝颜色变换范围解决了在拼接图像重叠区域中裂缝重复可视化显示的问题;在图像融合阶段,通过掩膜可视化法解决了图像融合导致的裂缝颜色显示变浅甚至消失的问题。最后,依据检测框之间、检测框与裂缝顶点的相对关系,实现了拼接图像中同一裂缝合并和重连接。研究结果表明:通过改进算法在有效提升匹配精度的同时,可以获得很好的拼接和融合效果,能够实现智能检测设备采集到裂缝图像的快速无缝拼接。Aimed at the concrete bridge crack images collected by automatic image acquisition equipment faces several serious problems including huge amount,low matching accuracy,poor stitching and fusion effect,an end-to-end solution for concrete bridge crack from matching,stitching,fusion to merging statistics was proposed.First,in the stage of crack matching,the conventional method of feature point matching was replaced by the combination of feature point matching and search set.Then,the moving direct linear transformation(Moving DLT)and singular value decomposition(SVD)was employed to calculate the local homography matrix and implement the mapping,which solved the problem that the global homography matrix was not enough to fit all the feature points.In addition,the Alpha weighted fusion processing algorithm based on distance transformation was adopted to realize the natural fusion of stitching lines.Based on the crack image data set collected in real engineering,the stitching and fusion effects of the moving direct linear transformation algorithm and the traditional OpenCV method were comprehensively compared and analyzed from the aspects of the alignment of crack stitching lines,the accuracy and integrity of crack visualization results,and the texture consistency of non-crack regions.Besides,in the image stitching stage,the problem of repeated visualization of cracks in the overlapping region of stitching images was solved by controlling the color transformation range of cracks in source image and target image.In the stage of image fusion,mask visualization method was used to solve the problem that the color display of cracks caused by image fusion become lighter or even disappeared.Finally,according to the relative relationship between detection boxes and crack vertices,the same crack was merged and reconnected in the stitching image.The results show that the improved algorithm can effectively improve the matching accuracy and achieve good stitching and fusion effect,and can realize the rapid and seamless stitching of

关 键 词：桥梁工程 裂缝拼接 移动直接线性变换 混凝土裂缝 图像融合 可视化 

分 类 号：U446.3[建筑科学—桥梁与隧道工程]