基于SURF算法的桥梁裂纹图像拼接技术

A Robust Approach for Bridge Crack Image Mosaic Based on SURF Algorithm

摘要: 针对现阶段桥梁健康检测中存在着多样化的裂纹问题，摄像机在确保清晰度的情况下很难获取一张高分辨率高精度的裂纹图像。本文在此问题上提出了基于SURF特征点的图像拼接技术研究在桥梁裂纹上的应用，对采集到的裂纹图像进行预处理，其中主要用了灰度处理突出特征点以及滤波方法去噪，利用SURF算法对裂纹图像进行特征点提取，通过欧氏距离的相似量方法进行特征点匹配，再采用RANSAC (Random Sample Consensus随机抽样一致性)算法剔除错误的匹配点对获得精确匹配，最后使用加权平均法融合图像实现图像拼接。本实验在光照、尺度变换、模糊等不同条件下，对SURF算法、ORB (Oriented Fast and Rotated BRIEF)算法和SIFT (Scale-invariant Feature Transform尺度不变特征变换)算法进行比较，研究结果表明，基于SURF算法裂纹图像拼接在不同环境下匹配精度更高，实用性和鲁棒性更强。因此，该算法在桥梁裂纹图像拼接具有较强的应用价值。

Abstract: Diversified crack difficulties occurred to the detection about bridge health at the present stage. Meanwhile, it is hard to obtain the high resolution and precision crack images accompanied with the insurance of the camera sharpness. Therefore, the image mosaic technology was applied to analysis of the bridge cracks based on the SURF (Speeded up Robust Features) algorithm. First of all, the captured crack images are mainly pretreated through gray processing to highlight the feature points and denoised by the way of filtering. Secondly, the feature points of the crack images are extracted by the SURF and matched through the similar quantity method of Euclidean distance. Thirdly, the RANSAC (Random Sample Consensus) algorithm is employed to eliminate the wrong match points and get an exact match. Finally, the weighted average method is used to fuse for the purpose of image mosaic. The SURF, ORB (Oriented Fast and Rotated BRIEF) and SIFT (Scale-invariant Feature Transform scale invariant feature transform) were compared by changing the environment factors like lighting, scale conversion, blurring in the experiment. The research results show that the image mosaic based on the SURF performs higher matching precision, more real-time and robustness in different environments. Therefore, it performs a strong application value in bridge crack image mosaic.