桥梁裂缝的分步式检测算法.docVIP

桥梁裂缝的分步式检测算法 摘 要 桥梁的生命周期是衡量整个交通工程状况的关键，裂缝又是桥梁常见的病害形式。本文以桥梁裂缝图像为研究对象，通过深入分析桥梁裂缝图像中目标、噪声和背景等特有的性质，提出一种新的桥梁裂缝的分步式裂缝检测模型。 整个分步式裂缝检测模型主要包括四步：首先，为减少图像中的冗余信息，提高检测效率，进行图像灰度化；其次，由于目标和噪声在裂缝图像中分布与形态的多样性，本文采用自适应均值-中值滤波的阈值分割算法对裂缝目标轮廓进行提取；但是，由于部分噪声和背景成块存在于裂缝图像中，本文采用扩大邻域窗口算法去除块噪声；最后，根据裂缝目标和伪裂缝噪声的形态弯曲程度相异的特性，采用最小二乘方差分布算法对伪裂缝加以去除。同时，对光照不均时细小裂缝的提取，提出了自适应阈值选择算法。 大量实验表明，分步式裂缝检测模型在提取轮廓的同时，有效去除了块噪声和伪裂缝噪声的影响，解决了桥梁裂缝图像中多种噪声交互的复杂情形，并且算法在检测精度和速度上有了明显的提高。 关键词：桥梁裂缝，分步式裂缝检测，自适应阈值，块噪声，光照不均 MULTI-STEP DETECTION ALGORITHM OF BRIDGE CRACK ABSTRACT The bridge life cycle is the key to measure the traffic engineering conditions, and cracks are the common deterioration problems of bridge form. The paper is base on bridge crack image as the research object. Through in-depth analyze the special properties of the target, noise and background of bridge crack image, this paper propose a new multi-step cracks detection model about bridge cracks. Multiple step format crack detection model mainly includes four steps. First of all, In order to reduce the redundant information in the image and improve the efficiency of detection, the image should be converted to gray image. Secondly, it is diversity of distribution and morphology of the target and noise in the cracks images. The target contour is extracted by adaptive mean - median filtering threshold segmentation in This paper. However, Because of the noise and background into pieces in the crack images, this paper uses the expanded neighborhood window algorithm to filter the block noise. The last, Because of the target and pseudo crack noise with different features of shape bending degree, using the least squares variance distribution algorithm for removing pseudo crack in this paper. Meanwhile, to extract tiny cracks in the uneven illumination, adaptive threshold selected model was presented. A large number of experiments show that the multiple step format crack detection model not only extract the crack shape, but also is effective to get rid of the effects