论文部分内容阅读
为了提高沥青路面裂缝的检测精度,针对裂缝的复杂性和多态性,提出了一种普适性较好的基于多尺度脊边缘的沥青路面裂缝检测算法。首先,对路面不同形状裂缝的多种脊边缘特征进行分析,并对利用高斯函数及其导数检测脊边缘的原理进行了推导;其次,根据脊边缘的多尺度特性,采用多个尺度的高斯滤波器对图像中的多尺度脊边缘进行检测,并对各个尺度的检测结果进行像素级融合;然后,结合裂缝的长度、宽度、方向、连续性、线性度等特征去除伪裂缝;最后,利用最小生成树算法实现了检测图像上不同位置裂缝的连接,使得检测出的裂缝更加接近实际裂缝的形态。研究结果表明:提出的算法可以对不同形状特征、尺寸和位置的裂缝目标进行有效检测,具有较高的检测精度和较好的检测效果,且该算法抗噪性能好、通用性强,是一种具有工程应用前景的高效的沥青路面裂缝检测方法。
In order to improve the detection accuracy of asphalt pavement cracks, a universal universal crack detection algorithm for asphalt pavement based on multi-scale ridge edges is proposed to deal with the complexity and polymorphism of cracks. First of all, the characteristics of various ridge edges of different shape cracks in pavement are analyzed, and the principle of detecting ridge edge by using Gaussian function and its derivative is deduced. Secondly, based on the multiscale features of ridge edge, Gaussian filter The edge of multi-scale ridge in the image is detected, and the pixel-level fusion of the detection results of each scale is carried out. Then, the false cracks are removed according to the characteristics of the length, width, direction, continuity and linearity of the crack. Finally, The spanning tree algorithm realizes the connection of the fractures at different positions on the detection image, so that the detected fractures are closer to the actual fractures. The results show that the proposed algorithm can effectively detect the fracture targets with different shape, size and position, has higher detection accuracy and better detection results. The proposed algorithm has good anti-noise performance and high universality Highly effective asphalt pavement crack detection method with engineering application prospect.