基于改进自适应加权均值滤波的桥梁裂缝测量

Bridge Crack Measurement Based on Improved Adaptive Weighted Mean Filtering

桥梁表面光照不均匀、对比度低、噪声干扰严重,采集到的裂缝图片较灰暗且含有大量混合噪声,因而单一的滤波方法不能达到很好的去噪效果。针对该问题,提出一种基于自适应加权均值滤波算法来实现桥梁裂缝图像滤波。提出了基于灰度拉伸和自适应加权均值滤波相结合的桥梁裂缝识别方法。利用灰度拉伸均衡图像的阴影和光照,改进的均值滤波来滤除复杂噪声,采用Otsu法和基于连通域滤波进行裂缝图像的二值化并去除孤立噪声点,结合裂缝骨架图,定量分析裂缝面积,长度和宽度。测试结果表明,通过与传统均值滤波、中值滤波比较,这种组合了两种特征优点的方法将图像去噪质量平均提高了1.87 dB,图像均方误差平均降低了5.173%,识别效果优于两种滤波算法独立使用的情况。

The bridge surface is not uniformly illuminated,the contrast is low,and the noise interference is serious.The collected crack images are gray and contain a lot of mixed noise,so a single filtering method cannot achieve good denoising effect.To address this problem,an adaptive weighted mean filtering algorithm is proposed to realize bridge crack image filtering.A bridge crack identification method based on a combination of gray-scale stretching and adaptive weighted median filtering is proposed.The grayscale stretching is used to equalize the shadows and illumination of the image,the improved mean filtering is used to filter out the complex noise,the Otsu method and the connected domain based filtering are used to binarize the crack images and remove the isolated noise points,and the crack skeleton map is combined to quantitatively analyze the crack area,length and width.The test results show that this method,which combines the advantages of both features,improves the image denoising quality by an average of 1.87dB and reduces the image mean square error by an average of 5.173%compared with the traditional mean filter and median filter,and the recognition effect is better than the two filtering algorithms used independently.