基于改进的Hough变换及小波分解的废旧机织物密度检测

Density detection of scrap woven fabric based on improved Hough transform and wavelet decomposition

为实现机织物密度的高精度检测,方便后续废旧织物纱线拆解与回收工作,首先将采集的原始图像进行裁剪与滤波处理,并针对织物纱线间隙阴影范围过大容易造成直线误检的问题,对传统Hough直线检测算法进行改进,限定角度检索范围并增加距离检测函数bwdist,通过检测出的阴影连通区域中心线实现图像直线检测,进行图像偏角矫正。随后通过能量曲线的方法确定小波变换的最优分解级数,在小波变换中增加Graythred函数,确定二值化最佳阈值,提高经小波变换处理后图像的可识别精度,并设定0和1像素点比例关系确定最佳阈值,实现图像平滑处理。试验验证表明:与人工计数结果对比,算法改进前后经密的识别平均相对误差由5.14%降低至2.86%,纬密的识别平均相对误差由7.87%降低至4.61%。认为:通过对Hough直线检测和小波变换算法的改进,可以提高机织物密度的检测精度。

To achieve high-precision detection of woven fabric density,facilitate the subsequent disassembly and recycling of yarn from waste fabric,the captured original images were first cropped and filtered.In response to the problem of excessive shadow range in the fabric yarn gap,which can easily lead to false detection of straight lines,the traditional Hough line detection algorithm was improved by limiting the angle search range and adding the distance detection function bwdist.Image line detection was performed by detecting the centerline of the connected shadow area,and image skew correction was carried out.Subsequently,the optimal decomposition level of wavelet transform was determined through the method of energy curve.The Graythred function was added in the wavelet transform to determine the optimal threshold for binarization,which could improve the recognition accuracy of the image after wavelet transform processing.The optimal threshold was determined by setting the proportional relationship between 0 and 1 pixel points to achieve image smoothing.Experimental verification showed that compared with manual counting results,the average relative error of warp density recognition was reduced from 5.14%to 2.86%,the average relative error of weft density recognition was reduced from 7.87%to 4.61%before and after the algorithm improvement.It is considered that the detection accuracy of woven fabric density could be improved by the improvement of Hough line detection and wavelet transform algorithms.