机器视觉的低照度图像缺陷识别方法

A Defect Recognition Method of Low-light Image with Machine Vision

对低照度图像实施识别的过程中,现有低照度图像受自身导图以及光线影响,含有噪声图像,导致现有方法在图像照度在15~45 lx的范围内存在缺陷区域图像对比度较低的问题,为此提出一种基于机器视觉的低照度图像缺陷识别方法。通过灰度变换算法改变低照度图像中各像素点的实际灰度值,并通过线性变换增强低照度图像对比度。根据有雾图像与低照度图像的反转图像之间的相似性对低照度图像实施降噪处理,精确估计图像缺陷区域。通过SIFT特征描述低照度图像的特征点,对低照度图像的特征进行提取。根据提取的低照度图像特征,基于机器视觉构建低照度图像缺陷识别模型,实现低照度图像缺陷识别。为了证明基于机器视觉的低照度图像缺陷识别方法实现了缺陷区域对比度的提升,将原有方法作为对比实验方法,比较该方法与原有方法的缺陷区域对比度,结果证明该方法成功实现了缺陷区域对比度的提升,更适用于图像照度在15~45 lx的范围内的低照度图像缺陷识别。

To solve the low picture contrast problem of the defective area of an image with illumination within the range of 15-45 lx,a defect recognition method of low-light image with machine vision was proposed.The actual gray value of each pixel in the low-light image was changed by gray transformation algorithm and the contrast of the image was chanced through linear transformation.According to the similarity between the foggy image and low-light image,the low-light image was denoised so as to accurately estimate the defective area.Furthermore,the feature points of low-light image were described by SIFT feature and the feature of the image was extracted.Finally,a machine vision based model for defect recognition of low-light image was built according to the extracted feature.The results showed that this method could increase the contrast of the defective area of low-light image,especially of the image with illumination within the range of 15-45 lx.