基于梯度直方图的产品表面缺陷图像自适应阈值分割方法研究

为了提高产品表面缺陷检测精度并降低检测成本,文章以E型磁环表面缺陷的自动检测为例,分析了产品表面图像的特征,提出了一种利用梯度直方图信息的图像自适应阈值分割算法.通过实例,与原自适应直方图阈值分割算法和固定阈值分割算法相比较,结果表明:该算法采用可动态调整的形态学梯度算子计算梯度直方图,抗噪能力强;充分利用了图像的边缘信息和缺陷内部灰度分布不均匀的特性,适应性广;算法简单,具有实时准确分割图像的特点.

Gradient-histogram Based Adaptive Threshold Segmentation Algorithm for Defects Image of Product Surface

Taking for example auto-detection of surface defects of E-shape magnetic tach, this paper analyzes the characters of product surface images. An adaptive image segmentation algorithm based on information of gradient-histogram is presented to improve detection precision of product surface defects and reduce the cost. In the end, the algorithm is compared with auto- adaptive histogram threshold segmentation algorithm and fixed threshold segmentation algorithm. The results indicate the algorithm has strong anti-noise ability by using dynamic adjustable morphologic operator to calculate grad-histogram, has wide adaptability by making the best of the characters of edge information and uneven gray-level distribution in defects, and has the ability of segmenting real-time image exactly.