基于奇异熵的钢球表面缺陷特征提取研究

Study on Feature Extraction for Steel Ball's Surface Flaws Based on Singular Entropy

为评价钢球表面缺陷,应用小波分析、奇异值分解和信息熵相结合的方法提取钢球表面特征,基于奇异熵定义钢球表面图像特征参数,计算4种典型钢球表面缺陷图像的特征参数.结果表明:特征参数能反映图像的纹理特征,钢球表面损伤愈大,纹理愈复杂,其特征参数也愈大,特征参数可用于钢球表面缺陷的识别. Abstract： The character of steel ball surface is extracted to estimate its surface flaws by means of wavelet analysis,singular value decomposition and information entropy. The character parameter of steel ball surface image is defined based on the singular entropy. The character parameters of four typical steel ball surface flaw images are ealculated. It is shown that the character parameter describe the texture character of image, larger surface flaw, more complex surface texture and bigger character parameters. Therefore the character parameters can be used to identify steel ball surface flaws.