基于同态滤波和深度卷积模型的低照度工件表面粗糙度等级识别  

Surface roughness level recognition of low-illumination workpieces based on homomorphic filtering and deep convolution model

作　　者：洪琛 朱坚民[1] 黄之文[1] HONG Chen;ZHU Jianmin;HUANG Zhiwen(School of Mechanical Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学机械工程学院,上海200093

出　　处：《光学仪器》2021年第2期16-23,共8页Optical Instruments

基　　金：上海市科委科研计划项目(17060502600)。

摘　　要：为解决低照度下工件表面粗糙度等级识别正确率低的问题,提出一种基于同态滤波和深度卷积模型的低照度工件表面粗糙度等级识别的方法。该方法通过对不同照度下工件表面粗糙度图像进行等级识别,确定了同态滤波器的最优参数值,再将图像从RGB空间转换到HSV空间,在对V(亮度)分量进行同态滤波处理后,再将图像转回RGB空间并通过设计好的深度卷积模型对图像进行识别。实验结果表明:图像的亮度对比度得到了改善,图像的纹理细节更加显著;该方法简单、有效,对低照度工件表面粗糙度等级识别有很好的效果,识别正确率达到95%以上。In order to solve the problem of low accuracy in identifying the surface roughness of workpieces under low illumination,a method based on homomorphic filtering and deep convolution model is proposed.This method optimizes and determines the optimal parameter value of the homomorphic filter by identifying the level of the workpiece surface roughness image under different illuminances.The brightness and contrast of the image are improved.The texture details of the image are more prominent.The conversion method converts the image from RGB space to HSV space,performs homomorphic filtering on the V(brightness)component,returns to RGB space, and recognizes the image through the designed deep convolution model. The experimentalresults show that the brightness and contrast of the image are improved, and the texture details ofthe image are more prominent. The method is simple and effective. It has a good effect on thesurface roughness level recognition of low-illumination workpieces, and the recognition accuracyrate reaches more than 95%.

关 键 词：低照度图像 同态滤波 粗糙度等级 卷积神经网络 图像识别 

分 类 号：TH751[机械工程—仪器科学与技术]