对光源环境鲁棒的铣削表面粗糙度分类检测

Classification Detection of Milling Surface Roughness Robust to Light Source Environment

针对当前基于神经网络的铣削表面粗糙度视觉检测方法,大多结合加工工艺参数,且要人为地设计图像评价指标,并且拍摄过程对环境光源严重依赖的问题,提出了一种基于卷积神经网络的铣削表面粗糙度分类方法。该方法采用白天和黑夜的光照环境分别拍摄实验图片,经过端到端的图像分析,利用卷积操作和综合处理得出粗糙度分类模型,可对表面粗糙度进行分类预测。试验结果表明,该方法能够实现自动特征提取,避免了特征提取,且无需结合加工工艺参数,并对拍摄的光源环境不敏感,为表面粗糙度实现在线视觉测量提供了可能。

In order to solve the problem that most of the current visual detection methods of milling surface roughness based on neural network combine the processing parameters, design the image evaluation index artificially, and the shooting process depends heavily on the environmental light source, a milling surface roughness classification method based on convolutional neural network is proposed.In order to solve these problems, this paper proposes a milling surface roughness classification method based on convolutional neural network.In this method, experimental images are taken in the light environment of day and night respectively.After end-to-end image analysis, the roughness classification model is obtained by using convolution operation and comprehensive processing, and finally the surface roughness can be classified and predicted.Experimental results show that this method can automatically extract features without combining processing parameters, which can avoid feature extraction and is insensitive to the light source environment, which provides the possibility for the realization of online visual measurement of surface roughness.