基于深度学习的瓷砖色差分类方法研究

Research on Classification Method of Tile Chromatic Aberration Based on Deep Learning

瓷砖是人类建筑不可或缺的一种材料,而瓷砖品质最重要的指标之一就是色差,一批瓷砖中色差越小,品级越高。目前企业主要是由熟练工人在特定的光照及距离条件下进行筛查,劳动强度大并且效率低,没有固定的判断标准。因此提出利用基于深度学习算法的视觉检测系统代替人工进行瓷砖色差分类,首先,经过数字图像处理算法对采集的瓷砖图像进行预处理,将瓷砖本体从背景中分割出来;然后,利用卷积神经网络分别提取具有色差的两类瓷砖特征,通过监督学习实现瓷砖的色差分类;最后,设计了图形用户界面将上述所有图像处理算法及分类算法进行实现,开发可视化人机交互式操作和分类结果显示系统。实验结果表明:基于深度学习的视觉检测系统能够在瓷砖色差分类任务上实现准确分类,且分类效率高、分类结果可视化,具有重要的应用价值。

Tile is an indispensable material for human construction, and one of the most important indicators of tile quality is chromatic aberration. The smaller the chromatic aberration in a batch of tiles, the higher the grade. At present, tiles are mainly screened by skilled workers under specific lighting and distance conditions in enterprises, which is labor-intensive and inefficient without a fixed judging standard. Therefore, this paper proposes using the visual detection system based on the deep learning algorithm to replace the manual classification of tile chromatic aberration. First, the collected tile images are preprocessed by digital image processing algorithms, and the tile body is segmented from the background;then, the convolutional neural network is used. The network extracts two types of tile features with chromatic aberration respectively, and realizes the chromatic aberration classification of tiles through supervised learning. Finally, a graphical user interface is designed to implement all the above image processing algorithms and classification algorithms,and visual human-machine interactive operation and classification result displaying system are developed.The experimental results show that the visual detection system based on deep learning can achieve accurate classification in the task of tile chromatic aberration classification, and has high classification efficiency and visualization of classification results, which has important application values.