摘要
基于卷积神经网络的喷漆产品自动检测相比于传统产品检测可以极大地提高产品合格率以及降低检测成本.传统的产品检测需要有相关经验的人员进行检测程序且耗时耗力,且由于参与人员主观因素的影响,不能保证产品绝对的合格率.利用卷积神经网络的特性提取喷漆产品的图片表现特点并学习,然后建立模型训练其特征.为了提升检测速度,本文提出以轻量级的卷积神经网络为核心的卷积模型.将喷漆产品完整的区域图像通过卷积模型提取出来特征而得到最终的预测检测结果.
Automatic inspection of paint products will greatly improve the product pass rate and testing costs.Traditional product testing with experienced personnel testing,time-consuming and labor-intensive.And due to the subjective factors of personnel,can not guarantee the absolute pass rate of the product.In view of the picture performance characteristics of paint products,the image features and learning of products are extracted by CNN.And build a model to train its characteristics.In order to improve the detection speed,a convolution model with lightweight CNN as its core is proposed.The full area image of the paint product is loaded into the support vector machine to finally predict the results of the results of the region extracted by the convolution model.
作者
何仁杰
郭秀娟
HE Ren-jie;GUO Xiu-juan(School of electrical and computer science,Jilin Jianzhu university,Changchun 130118,China)
出处
《吉林建筑大学学报》
2020年第6期82-86,共5页
Journal of Jilin Jianzhu University
