摘要
已有的库存物资缺陷智能抽检过程存在准确率低、耗时长的问题,为解决这些问题,提出一种将神经网络应用于库存物资缺陷智能抽检中的方法。在采集库存物资图像的基础上,利用小波阈值的方式对图像实施去噪处理,并通过量化处理以及小波逆变换实现图像重构;对像素进行相关性分析,提取目标对象及对应场景信息,描述库存物资缺陷图像的局部特征;通过图像增强扩充样本数据,利用以遗传算法优化后的卷积神经网络实现库存物资缺陷智能抽检。实验结果表明,神经网络可以快速准确地实现库存物资缺陷智能抽检。
The existing intelligent sampling process of the inventory material defect has the problems of low accuracy,long problems.To solve these problems,a kind of neural network is applied to inventory material defects in the intelligent sampling method is put forward.On the basis of collecting stock image,the wavelet threshold is used to de-noise the image,and the image reconstruction is realized by quantization and inverse wavelet transform.The correlation analysis of pixels is carried out to extract the target object and the corresponding scene information and describe the local features of the defect images of inventory materials,The sample data is expanded by image enhancement,and the intelligent sampling of inventory defects is realized by using the convolutional neural network optimized by genetic algorithm.The experiment results show that the neural network can quickly and accurately realize the automatic sampling of inventory defects.
作者
宁克
刘志伟
侯滨
张晓惠
NING Ke;LIU Zhi-wei;HOU Bin;ZHANG Xiao-hui(State Grid Shanxi Materials Company,Taiyuan 030006,China)
出处
《信息技术》
2024年第4期137-142,148,共7页
Information Technology
基金
国家自然科学基金项目(21373158)。
关键词
神经网络
库存物资缺陷
智能抽检
小波去噪
特征提取
the neural network
inventory material defects
intelligent sampling inspection
wavelet denoising
feature extraction