摘要
由于分拣环境中塑料瓶形状、颜色多样,人工提取的特征往往存在着主观性,可能丢失重要的信息或特征之间的复杂关系,从而导致塑料瓶分类效果差等问题,研究提出了一种基于卷积神经网络的废弃塑料瓶颜色分选系统。收集了市面上常见的塑料瓶,使用工业相机采集图像,使用labelimg进行人工标注制作数据集,通过数据增强来增扩数据集,提高训练模型的鲁棒性。利用YOLOv7卷积神经网络模型进行颜色分类的检测识别,使用喷阀筛选目标颜色的塑料瓶。实验表明,实时检测准确率高达95.7%,具有高度自动化和智能化水平,在回收行业中具有较高的应用价值。
Due to the diverse shapes and colors of plastic bottles in the sorting environment,the manually extracted features are often subjective and may lose important information or complex relationships between features,resulting in poor classification effect of plastic bottles,etc.A color sorting system for waste plastic bottles based on convolutional neural networks is proposed.Firstly,common plastic bottles in the market were collected,images were collected with industrial cameras,labelimg was used to manually annotate the data set,and data enhancement was used to expand the data set and improve the robustness of the training model.The YOLOv7 convolutional neural network model was used for detection and recognition of color classification,and the spray valve was used to screen the plastic bottles with the target color.The experiment shows that the real-time detection accuracy is as high as 95.7%,with a high level of automation and intelligence,and has high application value in the recycling industry.
作者
倪瑞涛
邹红艳
倪超
朱瑞林
NI Rui-tao;ZOU Hong-yan;NI Chao;ZHU Rui-lin(College of Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing Jiangsu 210037,China)
出处
《林业机械与木工设备》
2024年第3期43-49,共7页
Forestry Machinery & Woodworking Equipment
关键词
颜色分类
废弃塑料瓶
YOLOv7
智能分类
color classification
discarded plastic bottles
YOLOv7
intelligent classification