期刊文献+

基于深度学习的垃圾分类检测方法 被引量:10

Garbage classification and detection method based on deep learning
下载PDF
导出
摘要 针对现有垃圾分类不清、人工检测难度大、环境差、容易出错等情况,研究了基于深度学习的垃圾分类检测方法。分别提出了基于YOLOv3、RetinaNet和Faster RCNN的垃圾分类识别方法,制作了所用的数据集(训练集和测试集)。搭建了基于三种方法的垃圾分类识别的实验平台,并设计了实验,使用制作好的垃圾训练集进行多次不同参数下的训练,在不同分类的垃圾测试集上进行多次测试。对训练过程和测试结果进行综合分析和比较,得到Faster RCNN算法有更高的可靠性和准确性,检测速度满足系统要求,RetinaNet算法效果较好,YOLOv3算法效果最弱。因此采用基于Faster RCNN的垃圾分类识别算法,很好地满足了垃圾异物识别模型的开发,实现了垃圾分类检测,且有效降低了人工成本,提高了干湿垃圾分类检测效率,从而降低了垃圾对环境的污染。 In view of the unclear garbage classification,difficult manual detection of garbage,harsh environment and easy to make mistakes in the garbage classification,a garbage classification and detection methods based on deep learning is studied.The garbage classification and identification methods based on YOLOv3,RetinaNet and Faster RCNN are proposed respectively,and the required data sets of training set and test set are made.The garbage classification and identification experimental platform based on the three methods is built.The experiment was designed.The prepared garbage training set was used for multiple trainings under different parameters and for multiple tests on the garbage test sets of different classifications.After comprehensive analysis and comparison of the training process and test results,it is found that the Faster RCNN algorithm has higher reliability and accuracy,and its detection speed can meet the requirements of the system.RetinaNet algorithm has a moderate effect,while YOLOv3 algorithm has the weakest effect.Therefore,the garbage classification and recognition algorithm based on the Faster RCNN can well meet the development of garbage foreign object recognition model,realize garbage classification and detection,and effectively reduce the labor cost.In addition,the efficiency of dry and wet garbage classification and detection is improved,which reduces the pollution of garbage to the environment.
作者 王小燕 谢文昊 杨艺芳 胡瑞 WANG Xiaoyan;XIE Wenhao;YANG Yifang;HU Rui(School of Science,Xi’an Shiyou University,Xi’an 710065,China;No.8 Oil Production Plant of Changqing Oilfield,Xi’an 710018,China)
出处 《现代电子技术》 2021年第21期110-113,共4页 Modern Electronics Technique
基金 陕西省科学技术厅面上项目(2020JM-543):非线性自适应图正则的子空间聚类算法研究。
关键词 深度学习 垃圾分类 目标检测 图像识别 算法分析 YOLOv3 RetinaNet Faster RCNN deep learning garbage classification object detection image recognition algorithm analysis YOLOv3 RetinaNet Faster RCNN
  • 相关文献

参考文献8

二级参考文献108

共引文献169

同被引文献98

引证文献10

二级引证文献10

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部