基于深度学习的垃圾图片处理与识别

2021-04-25 13:06:20齐鑫宇龚劬李佳航何建龙
电脑知识与技术 2021年9期

齐鑫宇 龚劬 李佳航 何建龙

摘要:针对城市环境卫生提出的对市民生活垃圾进行分类回收的要求,考虑计算机卷积神经网络在图片分类中的强大表现,提出了基于深度学习中卷积神经网络对垃圾图片处理以及输出识别的新模型与方法。针对目前图像局部特征表达存在的复杂性,模糊性等不足,采用特征多层池化以及系统神经网络学习的方式进行优化。同时在ResNet101模型的基础上设计并构建了基于CNN(Convolutional Neural Network)算法的新模型框架,此系统模型也能实现端与端的实时识别。新模型提高了对训练样本图像信息提取的精确度以及图片识别的准确率,实验表明识别准确率平均提高了10%。为未来实现人工智能垃圾分类提供图像识别模型基础。

关键词: 卷积神经网络;多层池化;垃圾分类;图像识别;实时识别

中图分类号: TP183        文献标识码:A

文章编号:1009-3044(2021)09-0020-05

开放科学(资源服务)标识码(OSID):

Deep Learning Based Garbage Image Processing and Recognition

QI Xin-yu1, GONG Qu2, LI Jia-hang1, HE Jian-long1

(1. School of Aeronautics and Astronautics, Chongqing University, Chongqing 400000, China;2. School of Mathematics and Statistics, Chongqing University, Chongqing 400000, China)

Abstract: In response to the requirement of urban environmental health to classify and recycle citizens' household garbage, a new model and method for garbage image processing and output recognition based on convolutional neural networks in deep learning is proposed considering the powerful performance of computer convolutional neural networks in image classification. For the shortcomings of the current image local feature representation such as complexity and ambiguity, feature multilayer pooling and systematic neural network learning are used for optimization. A new model based on the CNN (Convolutional Neural Network) algorithm is also designed and built on the basis of the ResNet101 model, and this system model can also achieve end-to-end real-time recognition. The new model improves the accuracy of image information extraction from training samples and the accuracy of image recognition, and experiments show that the recognition accuracy is improved by 10% on average. It provides the basis of image recognition model for future implementation of artificial intelligence garbage classification.

Key words:convolutional neural network;multilayer pooling;garbage classification;image recognition;Real-time identification

引言

隨着时代的进步,人民生活水平的提高,生活垃圾的急剧增加成为城市卫生保护工作的一大难点,其中有效垃圾的利用以及无效有害垃圾的回收成为处理垃圾问题的一个重点问题。许多城市提出了公民自行垃圾分类的要求,这在一定程度上缓解了城市垃圾分类处理问题的恶化,但由于垃圾种类繁多,分类标准复杂且难以准确区别,导致效果并不是很好。国内主要的回收厂大部分采用人工分拣,这种方式不仅效率低下,并且对劳动人员生命健康造成伤害[1, 2]。所以要想完全解决这个问题就要从科学生产力上来解决,利用人工智能自动化分拣,实现智能分类以及回收必然会成为未来垃圾处理的模式。……

登录APP查看全文