基于图像LBP特征与Adaboost分类器的垃圾分拣识别方法  

Waste Sorting Recognition Method Based on Image LBP Features and Adaboost Classifier

作　　者：陈昱辰 曾令超 张秀妹 钟广泽 Chen Yuchen;Zeng Lingchao;Zhang Xiumei;Zhong Guangze(Guangdong Baiyun University Big Data and Computer Science,Guangdong Guangzhou 510450)

机构地区：[1]广东白云学院,广东广州510450

出　　处：《南方农机》2021年第21期136-138,144,共4页

基　　金：广东大学生科技创新培育专项资金(pdjh2020b0668)。

摘　　要：生活垃圾的有效处理和相关高价值废品的回收再利用,已日益成为社会关注的焦点。目前垃圾分拣主要采用人工分拣,工作强度大、效率低,基于此,课题组提出了一种能自动识别垃圾并进行分拣的算法,以提取图像的LBP特征编码为基础,设计了基于LBP特征的级联分类器。仿真结果表明,通过对果皮和有害垃圾的123张样本图片进行LBP特征提取,训练出的具有15级弱分类器的Adaboost模型,能对测试集314张垃圾图片进行识别,识别率达到90.4%。The effective treatment of household garbage and the recycling of related high-value wastes have become the focus of increasing social development.In view of the current garbage sorting mainly adopts manual sorting,which has high work intensity and low efficiency,the research group proposed an algorithm that can automatically identify and sort garbage.Based on the LBP feature coding of extracted images,a cascade classifier based on LBP feature was designed.The simulation results show that through LBP feature extraction of 123 sample images of fruit peel and harmful garbage,the trained Adaboost model with 15-level weak classifier can recognize 314 garbage images in the test set,and the recognition rate reaches 90.4%.

关 键 词：LBP编码 级联分类器 ADABOOST算法 果皮识别 垃圾分类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]