基于改进SSD的食物浪费行为识别方法  

Food waste behavior recognition method based on improved SSD

作　　者：杨永闯 王昊 王新良 YANG Yong-chuang;WANG Hao;WANG Xin-liang(School of Physics and Electronic Information Engineering,Henan Polytechnic University,Jiaozuo 454003,China)

机构地区：[1]河南理工大学物理与电子信息学院,河南焦作454003

出　　处：《计算机工程与设计》2023年第8期2523-2530,共8页Computer Engineering and Design

基　　金：河南理工大学博士基金项目(B2017-57);河南省高校基本科研业务费专项资金基金项目(NSFRF200335);2021年度河南省高等学校重点科研基金项目(21B413005);2019年度河南省高等学校青年骨干教师培养计划基金项目(2019GGJS060);河南理工大学基本科研业务费专项基金项目(NSFRF210460)。

摘　　要：为更准确地识别现实生活中复杂环境下的食品浪费行为,提出一种改进单激发多盒检测器(SSD)算法,增强算法在检测遮挡对象时的鲁棒性。将SSD基础网络替换成Resnet,增强特征提取能力。以实际应用中数据集真实框大小为依据,重新设计SSD检测时默认框的生成比例。通过添加SE模块,使模型在训练过程中增强重要通道信息。利用双向融合特征金字塔网络(BiFPN)对不同检测层中的特征进行融合。实验结果表明,改进后的SSD目标检测算法在自制的浪费行为数据集中检测精度达到88.49%,相比原始SSD提高了5.09%。To identify the food waste behavior more accurately in the complex environment of real life,the SSD algorithm was improved to enhance the robustness of the algorithm in detecting overlapping objects.The SSD basic network was replaced by Resnet to enhance the feature extraction ability of SSD.According to the actual frame size of data set in practical application,the generation ratio of default frame in SSD detection was redesigned.By adding SE module,the model enhanced the important channel information in the training process.Bi-directional fusion feature pyramid network(BiFPN)was used to fuse the features in different detection layers.Experimental results show that the detection accuracy of the improved SSD is 88.49%in the self-made waste behavior dataset,which is 5.09%higher than that of the original SSD.

关 键 词：浪费行为检测 遮挡对象 单激发多盒检测器 特征提取 默认框比例 通道注意力模块 双向融合特征金字塔网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]