基于深度可分离卷积的果蔬识别方法研究  

Recognition method of fruits and vegetables based on depthwise separable convolution

作　　者：刘斌 陈刚 袁浩 李志海 LIU Bin;CHEN Gang;YUAN Hao;LI Zhi-hai(Hubei Institute of Logistics Technology,Xiangyang 441002,Hubei Province,China)

机构地区：[1]湖北物资流通技术研究所,湖北襄阳441002

出　　处：《信息技术》2023年第11期92-98,共7页Information Technology

基　　金：湖北省重点研发计划项目(2020BBB093);湖北省科技发展专项项目(KJFZC2022000001)。

摘　　要：在无人零售行业的果蔬检测过程中,针对现有检测算法出现的目标漏检、误检和检测速度慢的问题,提出了一种基于深度可分离卷积的改进Faster R-CNN检测算法。该算法首先使用一个与DIOU正相关的惩罚函数对候选框的置信度得分进行惩罚,然后通过对冗余框的加权平均融合候选区域的上下文信息,将算法中的标准卷积替换为深度可分离卷积。实验表明,该算法可有效选择候选区域,降低了目标被漏检和误检的概率,同时提升了算法的检测速度和识别精度,具有重要的理论和应用价值。In the fruit and vegetable detection scene of the unmanned retail industry,this paper proposes an improved Faster R-CNN algorithm based on depth separable convolution to resolve the problems of missed detection,false detection and slow detection speed in the existing detection algorithms.The algorithm first uses a penalty function that is positively related to DIOU to punish the confidence score of the candidate box,then fuses the context information of the candidate area by the weighted average of the redundant box,and finally replaces the standard convolution with depth separate convolution.The experiments show that the algorithm can effectively select candidate regions,reduces the probability of missed and false detection of the targets,improves the detection speed and recognition accuracy which has important theoretical and application value.

关 键 词：果蔬识别 目标检测 距离交并比 候选区域优化 深度可分离卷积 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]