基于深度学习的粮库测控技术研究进展  

Research Progress of Grain Depot Measurement and Control Technology Based on Deep Learning

作　　者：高辉[1] 甄彤[1] 李智慧[1] Gao Hui;Zhen Tong;Li Zhihui(Key Laboratory of Grain Information Processing and Control,College of Information Science and Engineering,Henan University of Technology,Zhengzhou 450001)

机构地区：[1]粮食信息处理与控制重点实验室,河南工业大学信息科学与工程学院,郑州450001

出　　处：《中国粮油学报》2021年第2期162-171,共10页Journal of the Chinese Cereals and Oils Association

基　　金：国家科技支撑计划(2017YFD0401004)。

摘　　要：粮库负责国家及地方粮食储备,粮库测控技术发展与储粮安全息息相关。基于Faster R-CNN、YOlO、SSD、CNN、BRNN、双向LSTM、深度自编码器及各种优化、级联、融合等网络,阐述了深度学习在车辆检测、车牌识别、车型识别及车辆跟踪中的应用;结合BP、改进CNN、multi-CNN、双流卷积等网络,阐述了深度学习在人员识别、行为识别中的应用;依据R-FCN、Faster R-CNN、YOLO、GANPSO-BP、改进LSTM网络等技术,阐述了深度学习在粮库烟火识别、粮虫识别、粮堆温度预测、粮食籽粒计数、粮库图像去雾的研究进展。总结各种传统算法的优缺点,阐述深度学习在新一代粮库测控技术的研究进展及其优缺点,并做出展望。Grain depot is responsible for national and local grain reserves,the development of measurement and control technology of grain depot is closely related to grain storage safety.Based on Faster R-CNN,Yolo,SSD,CNN,BRNN,bidirectional LSTM,Depth self encoder and various optimization,cascading,fusion networks,etc.,the application of depth learning in vehicle detection,license plate recognition,vehicle type recognition and vehicle tracking was elaborated;combined with BP,improved CNN,multi CNN,double flow convolution and other networks,the application of depth learning in personnel recognition and behavior recognition was elaborated;According to R-FCN,Faster R-CNN,YOLO,GANPSO-BP,improved LSTM network and other technologies,the research progress of in-depth learning in the recognition of fireworks,recognition of grain insects,temperature prediction of grain pile,grain count and image defogging of grain depot was described.The advantages and disadvantages of various algorithms were summarized,The research progress and advantages and disadvantages of deep learning in the new generation of grain depot measurement and control technology were expounded,and a prospect was made.

关 键 词：粮库烟火识别 粮堆温度预测 粮虫识别 粮库图像去雾 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]