基于语义分割的粮仓粮食数量变化动态监测方法
Dynamic Monitoring Methodology of Grain Quantity Variation in Granaries Based on Semantic Segmentation
李智 1张艳飞 1杨卫东 2但乃禹 1张蕙 3陈卫东 4荆世华 5邵辉 5任飞燕1
作者信息
- 1. 河南省粮食光电探测与控制重点实验室,郑州 450001;河南工业大学信息科学与工程学院,郑州 450001
- 2. 河南省粮食光电探测与控制重点实验室,郑州 450001;河南工业大学人工智能与大数据学院,郑州 450001
- 3. 河南工业大学信息科学与工程学院,郑州 450001
- 4. 河南工业大学粮食和物资储备学院,郑州 450001;粮食储运国家工程研究中心,郑州 450001
- 5. 浪潮数字粮储科技有限公司,济南 250098
- 折叠
摘要
为了高效、准确且低成本监测粮仓粮食数量变化情况,研究提出一种基于语义分割的粮仓粮食数量变化动态监测方法,利用深度学习技术对粮仓仓内摄像机采集的图像进行分析,实现对粮仓仓内粮食数量变化情况的动态监测.通过将监测结果与仓内近期业务数据进行比对,可及时发现违法违规行为线索并预警,提高日常监管的针对性和效率.本研究选取粮仓仓内监控摄像机采集的图像作为数据集,构建了基于DeepLabV3+的粮仓粮食数量变化动态监测模型,通过提取判断粮面变化的参照边界,利用参照边界像素值的变化判断仓内粮食数量变化情况,并通过引入基于MobileNetV2的特征提取网络,提高了模型识别的准确性和计算效率.实验结果表明,该模型平均交并比和平均像素准确率分别达到89.57％和94.53％,参数量为5.818 M,MIoU分别比PSPNet模型和UNet模型高0.95％和0.88％.通过对50个粮仓的测试分析,模型识别得到的仓内粮食数量变化情况与实际情况的一致性为96％,验证了该方法的有效性,为粮仓粮食数量的动态监测提供了新的思路.
Abstract
In order to efficiently,accurately and cost-effectively monitor changes in the quantity of grain in grain warehouses.In this paper,a dynamic monitoring model was presented for grain storage based on semantic seg-mentation.In this approach,deep learning techniques were utilized to analyze and process images collected by cam-eras inside the granary,enabling the dynamic monitoring of the changes in grain quantities inside the granary.Final-ly,the monitoring results were compared with recent business data to identify any illegal behavior during routine su-pervision and provide timely feedback to the grain storage regulators.The proposed approach enhanced the targeting and efficiency of grain inventory inspections.In this paper,the images collected by the monitoring camera in the granary were selected as the data set,and a dynamic monitoring model of granary grain quantity change based on DeepPlabv 3+was constructed.By extracting the reference boundary to judge the grain surface change,the change of grain quantity in the granary was judged by the change of the pixel value of the reference boundary,and by intro-ducing the feature extraction network based on MobileNetV2,the accuracy and calculation efficiency of model identi-fication were improved.The experimental results indicated that this model had the mean intersection over union and mean pixel accuracy reaching 89.57％and 94.53％,respectively,with the number of parameters of 5.818 M.This model improved the Mean Intersection over Union by 0.95％and 0.88％compared to PSPNet and UNet models,re-spectively.Through the testing analysis of 50 grain silos,the consistency between the change of grain quantity in the silo obtained by the model identification and the actual situation was 96％,demonstrating the effectiveness of this method and providing a new idea for the dynamic monitoring of grain quantity in grain silos.
关键词
深度学习/DeepLabV3+/粮面识别/语义分割Key words
deep learning/DeepLabV3+/grain surface recognition/semantic segmentation引用本文复制引用
基金项目
国家重点研发计划项目(2017YFD0401001-02)
河南省杰出青年基金项目(222300420004)
河南省重大公益专项(201300210100)
出版年
2024
NETL
NSTL
万方数据