基于电子地图的杂草图像处理与识别方法分析  

Analysis of Weed Image Processing and Recognition Methods Based on Electronic Map

作　　者：宋雅丽[1] SONG Yali(Department of Information Engineering,Anhui Industry Polytechnic,Tongling Anhui 244000,China)

机构地区：[1]安徽工业职业技术学院信息工程系,安徽铜陵244000

出　　处：《佳木斯大学学报（自然科学版）》2023年第6期66-69,共4页Journal of Jiamusi University：Natural Science Edition

基　　金：2022年度安徽省科研编制计划项目(2022AH053156)。

摘　　要：当前智慧农业中杂草对农业的威胁与日俱增,为了降低田间杂草对农作物的实际危害,研究在运用电子地图对杂草图像进行检测和处理的基础上提出了YOLOv4模型,并引入MobileNet提出MobileNet-YOLOv4模型,同时实验验证了二者的性能。实验结果表明,YOLOv4模型的平均准确率达到了98%。MobileNetV3-YOLOv4的平均准确率高达97%。同时,MobileNetV3-YOLOv4模型体积为44.4 MB,远低于原模型。综合来看,将MobileNetV3作为主干提取网络的YOLOv4模型具有较高的杂草图像识别准确性,在实际的除草设备中具有较强的实用性。The threat of weeds to agriculture in smart agriculture is increasing day by day.In order to reduce the actual harm of weeds to crops in the field,the YOLOv4 model was proposed based on the detection and processing of weed images using electronic maps,and the MobileNet YOLOv4 model was introduced.At the same time,the performance of both models was experimentally verified.The experimental results show that the average accuracy of the YOLOv4 model reaches 98%.The average accuracy of MobileNetV3-YOLOv4 is as high as 97%.Meanwhile,the MobileNetV3-YOLOv4 model has a volume of 44.4 MB,which is much lower than the original model.Overall,the YOLOv4 model using MobileNetV3 as the backbone extraction network has high accuracy in weed image recognition and strong practicality in actual weed control equipment.

关 键 词：电子地图 杂草 图像处理 识别方法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]