改进的YOLOv3算法在棉花识别中的应用  

Application of Improved YOLOv3 Algorithm in Cotton Identification

作　　者：依沙·吾阿提别克 古丽孜亚·艾布列孜 YISA Watbek;GULZYA Ablez(Department of Computer Engineering,Changji University,Changji Xinjiang 831100,China;School of Mathematics and Data Science,Changji University,Changji Xinjiang 831100,China)

机构地区：[1]昌吉学院信息科学与工程学院,新疆昌吉831100 [2]昌吉学院数学与数据科学学院,新疆昌吉831100

出　　处：《信息与电脑》2022年第13期175-177,共3页Information & Computer

摘　　要：棉花是一种密集性农作物,传统YOLOv3算法在识别密集性目标方面准确率较低。为了解决此问题,提出了一种基于改进的YOLOv3算法的棉花识别方法。在传统的YOLOv3算法框架基础上,先加上一道多尺度特征检测通道,使算法更能识别密集性目标,再自制棉花识别数据集,并使用改进的YOLOv3模型进行实验。结果表明,检测速度高达56.4 fps,目标精度为88.55%,可以完成实际环境中的棉花识别任务。Cotton is a kind of dense plant,and the accuracy of traditional YOLOv3 in identifying dense targets is low.In order to solve this problem,an improved target detection method of YOLOv3 cotton recognition is proposed in this paper.Firstly,based on the traditional YOLOv3 framework,a multi-scale feature detection channel is added to make the algorithm more able to identify dense targets.Secondly,the self-made cotton recognition data set is used in the experiment with the improved YOLOv3 model.The detection speed is as high as 56.4 fps and the target accuracy is 88.55%,which can complete the cotton recognition task in the actual environment.

关 键 词：棉花识别 YOLOv3 棉花检测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]