摘要
脐橙的高精度实时检测是智能化采摘的关键技术之一。采用YOLOv8s模型开展脐橙检测研究,并进一步从数据集扩充和模型改进两个方面解决误检和漏检问题。数据集扩充方面,通过加雾的方式扩充脐橙训练样本,使得模型能识别出图像中光线不佳区域处的脐橙。YOLOv8s模型改进方面,通过增加检测头和合并块注意力模块,以分别检测出被树叶遮挡和被其他脐橙遮挡的脐橙。实验结果表明,改进后的方法能获得更高的精确率、召回率和平均精确率均值。
The high-precision real-time detection of navel orange is one of the key technologies for intelligent harvesting.We use YOLOv8s to conduct research on the navel orange detection and fur-ther solve the issues of error and missed detections from two aspects:dataset expansion and model im-provement.In terms of dataset expansion,the training samples are expanded by adding fog,enabling the model to recognize navel oranges in areas with poor lighting in the image.In terms of improving the YOLOv8s model,a detection head and a merged block attention module are added to detect navel oran-ges obscured by leaves and other navel oranges.Experimental results show that the improved method a-chieves higher accuracy,recall,and average accuracy than directly using YOLOv8s.
作者
陈骁扬
罗宇航
汪成江
罗坤
黄帅永
喻玲娟
CHEN Xiaoyang;LUO Yuhang;WANG Chengjiang;LUO Kun;HUANG Shuaiyong;YU Lingjuan(School of Information Engineering,Jiangxi University of Science and Technology,Ganzhou Jiangxi 341000,China)
出处
《佳木斯大学学报（自然科学版）》
CAS
2024年第7期18-21,共4页
Journal of Jiamusi University：Natural Science Edition
基金
江西省大学生创新创业训练计划项目(202110407029)。
