基于叶片数量与类别占比的香蕉树长势检测  

Banana Tree Growth Detection Based on Leaf Quantity and Category Proportion

作　　者：米佳豪 蒋权 MI Jiahao;JIANG Quan(Guangxi Minzu University,Nanning 530006,China)

机构地区：[1]广西民族大学,广西南宁530006

出　　处：《现代信息科技》2023年第23期106-110,115,共6页Modern Information Technology

摘　　要：为了提高香蕉长势的分类准确度,依据香蕉生长特性,文章提出了一种基于叶片数量和类别占比的香蕉长势检测方法。针对香蕉树主要的3个长势区间,利用YOLOv5s目标检测算法为基础构建模型,在检测端根据模型的预测结果对置信度低于0.8的目标进行筛选,依据叶片数与类别占比的分类矫正方法对模型输出的低置信结果进行矫正。结果表明,经过矫正后的模型与基础模型相比分类更加准确。In order to improve the classification accuracy of banana growth,based on the growth characteristics of bananas,this paper proposes a banana growth detection method based on the number of leaves and the proportion of categories.Based on the YOLOv5s object detection algorithm,a model is constructed for the three main growth intervals of banana trees.At the detection end,targets with a confidence level lower than 0.8 are selected based on the predicted results of the model.The low confidence output of the model is corrected using a classification correction method based on the number of leaves and the proportion of categories.The results show that corrected model is more accurate in classification compared to the basic model.

关 键 词：长势检测 分类矫正 YOLOv5s 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]