| LU-Re Net植物叶片分割与计数模型 |
| |
| 引用本文： | 邹龙,王以松,魏昆鹏,彭啟文,王玮,王运乾.LU-Re Net植物叶片分割与计数模型[J].计算机工程与设计,2022,43(2):553-559. |
| |
| 作者姓名： | 邹龙 王以松 魏昆鹏 彭啟文 王玮 王运乾 |
| |
| 作者单位： | 贵州大学 计算机科学与技术学院,贵州 贵阳 550025;贵州大学 计算机科学与技术学院,贵州 贵阳 550025;贵州大学 公共大数据国家重点实验室,贵州 贵阳 550025;中国移动通信集团 设计院有限公司,河南 郑州 450003 |
| |
| 基金项目： | 国家自然科学基金项目(61976065、U1832605)。 |
| |
| 摘 要： | 为完成一种植物叶片分割与计数模型LU-Re Net,利用Xception module和Coord Conv改进模型编码器,通过改进的Re Net模型并参照Unet模型结构将CNN和RNN结合在一起。在CVPPP的A1、A2、A3基准数据集上,LU-Re Net模型取得了0.95的平均F1-score、95.54%的平均前景-背景Dice(FBD%)、0.95的平均召回率和1.91的平均计数差分绝对值(|Dic|)。在A2基准数据集上,该模型表现超越大部分最新算法。实验结果表明,该网络在拥有高分割准确率的同时具有良好的泛化能力。
|
| 关 键 词： | 叶片分割 植物表型 深度神经网络 卷积神经网络 叶片计数 |
LU-ReNet plant leaf segmentation and counting model |
| |
| ZOU Long,WANG Yi-song,WEI Kun-peng,PENG Qi-wen,WANG Wei,WANG Yun-qian.LU-ReNet plant leaf segmentation and counting model[J].Computer Engineering and Design,2022,43(2):553-559. |
| |
| Authors: | ZOU Long WANG Yi-song WEI Kun-peng PENG Qi-wen WANG Wei WANG Yun-qian |
| |
| Affiliation: | (School of Computer Science and Technology,Guizhou University,Guiyang 550025,China;State Key Laboratory of Public Big Data,Guizhou University,Guiyang 550025,China;Design Institute Limited Corporation,China Mobile Communications Group,Zhengzhou 450003,China) |
| |
| Abstract: | To complete a deep learning approach for the plant leaf segmentation and counting model LU-ReNet,the Xception model and CoordConv were improved in the model encoder,and CNN and RNN were combined through Unet and improved ReNet.On the A1,A2 and A3 benchmark data-sets of CVPPP,LU-ReNet achieves the average F1-score of 0.95,the average foreground background Dice(FBD%)of 95.54%,the average recall rate of 0.95,and the average count difference absolute value of 1.91(|Dic|).On the A2 benchmark dataset,the performance of the model surpasses most of the latest approaches.It shows that the model has high segmentation accuracy and generalization ability. |
| |
| Keywords: | leaf segmentation plant phenotype deep neural network convolution neural network leaf count |
| 本文献已被 维普 万方数据 等数据库收录！ |
|
