华中师范大学学报（自然科学版）2026，Vol.60Issue(2)：271-283,13.DOI:10.19603/j.cnki.1000-1190.2026.02.010
基于改进UNet3+的粗梗水蕨生长状态智能视觉监测及其GPI综合评估方法
Intelligent visual monitoring of growth status and comprehensive GPI evaluation method for Ceratopteris pteridoides based on improved UNet3+
摘要
Abstract
Ceratopteris pteridoides,an aquatic fern classified as a second-level nationally protected wild plant,holds significant ecological value.Scientifically monitoring its growth status is crucial for elucidating its endangerment mechanisms and formulating conservation strategies.In this study,computer vision technology was employed to collect and construct a comprehensive image dataset covering its entire growth cycle.Building upon the UNet3+image segmentation model,a Euclidean distance self-attention mechanism was introduced to enhance the model's efficiency in utilizing global contextual information and its ability to retain fine features,thereby improving segmentation accuracy for aquatic plants with complex leaf morphologies and slender stems.Based on the image segmentation results,morphological parameters such as canopy coverage,shape factor,and greenness index were extracted.A comprehensive growth status monitoring index(GPI)is constructed to enable quantitative analysis of the growth status of Ceratopteris pteridoides.The results indicated that the GPI values were less than 0.27 during the seedling stage,between 0.27 and 0.44 during the rapid growth stage,and greater than 0.44 at the maturity stage,demonstrating the effectiveness of GPI in distinguishing growth stages.These findings provide a technical reference for the intelligent monitoring of growth status in other aquatic plants.关键词
粗梗水蕨/图像分割/欧氏距离自注意力/生长状态监测/量化评估Key words
Ceratopteris pteridoides/image segmentation/Euclidean distance selfattention/growth status monitoring/quantitative assessment分类
农业科技引用本文复制引用
胡慧莉,叶曦,曾长立,董元火..基于改进UNet3+的粗梗水蕨生长状态智能视觉监测及其GPI综合评估方法[J].华中师范大学学报（自然科学版）,2026,60(2):271-283,13.基金项目
湖北省自然科学基金项目(2023AFB462). （2023AFB462）
