上海农业学报2025，Vol.41Issue(2)：113-118,6.DOI:10.15955/j.issn1000-3924.2025.02.17
基于改进灰度特征算法的单株作物高光谱图像批量分割研究
Batch segmentation of single crop hyperspectral images based on improved gray feature algorithm
摘要
Abstract
In order to solve the problem of high-precision batch segmentation of hyperspectral images in crop stress experiments,an image gray-scale transformation algorithm with improved gray feature was proposed,which combined the characteristics of hyperspectral multi-channel and high spectral bands,generated grayscale image by reflectance operations in the 450 nm,600 nm,670 nm,and 700 nm to 1 000 nm bands,and improved the contrast between the target plant and the background area in the gray feature image,reduced the difficulty of threshold segmentation,and then used the Otsu algorithm to segment the image.The results showed that the proposed improved gray feature algorithm combined with the Otsu algorithm achieved Rand index and Jaccard similarity coefficient values as high as 0.982 3 and 0.979 9,respectively,in different plant segmentation tasks.These results were better than those obtained using the commonly used extra-green,near-infrared and normalized difference vegetation index.This method effectively removed various backgrounds from hyperspectral images of crops and extracted green leaves while preserving the discoloration and stress-induced chlorosis,yellowing,and wilting of the leaves.Ultimately,it enabled the batch precise segmentation of hyperspectral images of individual plants.关键词
高光谱成像/图像处理/灰度特征/大津算法Key words
Hyperspectral imaging/Image processing/Gray feature/Otsu algorithm分类
信息技术与安全科学引用本文复制引用
王漫,吴杰,田明璐,班松涛,胡冬,李琳一..基于改进灰度特征算法的单株作物高光谱图像批量分割研究[J].上海农业学报,2025,41(2):113-118,6.基金项目
上海市科技兴农技术创新项目(2022-02-08-00-12-F01183) （2022-02-08-00-12-F01183）
