基于数码照片的草地植被覆盖度快速提取方法  

Quickly determining grassland cover using the digital image

作　　者：胡健波[1] 张璐[2] 黄伟[1] 吴世红[1] 刘长兵[1] 

机构地区：[1]水路交通环境保护技术交通行业重点实验室交通运输部天津水运工程科学研究所,天津300456 [2]河南理工大学测绘与国土信息工程学院,河南焦作454000

出　　处：《草业科学》2011年第9期1661-1665,共5页Pratacultural Science

基　　金：中央级公益性科研院所基本科研业务费专项资金(TKS090303)

摘　　要：草地植被覆盖度是草地健康与否的重要参数,提高草地植被覆盖度的计算效率和精度具有重要的意义。本研究提出了一种利用过绿特征植被指数和半自动阈值设定算法(半自动阈值法)的从数码照片中快速计算草地植被覆盖度的方法,并将该方法与最大似然法监督分类方法(最大似然法)和色度饱和度法进行了比较。对32张草地数码照片的植被覆盖度估算结果和准确结果进行回归分析,分别得到斜率、截距和回归系数这3个参数。结果表明,3种方法都未能考虑植物的非绿色组分而存在低估的问题。半自动阈值法的准确度堪比最大似然法,且没有后者存在的低覆盖率高估问题;而且该方法人工干预少,计算结果准确客观,适用性强;但是对绿色特征不明显的植物(如灰绿色植物)效果不佳。The improving efficiency and accuracy of vegetation cover is very important because the cover is a necessary parameter to estimate the health of grassland ecosystem. This study proposed a quickly calculating vegetation cover method by Excess Green Index and semi-automatic threshold （ST）. Efficiency of the algorithm was evaluated by comparing with Maximum-likelihood supervised classification （MC） and Hue-Saturation method （HS）. 32 digital images were used to test the efficiency of three methods. Vegetation cover estimated by each method was regressed with accurate cover obtained by visual interpretation via slope, intercept, and regression coefficient. The three methods did not consider non-green component of vegetation, resulting in slightly underrating problem. In conclusion, ST not only achieved similar accuracy with MC but also avoided overrating problem of MC when cover was low. ST showed little human intervention, and achieved accurate cover; however, it failed to extract grey-green vegetation.

关 键 词：数码照片 植被覆盖度 过绿特征 半自动阈值 

分 类 号：S812[农业科学—草业科学]