基于Segment Anything的荒漠草原植被覆盖度提取研究

Study on Extraction of Desert Vegetation Coverage Based on Segment Anything

植被覆盖度(Fractional Vegetation Coverage,FvC)是草原监测评价工作的重要指标,实时、快速、准确地采集FVC是草原监测评价工作的基础。文章以解决荒漠草原单位面积的植被覆盖度为目标,选取分割一切模型(Segment Anything Model,SAM)以打点或者画框的方式对单位面积的图像进行分割,计算分割出的植被所占像素点的总数T和图像总像素点数A,然后通过F=T/A计算单位面积的植被覆盖度,并制作植被覆盖度评估系统。实验的结果表明,基于SAM大模型可以使荒漠草原单位面积的植被覆盖度提取结果更加准确和可靠,减小了噪声对覆盖度提取的影响。

Fractional Vegetation Coverage(FVC)is one of the important indicators for grassland monitoring and evaluation.Real-time,rapid,and accurate acquisition of FVC forms the foundation of grassland monitoring and evaluation efforts.In this paper aims to solve the vegetation coverage of desert steppe per unit area,select the SAM model,segment the image per unit area in the way of dot or frame,calculate the total number of pixels of vegetation T and the total number of pixels A,and then calculate the vegetation coverage per unit area by F=TlA,and made the vegetation coverage assessment system.The experimental results show that the Segment Anything Model(SAM)model can make the extraction results of vegetation coverage per unit area more accurate and reliable,and reduce the influence of noise on the coverage extraction.