用SAM模型+多阶段优化，精准分割卫星图像中的橄榄树
Olive Tree Satellite Image Segmentation Based On SAM and Multi-Phase Refinement
- 基于SAM模型并结合田间树列对齐与可学习形状约束进行修正
- 分割准确率提升至98%，远超原始SAM的82%
- 适合农业遥感、智能林业监测等场景使用
在气候变化日益严峻的背景下，利用遥感技术实现橄榄树早期异常检测与及时干预，对维护橄榄生物多样性至关重要。本文提出一种创新的橄榄树卫星图像分割方法，通过引入基础模型与先进分割技术，融合分割一切模型（SAM）以精准识别农业地块中的橄榄树。该方法包括基于田间树列对齐的SAM分割结果修正，以及关于形态和尺寸的可学习约束。实验结果显示，该方法分割准确率达98%，显著优于初始SAM的82%性能。
原文摘要 · Abstract (English)
In the context of proven climate change, maintaining olive biodiversity through early anomaly detection and treatment using remote sensing technology is crucial, offering effective management solutions. This paper presents an innovative approach to olive tree segmentation from satellite images. By leveraging foundational models and advanced segmentation techniques, the study integrates the Segment Anything Model (SAM) to accurately identify and segment olive trees in agricultural plots. The methodology includes SAM segmentation and corrections based on trees alignement in the field and a learanble constraint about the shape and the size. Our approach achieved a 98\% accuracy rate, significantly surpassing the initial SAM performance of 82\%.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。