PDF(Pubmed)
Abstract：
The mango fruit plays a crucial role in providing essential nutrients to the human body and Pakistani mangoes are highly coveted worldwide. The escalating demand for agricultural products necessitates enhanced methods for monitoring and managing agricultural resources. Traditional field surveys are labour-intensive and time-consuming whereas remote sensing offers a comprehensive and efficient alternative. The field of remote sensing has witnessed substantial growth over time with satellite technology proving instrumental in monitoring crops on a large scale throughout their growth stages. In this study, we utilize novel data collected from a mango farm employing Landsat-8 satellite imagery and machine learning to detect mango orchards. We collected a total of 2,150 mango tree samples from a farm over six months in the province of Punjab, Pakistan. Then, we analyzed each sample using seven multispectral bands. The Landsat-8 framework provides high-resolution land surface imagery for detecting mango orchards. This research relies on independent data, offering an advantage for training more advanced machine learning models and yielding reliable findings with high accuracy. Our proposed optimized CART approach outperformed existing methods, achieving a remarkable 99% accuracy score while the k-Fold validation score also reached 99%. This research paves the way for advancements in agricultural remote sensing, offering potential benefits for crop management yield estimation and the broader field of precision agriculture.
摘要：
芒果果实在为人体提供必需营养方面起着至关重要的作用，巴基斯坦芒果在全球范围内备受追捧。对农产品不断增长的需求要求加强监测和管理农业资源的方法。传统的实地调查是劳动密集型和耗时的，而遥感提供了一个全面和有效的替代方案。随着时间的推移，遥感领域取得了长足的发展，卫星技术被证明有助于在整个生长阶段对作物进行大规模监测。在这项研究中,我们利用从使用Landsat-8卫星图像和机器学习的芒果农场收集的新数据来检测芒果园。我们在旁遮普省六个月的时间里从一个农场收集了总共2150棵芒果树样本，巴基斯坦。然后，我们使用七个多光谱波段分析了每个样本。Landsat-8框架提供了高分辨率的地表图像，用于检测芒果园。这项研究依赖于独立的数据，为训练更高级的机器学习模型提供优势，并以高精度产生可靠的发现。我们提出的优化CART方法优于现有方法，取得了显著的99%的准确性得分，而k-Fold验证得分也达到了99%。这项研究为农业遥感的发展铺平了道路，为作物管理产量估算和更广泛的精准农业领域提供潜在的好处。
