ABSTRACT

Remote sensing is based on the extraction of data, acquired by satellites or aircrafts, through multispectral images, that allow their remote analysis and classification. Analysing those images with data fusion techniques is a promising approach for identification and classification of forest types. Fusion techniques can aggregate various sources of heterogeneous information to generate value-added maps, facilitating forest-type classification. This work applies a data fusion algorithm, denoted FIF (Fuzzy Information Fusion), which combines computational intelligence techniques with multicriteria concepts and techniques, to automatically distinguish Eucalyptus trees from satellite images. The algorithm customization was performed with a Portuguese area planted with Eucalyptus. After customizing and validating the approach with several representative scenarios to assess its suitability for automatic classification of Eucalyptus, we tested on a large tile obtaining a sensitivity of 69.61%, with a specificity of 99.43%, and an overall accuracy of 98.19%. This work demonstrates the potential of our approach to automatically classify specific forest types from satellite images, since this is a novel approach dedicated to the identification of eucalyptus trees.

摘要

遥感基于通过多光谱图像对卫星或飞机获取的数据进行提取，从而可以对其进行远程分析和分类。使用数据融合技术分析这些图像是用于森林类型识别和分类的有前途的方法。融合技术可以聚合各种异构信息源，以生成增值图，从而促进森林类型的分类。这项工作应用了称为FIF（模糊信息融合）的数据融合算法，该算法将计算智能技术与多准则概念和技术相结合，以自动将桉树与卫星图像区分开。算法定制是在种植有桉树的葡萄牙地区进行的。在使用几种代表性方案定制并验证了该方法以评估其对桉树自动分类的适用性之后，我们在一块大砖上进行了测试，灵敏度为69.61％，特异性为99.43％，总准确度为98.19％。这项工作展示了我们从卫星图像自动分类特定森林类型的方法的潜力，因为这是致力于鉴定桉树的新颖方法。