期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2010
卷号：XXXVIII - Part 7B
页码：26-31
出版社：Copernicus Publications
摘要：Riparian forests play an important role in the ecological balance of river ecosystems. Given the narrow nature of these environments, medium resolution sensors such as Landsat have limited use. Conversely, products obtained from high-resolution images, such as Ikonos-2, have wide applications in riparian forest studies. The objective of this article is to describe a methodology for delineating riparian areas and extracting their biophysical parameters from an Ikonos scene. The methodology is divided into two stages. Firstly, the image is segmented into a riparian forest class and non-riparian classes using a segmentation algorithm and a river-based buffer. The segmentation package MAGIC (Map Guide Image Classification) was used to separate the riparian forest zones from the rest. In the second phase, texture features derived the co-occurrence matrix were used to estimate the biophysical parameters of the riparian forest. Allometric measurements were made in 70 plots of riparian area from both sides of the P andeiros River, located in Northern Minas Gerais, Brazil. These plots were used to calibrate and validate our models based on texture parameters. The forest structure variables included height, diameter at breast height, basal area, stem density, volume, canopy openness and leaf area index which were acquired by direct measurements in the field. The results show that MAGIC segmented the riparian environment with an accuracy of more than 85% when compared with the map obtained by visual image interpretation. The best results for modeling riparian structure were obtained respectively for volume and basal area (R 2 =0.66 and R 2 =0.61) using Angular Second Moment, Entropy, Infrared band, distance analysis of four pixels and a window of 11×11 pixels
关键词：Segmentation; Knowledge Base; Spatial; Texture; Analysis; Vegetation; High resolution
