REGIONAL FUEL MAPPING USING AN OBJECT-ORIENTED CLASSIFICATION OF QUICKBIRD IMAGERY
REGIONAL FUEL MAPPING USING AN OBJECT-ORIENTED CLASSIFICATION OF QUICKBIRD IMAGERY
复制标题
使用面向对象的 Quickbird 图像分类绘制区域燃料图
DOI：
--
复制
发表时间：
2005
期刊：
影响因子：
--
通讯作者：
J. Manzanera
中科院分区：
文献类型：
--
作者：
Arroyo;J. Manzanera
燃料负荷和组成的知识是至关重要的，以改善目前的防火和模拟程序，以减轻火灾对生态系统的负面影响。通常，从遥感图像生成燃料分布图是基于对中等分辨率传感器的分析，如大地卫星数据。本文提出了一种方法，以产生燃料类型地图的遥感数据在高空间分辨率。普罗米修斯系统，燃料类型分类适应欧洲地中海盆地的生态特征，被采用这项研究。燃料分布图来自QuickBird图像。这颗卫星是目前运行的分辨率最高的商业遥感卫星，可提供分辨率从0.61米到2.88米的多光谱和全色图像。这些照片拍摄于2002年7月，位于马德里地区的西北部。预处理包括使用地面控制点的正射校正和通过与全色数据的分辨率合并将多光谱数据恢复到70厘米像素。对于高分辨率图像，感兴趣的对象通常是表现出各种光谱特性的像素的集合。通过分类正确识别这些物体需要明确考虑每个像素的空间背景。我们使用面向对象的方法作为传统基于像素的技术的补充;它允许在分类过程中显式考虑空间背景。这种面向对象分析的第一步是多尺度图像分割。在迭代步骤中，开发了图像对象的分层网络。不同尺度上的图像信息的同时表示允许识别由较小的共同出现的子对象组成的图像对象。通过林分调查网络评估了通过这种方法创建的普罗米修斯燃料地图的准确性。结果令人鼓舞，表明对高分辨率图像进行面向对象的分类有可能制作准确和高度精确的燃料图。
The knowledge of fuel load and composition is critical for improving current fire prevention and modeling programs and to alleviate the negative effects of fire on the ecosystem. Commonly, the generation of fuel maps from remote sensing images has been based in the analysis of medium resolution sensors, such as Landsat data. This paper presents a methodology to generate fuel type maps from remote sensing data at a high spatial resolution. The Prometheus system, a fuel type classification adapted to the ecological characteristics of the European Mediterranean basin, was adopted for this study. Fuel maps were derived from QuickBird imagery. This satellite is the highestresolution commercial remote sensing satellite now operating, offering multispectral and panchromatic imagery ranging from .61 to 2.88 m resolution. The images, dated from July 2002, was located in the NW of Madrid Region. Preprocessing consisted of orthorectificatation using ground control points and resampling of the multi-spectral data to 70 cm-pixels through a resolution merge with the panchromatic data. With high-resolution imagery, objects of interest are often aggregations of pixels exhibiting a variety of spectral properties. Correct identification of these objects through classification requires explicit consideration of the spatial context of each pixel. We used object-oriented approach as a complement to traditional pixel-based techniques; it allowed explicit consideration of spatial context during the classification process. The first step of this object-oriented analysis was multi-scale image segmentation. In iterative steps, a hierarchical network of image objects was developed. The simultaneous representation of image information on different scales allowed the identification of image objects composed of smaller, co-occurring sub-objects. The accuracy of the Prometheus fuel map created through this approach was assessed through a network of stand surveys. Results were encouraging and suggest that objectoriented classification of high-resolution imagery has the potential to create accurate and highly precise fuel maps.