中文摘要：

红外遥感数据是获取火灾参数的主要方法，中分辨率成像光谱仪（MODIS）因具有广泛的光谱覆盖波段和强大的动态监测能力被用于森林火灾监测。传统方法主要是基于光谱特征的火烧迹地识别及制图，而描述火灾蔓延的边缘火线对于动态火灾监测具有更重要的意义。本文基于多波段影像梯度的图像视觉分析，提出一种自动检测林火边界火线特征的方法。通过与传统边缘算子的检测结果对比，并结合火烧迹地的归一化植被指数NDVI叠置验证，本方法在融合多波段影像的火灾光谱信息的同时，增强了火线变化的空间细节表现能力，提高了边缘定位性能，在动态火灾制图l和监测方面是有效可行的。

英文摘要：

Remotely sensed infrared images are often used to estimate wildfire parameters. MODIS （moderate resolution imaging spectroradiameter） is widely applied to monitor forest wildfire due to its multispectral images and capability of real-time acguirlng fire data. The traditional mapping methods of burn scar and hotspots depend on its particular radiant and thermal spectral characters. In dynamic fire monitoring process, the fire line detection of the edge of burn scar is more important than burn area estimation. An automatic method for extracting fire line parameters was developed in this paper based on muhispectral imagery. The method integrated the fire edge information of muhispectral images, and as well as improve spatial details mapping of fire line by multi-bands image gradient magnitude calculation. Superimposition of the fire lines on the edge of burn scar at Normal Difference Vegetation Index （NDVI） image produced a coincident result which suggests the algorithm here on infrared images is better than sobel edge-detection operators. The method was also proved with good performance and efficient in dynamic fire mapping and monitoring.