一种增强的基于上下文火点遥感影像识别方法

传统的火点遥感影像识别方法大多采用阈值法,但阈值的选择受区域、季节以及云天状况等多种因素的限制,因而在实际监测中往往效果不佳。针对这些问题,提出了一种增强的基于上下文信息的火点遥感影像识别方法,考虑了火点与其相邻像元之间的内在联系,在火点背景像元的确定及真实火点的判据选择等方面做了改进,在此基础上确定一组火点判据。该方法基本不受区域、时间等因素的限制,对面积较小的火点识别较为敏感,在实验中取得了较好的效果。

An Enhanced Contextual Fire Detection Algorithm Based on Remote Sensing Images

Most of the present satellite-based fire detection methods are based thresholds.The thresholds adopted in these methods are generally limited to a specific area,or certain seasonal weather conditions,therefore the actual effects are sometimes unsatisfactory especially in the detection of small-scale mountain forest fire.In response to these problems,an enhanced contextual algorithm for detection of forest fire(ECFDA) was introduced in this paper.Based on the relationships between the fire pixel and its background pixels,this algorithm made improvements on the confirmation of the neighbouring pixels of potential fire and established a criterion for the detection of fire.This method was adaptive to different areas and seasons and was sensitive to small-scale mountain forest fire.The results in our experiments have confirmed the potential of this method.