Impact of point spread function of MSG-SEVIRI on active fire detection
Impact of point spread function of MSG-SEVIRI on active fire detection
复制标题
MSG-SEVIRI 点扩散函数对主动火灾探测的影响
DOI：
10.1080/01431160802609726
复制
发表时间：
2009
影响因子：
3.4
通讯作者：
F. Gonzalez
中科院分区：
文献类型：
--
作者：
Abel Calle;J. Casanova;F. Gonzalez
传感器的空间分辨率是一个经常以不恰当的方式描述的概念，通常通过图像捕获过程中的采样距离来识别。调制传递函数 (MTF) 的形状对基于辐射均匀分布的应用的结果没有影响。然而，在高温事件（HTE）的情况下，为了量化辐射亮度，像素内部燃烧区域的空间位置是一个需要解决的关键问题。在面向火灾探测的算法和双谱过程的应用中都应考虑点扩散函数（PSF）。本文分析了气象卫星第二代旋转增强型可见光和红外成像仪 (MSG-SEVIRI) 传感器的 PSF 对热火灾参数测定的影响。分析了中红外 (MIR) 3.9 μm 光谱带中 PSF 对亮温 (BT) 的影响以及检测算法。还分析了由于所涉及的 PSF 不重合而导致的双谱技术检索的火灾温度误差。获得的结果显示 3.9 μm 光谱带中的 BT 存在约 20 K 的差异，具体取决于像素内的火灾位置。最后，对燃烧面积最小尺寸的检测概率进行了分析，发现燃烧面积为 10 公顷的火灾检测概率为 90%，而检测面积为 4 公顷的火灾概率为 50%。
The spatial resolution of sensors is a concept frequently described in an inappropriate way, usually identified by the sampling distance in the image capturing process. The shape of the modulation transfer function (MTF) has no influence on the results in applications based on homogeneous distribution of radiance. However, in the case of high-temperature events (HTEs), the spatial location of the burning area inside the pixel is a key issue to solve, in order to quantify the radiance. The point spread function (PSF) should be considered both in fire detection-oriented algorithms and in the application of bispectral processes. This paper analyses the impact of the PSF of the Meteosat Second Generation Spinning Enhanced Visible and Infrared Imager (MSG-SEVIRI) sensor on the determination of thermal fire parameters. The PSF influence on the brightness temperature (BT), in the mid-infrared (MIR) 3.9 μm spectral band, on detection algorithms is analysed. Errors in the fire temperature retrieved by the bispectral technique, due to non-coincidence in the PSF involved, are also analysed. The results obtained show a difference of around 20 K in the BT in the 3.9 μm spectral band, depending on the fire location inside the pixel. Finally, the probability detection of the minimal size of the burning area was analysed, and revealed that there is a 90% probability of detecting a fire with a burning area of 10 ha whereas an area of 4 ha is detected with a probability of 50%.