利用遥感监测亚像元分解遗传算法估算森林火灾面积

牛顿迭代法是林火亚像元分解的常用算法,但其火区面积计算精度取决于背景温度的估算精度,遗传算法具有全局寻优能力,本文尝试以火区面积百分比Af、火区温度Tf、背景温度Tbg为优化变量,采用RAGA遗传算法对林火亚像元辐射率平衡方程全局寻优求解。结果表明：用RAGA遗传算法比牛顿迭代法计算的结果更符合实际,迭代误差低2个数量级,用此方法对2005年四川省木里县三处森林火灾某时刻燃烧区面积进行估算,其结果对火情通报和火灾扑救工作具有重要的参考价值。

Estimate of Forest Fire Area by Using RAGA Genetic Algorithms in Sub-pixel Decompound Based on Remote Sensing Monitoring

Newton＇s iteration method was the common algorithms on the sub-pixel decompound of the forest fire. The computing precision of the fire area lied on the estimating precision of background temperature, when the estimating error of the background temperature （Tbs） was to 0. 5 1K, one times difference of the fire area to pixel size percent （Af） would caused. The genetic algorithms has the capability of entirely optimizing, with the background temperature （Tbg） , fire temperature （Tf） and fire area percent （Af） as the optimizing variables. The RAGA genetic algorithms were used to optimize the entire results of radiance balance equations on the sub-pixel deeompound. The result indicated that the equations evaluation precision was two orders of magnitude higher than the Newtong iteration method. Finally, three forest fire areas in Mull County of Sichuan Province was estimated with the RAGA resoectivelv to orovide the imoortant references for outting out of the forest fires.