火烧迹地信息遥感提取研究进展与展望

Research Progress and Prospect of Remote Sensing on Extracting Burned Areas Information

火烧迹地是描述森林燃烧最重要的信息之一,也是碳循环扰动和全球变化研究的重要参数。重点对近年来国内外广泛使用的火烧迹地产品的解译标志和模型算法进行了归纳总结,以往研究主要侧重于火烧迹地提取模型算法开发,忽视了复杂地形下阴影、纹理等细节特征的作用,限制了人们对火烧迹地的准确提取和科学评估。针对该问题提出了对火烧迹地信息提取改进途径的展望,基于模拟真实地表的DEM参数,优化地形校正效果,建立阴影校正植被指数;明确灰度纹理产生的条件,尝试建立颜色空间和灰度纹理映射方案,实现颜色纹理的定量解译;耦合阴影校正植被指数与颜色纹理,减少对小面积火点的遗漏和混淆信息误判,基于高分卫星数据实现长时间序列高精度火烧迹地信息提取,对探讨全球火烧迹地产品研发具有典型代表意义。

The emerging evidence as regards the issue of the increasing degree of harm that forest fires cause to the environment, economy and society has garnered significant attention from both the national and local governments, especially considering China’s rapid economic growth, urban economic development, continuous warm and dry climate and combustible accumulation caused by logging ban. Undoubtedly, burned area information is one of the most important information used for describing forest fires and is an important parameter for the study of carbon cycle disturbance and global change. by summarizing the interpretation marks and model algorithms of the products which are still widely used at home and abroad in recent years, previous studies have focused on developing models for extracting burned area algorithms rather than emphasising the effects of detailed features such as shadows and textures in complex terrain, which has restrained accurate extraction and scientific evaluation of burned area information. Therefore,In view of this problem, we put forward the prospect of improving the way to extract the burned area information, the effect of a terrain correction model is first optimised based on digital elevation model parameters for simulating the earth’s surface by spectral vegetation test in sunlit and shadow slopes. Meanwhile, vegetation indices are established upon shading correction. Then, by defining the conditions for generating grey texture using the colour co-occurrence matrix and Euclidean space, quantitative interpretation of colour textures is achieved. Finally, the random forest model is optimised. By expounding the influence of coupling of vegetation indices upon shading and colour texture correction on extracted burned area, decrease in missing small-area fire points and erroneous judgement of mixed information to high-precision extraction of burned area information can be achieved. It has typical representative significance in exploring the research and development of burned area products worldwide.