光学影像纹理信息在林业领域的最新应用进展

The latest applications of optical image texture in forestry

随着光学卫星影像空间分辨率的不断提高,影像纹理特征的重要性日益凸显。然而,纹理是一个非常复杂的空间属性,会随着太阳/观测角度、地形、感兴趣目标及其所处环境的不同而发生显著的变化。此外,不同纹理变量的选择及相应输入参数的设置,如窗口大小、像元间距、方向以及量化等级等都可能在一定程度上决定着影像纹理的利用价值。如何有效利用纹理量及其优化组合是一个值得深入探讨的问题。鉴于此,本文首先全面回顾了影像纹理特征在森林分类、森林结构参数反演以及森林生物量与碳储量的遥感估算等方面的最新研究与应用,并从不同角度肯定了光学影像纹理在林业遥感领域的应用潜力。此外,从纹理变量及其4大输入参数的选择和最优变量组合的判别方面,总结并剖析了当前研究领域中所存在的关键问题,权衡利弊并给出了相应的建议,为相关研究人员将影像纹理信息更有效地应用于林业领域提供参考。

Image texture is becoming more and more important with the increasing spatial resolution of optical satellite images. However, it is a very complex spatial attribute that can vary significantly with solar/ viewing geometries, topographic conditions, and the object of interest as well as its location. Besides, the selection of texture variables and the set of their corresponding input parameters, for instance, window sizes, inter-pixel distances, directions and quantization levels, determine the efficiency of image texture. How to apply texture variables and optimize their combination is worth further studying. Therefore, we firstly reviewed the latest researches and applications of image texture in forest classification, inversion of stand structure parameters, and estimation of forest biomass and carbon storage, and prospected the potential of image texture in the remote sensing of forestry from different aspects. In addition, we summarized the critical problems in the current research field based on the selection and optimal combination of texture variables and their input parameters. Some suggestions have also been proposed for further effective applications of image texture in the field of forestry.