一种新的准确提取离散采样点光谱值的方法

在建立遥感信息反演模型中，是利用离散的采样点实测数据与相应的影像对应像元的光谱值建立关系，从而实现目标信息反演。准确提取光谱值是建立模型的关键，提取光谱值通常采用的方法是把目标点图层转化为感兴趣区ROI，然后保存ROI为ASCII。用ENVI软件依据原始坐标提取采样点光谱值，分析提取的坐标和光谱值发现，提取出的部分采样点的坐标和原始坐标不一致，即光谱值非本像元，以致基于此而建立的反演模型不能真实反映目标属性与光谱值的关系，即模型无意义。我们把像元均分为4个区域，总结规律发现只有采样点落在像元左上角区域时，提取的光谱值为本像元光谱值。在上述方法的基础上，深入研究其提取目标点坐标和光谱值的原理，并总结规律。介绍了实现提取采样点在遥感影像上对应像元光谱值的一种新方法，该方法首先是提取采样点所在像元的4个顶点坐标之一，通过对比原始坐标和提取坐标的经纬度差值判断采样点在像元中分布所属区域，采用对称原则进行调整采样点在像元中分布位置，最终使全部采样点均分布在所属像元的左上角区域，最后再次进行光谱值提取，经验证提取的光谱值准确无误。通过OLI，TM和ETM+影像验证，结果表明该方法能够有效准确提取离散点光谱值，同时原理清晰，操作简单可行，适用性较强，为遥感影像提取离散点光谱值提供了新思路。

A New Method of Accurately Extracting Spectral Values for Discrete Sampling Points

In the establishment of remote sensing information inversion model, the actual measured data of discrete sampling points and the corresponding spectrum data to pixels of remote sensing image, are used to establish the relation, thus to realize the goal of information retrieval. Accurate extraction of spectrum value is very important to establish the remote sensing inversion mode. Converting target spot layer to ROI (region of interest) and then saving the ROI as ASCII is one of the methods that researchers often used to extract the spectral values. Analyzing the coordinate and spectrum values extracted using original coordinate in ENVI, we found that the extracted and original coordinate were not inconsistent and part of spectrum values not belong to the pixel containing the sampling point. The inversion model based on the above information cannot really reflect relationship between the target properties and spectral values; so that the model is meaningless. We equally divided the pixel into four parts and summed up the law. It was found that only when the sampling points distributed in the upper left corner of pixels，the extracted values were correct. On the basis of the above methods, this paper systematically studied the principle of extraction target coordinate and spectral values, and summarized the rule. A new method for extracting spectral parameters of the pixel that sampling point located in the environment of ENVI software. Firstly, pixel sampling point coordinates for any of the four corner points were extracted by the sample points with original coordinate in ENVI. Secondly, the sampling points were judged in which partition of pixel by comparing the absolute values of difference longitude and latitude of the original and extraction coordinates. Lastly, all points were adjusted to the upper left corner of pixels by symmetry principle and spectrum values were extracted by the same way in the first step. The results indicated that the extracted spectrum values of all points were accurate. Experiment on OLI (Operational Land Imager), TM and ETM+ images showed that this method can accurately extract the discrete spectrum value, and as well, clear principle, simple and feasible operation, strong applicability, This paper provides a new idea for remote sensing image extraction of discrete point spectrum.