基于最小二乘配置理论的点面观测数据融合研究

随着对地观测技术发展，大地测量中有越来越多栅格形式的DEM面观测数据。与点观测数据相比，面观测精度偏低，一般需要融合点观测来提高精度。目前点面观测数据融合主要是根据点观测与面观测中对应点观测的差值，再采用插值回归类方法消除面观测的系统误差，没有考虑DEM系统误差在空间上的复杂变化及偶然误差在空间上的相关性，难以取得很好的融合效果；而最小二乘配置理论能同时处理观测对象的随机信号与倾向参数。本项目将把该理论引入到点面观测的融合中，建立顾及面观测误差特性的点面观测数据融合理论与方法。具体研究：⑴DEM面观测的误差特性及随机模型；⑵基于最小二乘配置理论的点面观测数据融合平差模型；⑶点面观测数据融合的平差模型解算与局域性系统误差的去除；⑷点面观测数据融合理论与方法在DEM、区域地表形变监测等典型领域中的应用。本项目的研究将平差理论从处理点观测数据拓展到处理面观测数据并解决DEM精度提升等实际问题。

With the development of earth observation technology, there are more and more area observations in geodesy, such as digital elevation models (DEM) represented by raster. Compared with point observations, the accuracy of area observations is low. In most cases, point-wise and area-wise observations need to be fused in order to improve the accuracy of DEM. The fusion of point-wise and area-wise observations is currently mainly based on the differences between point observations and those of the corresponding points in the area observations, and then a kind of interpolation and regression methods is applied to eliminate systematic errors of area observation. These types of methods do not take into account complex spatial variations of systematic errors and the spatial correlation of random errors for DEM. Therefore, it is difficult to achieve a good fusion effect. It is well known that least squares (LS) collocation can be used to process simultaneously random signals and deterministic trends of the observed objects. The purpose of this project is to introduce LS collocation to fuse point-wise and area-wise observations and establish an adjustment model to process simultaneously point-wise and area-wise observations. More specifically, this research project will focus on the following aspects: 1. To study error characteristics and random models of DEM area observations, 2. To construct trend and stochastic models and apply LS collocation to fuse point-wise and area-wise observations, 3. To apply variance component estimation to determine optimal weightings for point-wise and area-wise observations and to achieve the optimal removal of local systematic errors, and 4. The proposed method of data fusion will be applied to typical applications, such DEM and regional area deformation monitoring. The research in this project can also be said to extend the adjustment theory from processing point observations to area observations for problems of practical importance, such as accuracy improvement for DEM.