大钝角剖分与最小二乘法约束迭代算法优化TIN模型研究

TIN模型通过从不规则分布的数据点生成的连续三角面来逼近地形表面，就表达地形信息的角度而言，TIN模型的优点是它能以不同层次的分辨率来描述地形表面。但在相关软件对离散数据生成TIN模型时会出现边缘失真现象，特别是当一些离散数据出现凹区域时，凹区域地形逼近失真现象相当明显。而在这些凹区域大多生成的三角网以钝角三角形居多，本文对离散数据TIN模型生成DEM时存在凹区域地形逼近失真现象进行了研究。提出一种大钝角剖分与最小二乘法约束迭代算法优化TIN模型。首先对原始TIN模型中大于100°小于1800的钝角三角形进行遍历记录，然后利用最小二乘法约束迭代算法进行优化，通过此算法优化后的TIN模型生成DEM数据，分别从小区域面积稳定性与剖面拟合实际地形进行了试验与精度对比，试验显示这种算法是可行的。

Study about Large Obtuse Angle Dividing and Least- Squares Constraint Iteration Algorithm Optimization TIN Model

TIN model by irregular distribution of data points that produce Continuous triangle surface are generated to approximate the terrain surface, . on the expression of terrain information point of view, the advantages of the TIN model is that it can be described by different resolution terrain surface. Presence of edge distortion will but the generated TIN model of discrete data in the related soft- ware, especially when some discrete data appears concave, concave terrain approximation distortion phenomenon is quite obvious, and triangulation are mostly in the concave region mostly obtuse triangle, the research on discrete data model to generate DEM in TIN con- cave terrain approximation distortion phenomenon was studied, the model of TIN triangulation for concave regions mostly appear in the edge and obtuse to automatically partition also rule generation acute triangle into actual terrain, presents a large obtuse subdivision and least square method constrained iterative algorithm to optimize TIN model algorithm. The basic idea of the algorithm is based on the o- riginal TIN model is more than 100 degree and less than obtuse triangle was defined as 180 degrees of the large obtuse angle, and iter- ate recorded by the computer program, and then using the least square method constrained iterative algorithms for optimization, through the TIN model to generate DEM data the algorithm is optimized, respectively from the area stability and profile fitting actual terrain were compared with precision, experiments show that this algorithm is feasible.