贝叶斯-马尔可夫随机场（Bayes-MRF）算法由于其在解决相位解缠方面的优异性能，因此备受关注。然而现有算法单纯依靠相位信息，受城市复杂地形以及噪声影响，导致解缠精度降低，生成的城市数字地表模型（DSM）无法满足实际应用。针对上述问题，本课题将在现有算法中引入城市地物类别分析，重点研究：（1）将相位解缠问题转化为基于相位周期和城市地物类别的像素标记问题，并构建关于该标记问题的最大后验-马尔可夫随机场（MAP-MRF）框架；（2）建立城市地物类别标记的能量函数，有效避免城市地物过分割或者不完全分割现象；（3）建立相位周期标记的能量函数，可以兼顾同类城市地物内部相位光滑特征和不同城市地物之间相位突变特征；（4）设计结合两种标记变换的最佳能量优化算法，有效解决能量函数欠收敛问题。通过本课题的研究，将对提高城市复杂地形及高噪声区域的解缠精度，更好地反演城市DSM，具有重要的理论意义和应用价值。

Bayes-MRF algorithm is widespread applied owning to its excellent performance in solving phase unwrapping problem. However, the current algorithm relies solely on the phase information, which leads to a reduction of unwrapping accuracy due to the influence of complex terrain and high noise in urban district. Aiming at this problem, this research will introduce urban land cover classification into the algorithm. The following topics are involved in the research: (1) transforming the phase unwrapping problem into the pixel labeling problem based on phase cycle and urban land cover, and building the MAP-MRF framework to solve this pixel labeling problem; (2) establishing the energy function based on urban land cover to avoid the over segmentation or imperfect segmentation phenomenon of urban objects effectively; (3) establishing the energy function based on phase cycle to take account of both the smooth characteristics inside the same urban objects and the discontinuous characteristics at urban object boundaries; (4) designing the best energy optimization algorithm to exchange the two types of labels, which can solve the energy function problem under convergence effectively. This research can systematically improve the precision of phase unwrapping in urban district with complex terrain and high noise and can acquire urban DSM more profitably.