基于特征线段分析的建筑物面识别方法

将建筑物划分为单个的建筑物面进行处理可以有效降低街景图像理解的复杂度. 因此, 文中提出一种基于建筑物特征线段累积的动态规划方法用于检测街景图像中各个建筑物面的信息. 该方法结合系统工程分析问题的方法, 根据图像建筑物区域直线特征丰富的特点, 分析不同朝向类型的建筑物面中水平直线段透视投影后的变化, 建立相应的特征累积数学模型量化建筑物区域蕴含的直线段特征, 并给出相应的算法优化平滑特征线段累积结果, 最后证明了根据特征线段累积结果识别建筑物面的过程符合动态规划最优性原理, 通过动态规划的方法求解识别目标区域内所包含的各类建筑物面. 实验证明, 该文的方法可以准确的识别街景图像中各类建筑物面, 而且算法复杂度更低, 处理速度更快.

Building facet recognition by analyzing feature line segment

It would be more easily to process the image of street scene by recognizing every building facet first. In this paper, a novelty method which recognizes every building facet by analyzing feature line segment of buildings with technique of system engineering is proposed. Beginning, the feature line segments extracted from the area of buildings are accumulated with the mathematics' model which deduced by the relation of horizontal lines between the real world and image. Then the result of feature accumulation is refined by a new method. At last, each building facet could be recognized by dynamic programming because the building facet recognition by analyzing feature line segments has been proved to comply with the optimality principle. The experiment shows that our method could recognize building facet exactly in many complex environments and need less time in the fields of building facet recognition.