基于遥感数据提取建筑物的主动轮廓改进模型

An improved active contour model to extract buildings based on remotely sensed data

针对传统的主动轮廓模型在提取建筑轮廓时不能区分与其具有相似反射率的地物和需要建筑物附近的初始轮廓等问题,从两方面对其进行了改进:1)将原始的单一数据改为利用高分辨率遥感影像与激光雷达(LiDAR)数据融合后的数据进行建筑物的提取;2)在原始模型的能量公式中加入比例系数,控制各波段在能量泛函中的比重.采用Matlab编程实现了所提出的算法,并对徐州市两个地区的快鸟(Quickbird)影像进行了分析.结果表明:改进后的模型可以很好的完成建筑物轮廓的自动提取,并且具有对噪声不敏感、不需要建筑物附近的初始轮廓和隐式改变拓扑结构的优点,达到了较好的效果,证明了改进主动轮廓模型的可行性.所提出的算法为建筑物的轮廓提取提供了有效手段.

Considering the problems that the buildings can＇t be distinguished from the objects with the same reflectivity and the initial contours nearby buildings are needed in the traditional- ly active contour model, two different methods are provided to improve the traditionally active contour model in the paper： 1） the high resolution remotely sensed images and LiDAR data are fused to replace the traditionally single data in the improved model; 2） scale coefficients are added into the initial model to control the proportion of each band in the energy functional. The improved model was implemented by Matlab coding and two experiments were performed to e- valuate the performance of the improved model using Quickbird images of two different areas in Xuzhou. Experimental results indicate that the improved model can automatically detect the buildings＇ contours with the advantages of low sensitivity to noise, no initial contours nearby buildings and changing topologies implicitly, and achieve good results, hence providing effec- tive methods for extracting buildings＇ contours.