论文部分内容阅读
目的基于中高分辨率影像进行大范围的农村建筑区提取时,由于影像分辨率的限制以及农村建筑区规划自身规划特点等因素,造成了传统变差函数方法的高错分误差。为了准确提取农村建筑区,为后续获取建筑密度和人口密度等工作建立基础,提出了一种基于迭代P参数法的阈值确定方法。方法通过设定亮度阈值,在变差函数纹理计算中为满足条件的像元点赋以权值。本文方法确保在4个方向都满足条件的像元点(认为是建筑区)获得较大的变差函数值加成,而仅在一个方向或者没有方向满足条件的像元点(认为是非建筑区)获得较小加成或不变,以此改进传统变差函数方法,抑制了农村建筑区与周边非建筑区的混淆。结果以Radarsat-2的多个极化波段影像为数据源进行了实验,改进变差函数方法在实验区1与实验区2的各个波段平均检测率分别为91.58%和90.11%,平均错分误差分别为19.83%和31.87%。结论与传统变差函数方法以及最小距离法相比,既保证了较高的检测率,同时显著降低了错分误差,不足之处是在建筑区与非建筑区的边缘处以及与建筑区具有相似纹理特征的非建筑区处出现错分,需要进一步的研究和完善。
The purpose of this paper is to extract the high-resolution image of a large area of rural construction area. Due to the limitations of image resolution and the planning features of the planning of rural building area, high error-error of the traditional variogram method is caused. In order to accurately extract rural construction area and establish the foundation for subsequent acquisition of building density and population density, a threshold determination method based on iterative P-parameter method is proposed. The method assigns weights to pixel points satisfying the condition in the variogram texture calculation by setting the brightness threshold. The proposed method ensures that pixel points that satisfy the condition in all four directions (considered as building areas) gain larger values of variogram additions and that pixel points that satisfy the condition in only one direction or no direction (considered as non-building areas ) To obtain smaller additions or invariances, thus improving the traditional variogram method and restraining the confusion between the rural construction area and the surrounding non-construction area. Results Radarsat-2 images from multiple polarimetric bands were used as data sources. The improved variogram method was 91.58% and 90.11% respectively in each band of experimental zone 1 and experimental zone 2, and the average misclassification error Respectively 19.83% and 31.87%. Conclusion Compared with the traditional method of variogram and the method of minimum distance, it not only guarantees a high detection rate but also significantly reduces the misclassification error, which is similar to the building area and the edge of the building area The misclassification of texture features at non-building areas requires further research and refinement.