摘要
为解决矢量数据与遥感影像的变化检测难题,提出了一种融合光谱异质度与纹理异质度的变化检测方法。以矢量数据为约束,对遥感影像进行分割获取像斑;提取像斑的光谱直方图与局部二值模式(LBP)纹理直方图;采用G统计量度量直方图的距离,利用像斑与其他同类像斑特征距离的平均值分别构建像斑的光谱异质度与纹理异质度,加权组合光谱异质度与纹理异质度构建像斑异质度;利用大津法获取各地物类别的异质度阈值,比较像斑的异质度与对应类别的异质度阈值,对像斑进行变化/未变化判别。在IKONOS遥感影像上的试验验证了本文方法的有效性。
In order to solve the change detection problem using vector data and remote sensing imagery,a change detection method by integrating spectral and textural heterogeneity was proposed in the paper. Image segmentation under the constraint of vector data was employed to get image objects.Spectral histogram and lbptextural histogram of the object were extracted.G statistic was used to measure the distance between two histograms.The spectral and textural heterogeneity were built by the average distance between the object and other objects with the same class respectively. The object heterogeneity was calculated by the weighted combination of spectral heterogeneity and textural heterogeneity. The changed/unchanged label of the object can be decided by comparison the object heterogeneity with heterogeneity threshold of correspond class which can be calculated by otsu method. The experimental result on IKONOS image shows the effectiveness of the proposed method.
出处
《测绘通报》
CSCD
北大核心
2017年第S2期28-33,共6页
Bulletin of Surveying and Mapping
基金
测绘地理信息公益性行业科研专项(201512026)
四川省测绘地理信息局2017年科技支撑计划(J2017ZC06)
