基于多尺度图像库的遥感影像分割参数优选方法

Selection of best-fitting scale parameters in image segmentation based on multiscale segmentation image database

寻找合适的分割参数是面向对象影像分析的首要步骤。试错策略的目视分析法在实际应用中被广泛采用,但其无法分析大量分割结果,难以找到最佳分割参数。针对此问题,提出一种基于多尺度图像库的分割参数优选方法,并以SPOT5影像为例开展了多尺度分割参数优选实验。结果表明,该方法能够有效找到合适的分割参数,并能对地物目标的多尺度特性做全面分析。基于多尺度图像库的分割参数优选方法简单有效,可为工程化影像解译中分割参数选择提供支持。

Finding the best - fitting parameters in image segmentation is of great importance for object - oriented information extraction. The try -and -error strategy and visual analysis on multiresolution segmentation are widely used in real practice, but they cannot analyze a large number of segmentation results. This paper proposes a procedure for picking segmentation parameters based on muhiresolution segmentation image database and visual analysis. The experiment of multiresolution segmentation on SPOT5 image shows that the proposed procedure is capable of and - error finding the best fitting parameters. The procedure is more efficient and effective than the traditional try - strategy, and there is good potential for the procedure to be used in practical image analysis application.