改进Otsu与形态学相结合的水体信息提取

Water Information Extraction Based on Improved Otsu and Morphology

水体提取对于洪涝灾害评估有着重要的意义。针对传统的Otsu分割阈值会偏向方差大的一类,因此分割结果包含大量的伪水体,给后续处理增加了极大困难的问题,提出目标权重改进的Otsu和结合高程连通标定的形态学优化方法。通过实验证明,基于目标权重改进的Otsu可以有效地平衡分割阈值,计算的分割阈值更接近真实分割阈值。利用改进Otsu提取水体的初始分布图,并采用结合高程连通标定的形态学优化方法对其进行优化,得到最终的水体分布图。对实验结果分析得出,该水体提取方法可以充分发挥改进Otsu和结合高程连通标定的形态学优化方法在提取水体方面的各自优势,大大提高了水体提取的精度。利用该方法对覆盖彭泽县的影像数据进行水体提取,验证了该方法提取水体的可靠性。

Water body extraction is of critical significance for flood disaster assessment. The threshold segmentation results based on the traditional Otsu method are usually biased towards the classes with larger variance,and the segmentation results contain a large number of pseudo water bodies,which increases great difficulty to the subsequent processing. In this paper,an improved Otsu method based on target weights and a morphological optimization method combined with elevation connectivity calibration are adopted as a combined method. The results show that the improved Otsu based on the target weight can effectively balance the segmentation threshold,making the calculated segmentation threshold closer to the real value. First,the improved Otsu is used to extract the initial distribution of the water body,and then the morphological optimization method of elevation connectivity calibration is used to optimize it,and the final water body distribution map is obtained. The experimental results show that based on the combined method proposed in this paper,the advantages of the improved Otsu and the morphological optimization method combined with elevation connectivity calibration are fully utilized,and the accuracy of water body extraction has been greatly improved. The proposed method was applied to water body extraction in the image of Pengze county,which verified the reliability of it in water body extraction.