一种超像素区域相似性度量的遥感信息提取算法

发布时间：2018-05-13 15:12

本文选题：图像处理 + 遥感信息提取　；参考：《激光与光电子学进展》2017年08期

【摘要】：为了解决目前主流的显著性检测算法在复杂多目标遥感图像中检测能力不足的问题,提出一种基于超像素区域相似性度量的显著目标提取算法。该算法利用简单线性迭代聚类方法对原始图像进行超像素分割,通过基于图论的视觉显著性方法检测出显著超像素,并对其修正得到显著目标提取的训练样本,进一步逐层计算全体超像素区域与显著超像素区域的相似性并转化为超像素区域的隶属度值,最后实现对整幅超像素图像的显著目标提取。实验结果表明,该算法具有较高的准确率和召回率,能更加有效地检测出遥感图像中的显著目标,提取效果优于主流的显著性检测算法,还可以有效应用于复杂多目标的遥感图像显著目标信息提取中。
[Abstract]:In order to solve the problem of lack of detection ability in complex multi-target remote sensing images, a significant target extraction algorithm based on the similarity measurement of super pixels is proposed. This algorithm uses simple linear iterative clustering method to segment the original image with super pixel, and through the graph based visual display. The remarkable super pixel is detected by the novel method, and the training sample which is extracted from the significant target is obtained. The similarity between the super pixel area and the significant super pixel region is calculated by the layer by layer, and the membership degree of the super pixel region is transformed. Finally, the extraction of the significant target for the entire amplitude super pixel image is realized. The experimental results show that the algorithm is used. With high accuracy and recall, the significant target in remote sensing image can be detected more effectively. The extraction effect is better than the mainstream detection algorithm, and it can be effectively applied to the extraction of the significant target information of the complex multi-target remote sensing images.

【作者单位】：中国科学院遥感与数字地球研究所数字地球重点实验室;中国科学院大学;
【基金】：国家863计划(2015AA7026087) 国家自然科学基金(41501489) 国家科技支撑计划(2015BAB05B05-02) 中国地质调查局项目(12120113089200) 中国科学院遥感与数字地球研究所所长青年基金(Y6SJ1100CX) 高分辨率对地观测系统重大专项(民用部分)(30-Y20A37-9003-15/17)
【分类号】：TP391.41

【相似文献】