基于知识图谱的遥感图像检索定位方法

Remote Sensing Image Retrieval and Location Method Based on Knowledge Graph

为了快速、准确地在海量遥感图像库中检索出所需的遥感图像资源,提出了一种基于知识图谱的遥感图像检索定位方法。为了充分利用遥感图像中所蕴含的地理语义信息,方法中使用知识图谱节点代表遥感图像,根据图像间的相似程度和地理位置在节点间建立关系。通过知识图谱缩小图像检索空间,加快检索速度,准确地定位遥感图像所示区域的地理信息,并返回最相似的图像。将此方法应用于卫星自主导航领域中,与传统的线性检索相比,图像检索的时间缩短了50%,准确率也得到提高。

In order to quickly and accurately retrieve the required remote sensing image resources from the massive remote sensing image database,a remote sensing image retrieval and location method based on the knowledge graph is proposed.To make full use of the geographic semantic information contained in the remote sensing images,the nodes of the knowledge graph are used to represent the remote sensing images,and the relationship between the nodes is established according to the similarity degree and geographic locations of the images.The knowledge graph can reduce the image retrieval space,accelerate the retrieval speed,accurately locate the geographic information of the area shown in the remote sensing images,and return the most similar image.The method is applied to the satellite autonomous navigation field for verification.The results show that,compared with the traditional linear retrieval method,the image retrieval time is shortened by 50%,and the accuracy rate is improved.