【摘要】 针对航天遥感影像数据量大，存在重复和低质量影像，导致重复处理和存储，进而浪费计算、存储资源的问题，提出一种改进的航天遥感影像数据清洗方法。采用一种改进的DBSCAN聚类算法，基于遥感影像编目数据，对相似重复影像进行检测，形成相似重复记录集；对上一步确定的潜在相似重复影像执行影像匹配，根据匹配结果确判相似重复影像；最后，对相似重复影像进行质量评价，根据评价结果确定清洗策略，依据策略对数据进行清洗，剔除重复和低质量影像。可对海量原始遥感影像数据中重复、低质量数据进行准确、快速、有效清洗。实验结果表明，可有效识别相似重复遥感影像，并对影像进行准确分级，最终去除重复和低质量影像。更多还原

【Abstract】 Aiming at the problem of the large amount of space remote sensing image data, the existence of a large number of repeated and low-quality images, which leads to repeated processing and storage, thereby causing waste of computing and storage resources, an improved space remote sensing image data cleaning method is proposed. This method firstly utilizes an improved DBSCAN clustering algorithm to detect similar duplication images based on remote sensing image cataloguing data to form a similar duplication record set. Image matching is then performed on the potentially similar duplication images identified in the previous step, according to the matching results, the similar duplication images are determined. Finally, the quality of similar duplication images is evaluated to determine the final cleaning strategy, and the data can be cleaned and the duplication and low-quality images are eliminated according to the strategy. Through the above processes, the duplication and low-quality data in the massive original remote sensing image data could be accurately, rapidly and effectively cleaned.The experimental results show that the proposed method can effectively identify similar duplication remote sensing images, and can accurately grade the images, and finally remove the duplication and low-quality images.更多还原