SAR影像古遗址自动检测方法研究

Remote sensing application in archaeological investigations has been an emerging interdisciplinary with great potential, which has drawn widespread attentions of both domestic and international scholars. As one of the most advanced remote sensing techniques, SAR is an indispensable scientific archaeological method and has accounted for a number of important archaeological discoveries, thanks to its unique observation advantages. However, the vast majority of the applications of SAR in archaeology have been undertaken within a visual interpretational framework. With operations in a low automatic level, this traditional approach to detect historical sites over extensive areas, with huge and complex SAR datasets, is becoming to be more problematic. Thus, in this project, the automatic detection method of archaeological sites in SAR images will be studied. The research will choose the region of the Juyan historical sites as the study area, which is located in Ejin Banner of the Alxa League, Inner Mongolia, China. The detection scheme will consider a two-stage processing flow: prescreening and discrimination. The former will be largely based on the most practical detection method-Constant False Alarm Rate (CFAR) method. The latter should consist of an effective selection of historical site features and a proper design of discrimination rules. In addition, the distribution data of the historical sites in the study area will be constructed, via visual inspection and fieldwork, to guide the development and to test the performance of the proposed detection method. The primary goal of the project is to develop an effective automatic detection method of historical sites in SAR images that can be technically contributing to the rapid inventories of historical sites across large areas. This project will boost SAR applications in archaeology, and further advance and deepen the research on the application of remote sensing in archaeology.

遥感考古是一门极具发展潜力的新兴交叉学科，得到了国内外学者的广泛重视。SAR作为最重要的前沿遥感技术之一，凭借其独特优势已取得了多项令人瞩目的考古成果，是重要的科技考古手段。然而，目前SAR遥感考古研究仍以人工判读为主，自动化程度低，不利于大数据下大区域范围内古遗址的快速有效探测。本研究以内蒙古阿拉善盟额济纳旗境内的居延遗址群为研究区, 以实用性较强的恒虚警（CFAR）检测为基本手段，采取二级递进的（预检测+鉴别/辨识）检测处理流程，通过选取有效的古遗址辨识特征和建立合理的辨识规则，进行SAR影像古遗址的自动检测研究。通过人工判读和田野调查的方式构建研究区内的古遗址分布数据，以指导自动检测方法的研究和评价检测结果。项目整体目标是，发展一种有效的SAR影像古遗址自动检测方法，为大区域范围内古遗址的快速编目提供技术方案。该研究将会促进SAR在考古中的应用，推动遥感考古研究的进一步发展与深入。

古遗址是人类历史信息的重要载体，是不可替代的宝贵资源。然而，由于战争、自然环境以及人类活动等的影响，大量古遗址在全球范围内正面临着加速破坏、甚至消失，提前探测和加快保护这些古遗址的任务变得愈发紧迫。本项目基于合成孔径雷达（SAR）这一前沿对地观测手段，以内蒙古阿拉善盟额济纳旗境内的居延遗址群为研究区，开展了古遗址的自动检测研究。利用多频段、多极化、多角度高分辨率SAR数据，并结合高分辨率可见光遥感数据、无人机数据、地基三维激光扫描仪数据，以及考古资料和田野调查数据，项目开展了以下主要研究内容：1）考古资料收集与居延遗址空间分布制图；2）SAR 影像斑点噪声抑制与图像增强分析；3）多频段、多极化、多角度SAR影像古遗址探测能力评估和分析；4）古遗址SAR影像形态特征分析；5）SAR 影像统计建模分析；6）SAR影像古城址自动检测算法设计。 通过上述研究,项目取得了如下成果：1）初步完成了居延遗址空间分布制图，将为国内外居延考古研究提供关键数据和空间框架；2）对比分析了当前广泛使用的多种斑噪抑制与图像增强算法，发现计算简单、便于操作的均值滤波器更适用于戈壁古遗址目标的SAR影像探测/检测； 3）较为全面的阐明了雷达频段、极化、角度参数对戈壁地区古遗址探测能力的影响；4）明确了古遗址在多频段、多角度SAR影像上的形态特征和可识别性；5）确定了对数-正态分布对居延遗址区内的自然地物杂波在不同频段、不同角度的SAR影像上均具有较强的建模能力；6）发展了星载SAR影像古城址自动检测算法；7）给出了星载SAR影像古城址自动检测方案。本项目的研究结果，进一步完善了SAR在考古中的应用，为大区域范围内古城址的自动检测提供了一种手段。