A convenient archaeological ruins identification method through elevation information extraction from CORONA stereo pairs

Abstract

Three-dimensional (3-D) stereo images can be generated via computer-based image processing of CORONA stereo pairs. To a certain extent, important terrain and surface feature data extracted from these stereo images can improve the survey of archaeological sites and the identification and mapping of major landscapes. In this study, we focused on the identification of the archaeological ruins of Liangzhu City. An optical stereo model (red/blue stereo image) of the Liangzhu site was created through computer-based mosaicking and processing of CORONA remote-sensing stereo pairs taken in the 1960s and 1970s. By importing the optical stereo model into mobile phones, tablet computers, and other mobile devices, the research team undertook real-time locating of ruins via human observation, on-site investigation, and image overlay during a field survey and identified several Liangzhu-period dams, some of which have been confirmed via archaeological field investigations. The research team later applied the same method to the identification of tombs at the site of the mausoleums of the six emperors of the Southern Song dynasty. The results further prove that this method is feasible and reliable and can be widely promoted and used for the identification of archaeological ruins.

Introduction

The use of remote-sensing images for field surveys and exploration has become an important step in the preliminary stage of archaeological research [1,2,3]. The discovery and identification of archaeological ruins cannot be achieved without obtaining a balance between factors such as the timeline of landform evolution, data accuracy, easy access to data, efficient data processing, and convenience in interpreting remote-sensing images for archaeological field workers [4,5,6]. For archaeologists, the need is to combine traditional and scientific and technological archaeology to discover and study sites, and on this basis, to provide an important foundation for realizing the protection and inheritance of cultural heritage sites. To meet this demand, archaeologists need a platform or system that is intuitive, easy to operate, and effective.

In recent years, many advanced geospatial tools have been developed, but these tools are not only expensive but also require specialized training before use, which may be within the grasp of researchers but are not amenable to ordinary archaeologists [7]. With the development of global positioning systems (GPSs), ordinary users can pinpoint the locations of sites using a wide range of available navigation applications (such as Google Maps and WAZE) and professional geographic information system (GIS) software (such as ArcGIS and Map Plus), and ordinary mobile phone software provides increasingly more help in archaeological work [8,9,10,11]. Some of these applications support user-defined layers, which allow users to import their data for real-time locating.

Based on the status quo, in this study, we utilized remote sensing technology to develop a concrete operational platform that is easy to implement, and we propose a new research methodology that addresses the research needs of archaeologists. This platform addresses the need for archaeologists to consider the historical topography by using historical remote sensing images taken before large-scale urbanization took place and generating a 3-D model using historical remote sensing images to create a digital sand table model based on the past situation. In this study, our research team applied the well-established localization function to archaeological field surveys and exploration. By importing the required image layers into the mobile device in advance, we were able to observe the features of the archaeological sites in the study area in the field based on the topographical features observed in the images. It should be noted that this platform is easy to operate and portable, and it can be used not only for indoor research but also for field surveys, making field surveys and research simpler and more efficient.

In this platform, the micro-geomorphology is an important element in the identification of sites. Micro-geomorphology refers to the small undulations and morphological changes of the ground surface, which play an important role in the study of archaeological sites. The micro-geomorphology can not only reflect the formation process and historical background of a site but can also provide a basis for site investigation and protection. The wetland environment in the Jiangnan region of China is a typical area that reflects the role of micro-geomorphic features in the study of archaeological sites. The Liangzhu site is located in the Jiangnan wetland environment, and this site has been well studied, so it is an ideal example of archaeological research using micro-geomorphic features. However, the acquisition of micro-geomorphic information encounters the challenges of strong concealment, complicated data processing, and high difficulty in interpretation. Research on the Liangzhu site faces the same challenges. First, the Liangzhu area has a complex natural environment, which includes abundant lakes, rivers, and hills. Second, the Liangzhu culture was present thousands of years ago, and the micro-geomorphology of the region has changed significantly. In addition, based on the fact that most of the sites in the Liangzhu area were built on artificially constructed terraces, researchers believe that elevation is a key indicator for distinguishing the types and grades of Liangzhu ruins as it can better reflect the original distribution pattern of the ruins in this area and is more effective in identifying ruins compared to high-definition satellite images and digital line graphs [12]. Water management was an important social activity for agricultural settlements. The builders of many now ruined structures similar to those at the Liangzhu site acquired a good knowledge of local hydrological features before they started construction. Liu Jianguo built digital elevation models (DEMs) of the Qujialing, Shijiahe, and Chenghe sites and other major prehistoric settlements on the Jianghan Plain. By identifying and analyzing the performance of the settlements in terms of the hydrological environment, he was able to explain the structures and functions of these archaeological ruins [13]. Therefore, to some extent, elevation information can be used as an indicator of ruin patterns. However, as the DEM was created after 2000, it can reflect the topography of the modern environment and enable archaeological predictions to be made, but the changing surfaces of the plains and lowland and upland areas of the Yangtze River Delta, as well as the changes in the vegetation and buildings, have affected the accuracy of the identification of archaeological ruins to varying extents [14]. Furthermore, Galiatsatos et al. attempted to obtain a high-resolution digital surface model (DSM) from CORONA images and evaluated the accuracy of this DSM in comparison to the Shuttle Radar Topography Mission (SRTM) with a resolution of 90 m, the Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model (Aster GDEM) with a resolution of 15 m, and the ALOS World 3D (AW3D) model with a resolution of 5 m. The results demonstrated that the feature detectability of this DSM was comparable to that of AW3D [15]. Subsequent research by Watanabe demonstrated that even AW3D-type DSM approaches are unable to detect features smaller than approximately 20 m in size [16]. This represents a limitation in the application of DSM in archaeological site detection.

Considering the temporal and environmental characteristics of the Liangzhu site, researchers need to study this site by considering the micro-geomorphology of the historical period. This can be achieved via photogrammetry, and currently, remote sensing archaeological research can also be realized using stereo pair images from the historical period. At present, Keyhole (KH) satellites are the only source of historical remote-sensing images that have a high resolution and that can be acquired easily [17, 18]. Among them, the CORONA images (mostly KH-4B images with a ground resolution of 6 feet) were mainly taken in the 1960s, when the study area had not yet undergone massive urban expansion and intensive agricultural development and the landforms could still reflect the early surface features, thus providing an important basis for the identification of archaeological ruins [3, 19].

CORONA images can display an archaeological landscape that no longer exists [20]. In addition, CORONA data consist of stereo pairs, from which 3-D images of landscapes can be generated. Therefore, CORONA historical satellite images have been widely used in archaeology studies since they were declassified in the late 1990s [21]. These images have mainly been used in archaeological investigations in the Middle East [22,23,24,25,26,27,28,29,30,31] since archaeological sites dominate these alluvial landscapes, which have relatively heterogeneous land cover but less vegetation and thus are easily identified in panchromatic images with a high-resolution [32]. However, there are few other generally discussed cases in Europe [33]. From methodological perspectives, most of the utilization has concentrated on discovering sites that have disappeared through manual interpretation by archaeologists. Recent developments have taken advantage of the large coverage of CORONA data to create historical land cover databases [17, 34]. However, only a few previous studies utilized stereo pairs collected by CORONA. Casana et al. adopted ortho-rectification techniques developed by various scholars and developed an online platform called Sunspot. This platform is mainly for creating a global CORONA atlas [17], but it has also been proven to be useful for generating high-resolution DEMs via photogrammetry [23, 24].

Therefore, we selected stereo pairs of CORONA historical images and used remote sensing and GIS technology to determine the stereo effect of the archaeological site so as to obtain micro-geomorphic information about the area and carry out field investigation and analysis.

Research area and data

Named after the town of Liangzhu in the Yuhang district of Hangzhou, Zhejiang, where it was discovered, the Liangzhu culture dates back about 4300–5300 years. It represents the peak of Late Neolithic social and political development in China and is regarded as the culmination of the Neolithic Age in the Taihu Lake area in the lower reaches of the Yangtze River [35]. As a testimony to the 5000 year history of Chinese civilization, the archaeological ruins of Liangzhu City (referred to hereinafter as the Liangzhu site) were added to the United Nations Educational, Scientific and Cultural Organization (UNESCO) World Heritage List in 2019. The study area features a relatively visible and concentrated distribution of Liangzhu ruins, and most of the Liangzhu-period ruins in the Taihu Lake area are clay or stone terraces that rise several meters above the ground surface. The terraces vary in height, indicating their grade and size. Water bodies and rice fields are distributed in the lower regions around the terraces. This pattern has remained largely unchanged, and these terraces have been in use ever since they were built. Instead of leveling up the terraces and building new ones, people of later generations built upon the original terraces and carried out planting, processing, and other activities on the periphery of the terraces, thus preserving the early pattern of the archaeological ruins, which is helpful for the identification of archaeological ruins using remote-sensing images [36].

Some researchers have adopted remote sensing technology to conduct research on the Liangzhu site and have proven its effectiveness. In particular, historical CORONA remote-sensing images and DEM data have played important roles in assisting the search for the walls, inner city ruins, and outer regions of the ancient city at the Liangzhu site. GIS elevation models and CORONA images have also provided key clues for the search for the overall structure of the massive hydrological system outside the ancient city and for analysis of the structure, proving that the Liangzhu site has a palace/inner city/outer city structure and a large hydrological system on its periphery. Thus far, most of the remote-sensing images used for archaeological research on this site have been two-dimensional (2-D) images [16]. When using CORONA remote-sensing stereo pairs for archaeological ruin identification, well-known scholars such as Watanabe Nobuya successfully extracted some probable archaeological features and potential water flows. However, all the GCPs were collected from Google Earth, which constrained the precision of the registration. By utilizing CORONA stereo pairs and through the application of remote sensing technology already available for the Liangzhu area, it is easy to see that although remote sensing data, especially high-resolution categories such as CORONA, are mainly utilized to support field survey work, the entire survey procedure largely depends on indoor image investigation in addition to site identification. Therefore, all of these efforts are desktop approaches and are time-consuming. On-site efficiency does not seem to be a priority. Therefore, this paper provides an easy-to-operate and effective method for archaeological field investigations by processing CORONA stereo image pairs.

By comparing the images of the Liangzhu Ancient City in different periods (Fig. 1), it can be seen that the CORONA image, with clear layers and surface features, is as good as the one taken by modern commercial high-resolution satellites in terms of the details of houses and streets.

Fig. 1

Liangzhu Ancient City in different periods. The yellow circles represent the walls of Liangzhu Ancient City. a CORONA image photographed in 1968; b photographed in 1972; c photographed in 2010; d photographed in 2022

Methods and data processing

The use of remote-sensing images to identify and analyze archaeological ruins should be based on a thorough understanding of the specific needs of the archaeological research and a good knowledge of remote-sensing images that are widely used in the field of archaeology. Appropriate computer-based image processing techniques can then be used to make the most out of the remote sensing data and obtain in-depth research results. However, in previous archaeological research, CORONA images were often used as 2-D images for interpretation [37]. Furthermore, when stereo pairs were derived from CORONA images, they were often interpreted indoors using a conventional stereoscope. This method not only imposes strict requirements regarding observation sites and equipment but also requires experienced satellite image interpreters, which has to some extent increased the difficulty of archaeological information interpretation [38].

This paper (Fig. 2) addresses the aforementioned problems. First, we obtained remote sensing images of the study area, pieced them together, and applied computer processing, including image alignment and adjustment, mosaicking of multiple images, and image registration based on a base image. After checking the registration accuracy, we imported the stereo pairs into optical stereo modeling software to produce a red/blue stereo image. After ensuring that the stereo image had a satisfactory stereo effect, we prepared for the field survey and exploration of the archaeological ruins by importing the optical stereo model into mobile phones, tablet computers, and other mobile devices and entering its coordinates.

Fig. 2

Research framework and technical route

Acquisition of stereo pairs

Elevation changes in the Earth’s surface are more visible in CORONA stereo pair images than in single images, making it easier to identify surface features. However, the extraction of DEMs from CORONA stereo pairs using photogrammetric methods still has many limitations and is rarely used in practice because of its low efficiency [39]. It is much simpler if a stereoscope is used to observe the CORONA images. Moreover, this observation is very intuitive. Therefore, the first step is to obtain stereo pairs of the study area.

The stereo pairs of the study area were mainly acquired from the website of the United States Geological Survey (USGS). As the study area is located exactly at the junction of two serial images, it was necessary to obtain two front-view images (DS1106-2086DA078 and DS1106-2086DA079) and two rear-view images (DS1106-2086DF071 and DS1106-2086DF072).

Three-dimensional image adjustment

The acquired stereo pairs cannot be directly used to build an optical stereo model. Because of factors such as different lighting conditions, there may be obvious differences in the hue and shade of images of different areas, which affect the image interpretation and result in patches and seams in the composite image, or in severe cases make the composite image obsolete. As CORONA images are high-resolution panchromatic images, the gray-level information about the surface features is important for computer-based image interpretation [37, 40]. Therefore, CORONA images need to be processed to become clear and evenly colored. Tests have shown that the best stereo effect was achieved by ensuring color consistency between the stereo pairs used for producing a 3-D stereo image.

First, we smoothed and feathered the images through histogram equalization. Then, to ensure that the final images were balanced in terms of color, without seams, ghosting, or other defects, the images were processed using professional image editing software and techniques such as radiation correction, de-overlapping, hue adjustment, brightness adjustment, and contrast adjustment in order to make the image pairs consistent in terms of the gray value before exporting the stereo pairs. The images before and after hue adjustment are shown in Fig. 3 against the base image.

Fig. 3

Images a before hue adjustment and b after hue adjustment

Image mosaicking

The KH-4B image purchased from the USGS website is usually split into four images numbered with suffixes (a, b, c, and d), and about 1/5–1/4 of each image overlapping. The images covering the extent of the study area were merged into a composite image, and the decision of whether to rotate was based on comparison with the actual surface features. It should be noted that mosaicking images manually is not efficient for a large study area covered by a large number of images. We wrote and tested a Python program for this purpose, and it can be used when necessary.

Image registration

As a key step in 3-D model construction, image registration is required to give a 3-D stereo image accurate geographical coordinates [41]. As the global navigation satellite system had not yet been developed in the 1960s, there were no ephemeris data to refer to when images were taken by KH satellites. As a result, the CORONA images do not carry any location and direction information. Affected by the cruising altitude and attitude, operating speed, and the Earth’s curvature, the built-in sensors may also generate varying degrees of geometric distortion of the images, resulting in a mismatch between elements on these images and their counterparts on the ground in terms of the shape, size, and height. Therefore, we had to assign accurate geographical coordinates to the CORONA images before using them.

For this task, the method of correcting distorted CORONA images against a base image with accurate geographical coordinates was used. This registration method involves the selection of the base image and same-name control points. If a modern satellite image were used as the base image to register historical images, it would be difficult to select same-name control points and to ensure registration accuracy. Therefore, a relatively early image that covers the extent of the study area and has accurate geographic coordinates should be selected as the reference image.

In addition, the distortion in CORONA images is nonlinear, which means that the distortion of the different parts of an image varies greatly [42]. For this reason, the selected ground control points (GCPs) must have a relatively even distribution, reasonable quantity, and accurate coordinates. In terms of the quantity, GCPs should be selected based on the area that the images cover and the requirements for image registration accuracy. In terms of the spatial distribution, after the total number of GCPs is determined, efforts should be made to ensure that these GCPs, such as roads, bridges, buildings, and the meanders and center points of narrow rivers, are evenly distributed and easy to locate (Fig. 4).

Fig. 4

Selection of GCPs for image registration

Each composite image used in this study covers an area of about 3,000 km². After repeated tests and result comparison, we found that the ideal number of GCPs for such an image is 600. The exact number of points needs to be adjusted according to the actual situation. By selecting a suitable base image and reasonable GCPs, we estimated and corrected the distortion in the CORONA stereo pairs. To verify the effect of the CORONA image registration, we superimposed a slightly transparent version of the image onto a Google Maps image reflecting the modern landforms in the area to determine how well the two overlapped (Fig. 5). To maximize the registration accuracy, we ensured that the linear surface features (e.g., each road and river) of the two images matched.

Fig. 5

CORONA images a before and b after registration with the base image

Production of optical stereo model

Neither the printed CORONA images viewed using stereoscopic glasses nor the software-generated red/blue stereoscopic images viewed indoors on a computer screen can be directly applied to field surveys. Therefore, considering the purpose of this study, the shooting time of the CORONA images, and the resolution, price, and maneuverability of the images, we collected CORONA stereo pairs covering Zhejiang Province, and with the help of the StereoPhoto Maker software, we built an optical stereo model that can be directly used for site surveys.

StereoPhoto Maker is a compact but versatile stereo image editor and stereo image viewer (https://stereo.jpn.org/eng/stphmkr/). It supports a variety of image formats, including JPEG, BMP, TIFF, and MPO, and it can automatically batch-align hundreds of images and load them into the window. We used the software’s built-in stereo image editor to display the stereo images by opening the left and right images and setting the bands, usually the red and blue bands. The stereo image with a satisfactory stereo effect was then exported. By wearing red/blue glasses, we found that the stereo effect in the mountainous areas was stronger and that in the areas with minor topographic changes was weaker (Fig. 6). When the left and right images are set up, the red/blue stereo image may display a positive or negative landform. In the case of a negative landform, a positive landform can be displayed by transposing the left and right images.

Fig. 6

Left a and right b stereo images; red/blue stereo image c created in StereoPhoto Maker

Tests showed that provided that each selected GCP basically met the requirements for registration accuracy, the stereo effect presented by the red/blue stereo image was closely related to the density of the GCPs during image registration. Images with GCPs with distributions that were too sparse or too dense were more likely to exhibit small abnormal bumps or dents in areas where no GCPs were selected. Therefore, to achieve the best stereo effect, we found that it was necessary not only to ensure the accuracy and density of the GCPs during image registration but also to adjust specific GCPs based on the stereo effect, especially in areas with special terrain features (such as places where a mountain meets a plain).

Importing images into mobile devices

Undoubtedly, the obtained red and blue stereo images can be directly used for indoor interpretation and research, and then, the results of the indoor interpretation can provide clues for field investigation, which is of great help in the discovery of archaeological sites. However, through practice, we found that it takes a long time to carry out field investigations after indoor interpretation and then to apply the investigation situation to the indoor observations for correction. If the 3-D model can be combined with the survey anytime and anywhere, the identification efficiency will be improved and the error rate will be lower.

Since the stereo pairs of the study area were accurately aligned and 3-D models were generated, it is considered that the generated models can be imported into mobile devices with positioning functions and combined with existing applications with functions such as layer management and adding notes to realize the identification of the site on the basis of the topography and geomorphology in the field, as well as to make direct decisions regarding the preservation of the site in the context of the existing geomorphology.

Map Plus, a mobile GPS navigation and map application, was selected for use in this study. This powerful iOS-based free map application supports world map viewing, searching, navigation, custom maps, offline maps, GPS track recording, KML/GPX/SHP/DXF file editing and processing, favorites collection, and photo management (https://duweis.com/zhcn/mapplus.html). By entering the red/blue stereo image’s xmax, xmin, ymax, and ymin coordinates (Fig. 7), we imported the image into Map Plus on a tablet computer to match the image with the real geographic location.

Fig. 7

Importing the red/blue stereo image into Map Plus (the blue dot on the left is the current user's location, the list on the right shows the images that have been imported)

Field survey and adjustments

Using red/blue glasses, we performed on-site stereo observations to fully interpret the topography and surface features on the historical remote-sensing images, which facilitated the search for archaeological ruins. During the field survey, tablet computers with a satellite system were used to conduct real-time locating by enabling the display location feature in Map Plus. The imported stereo image was displayed on the screen in the positive coordinate direction (Fig. 8). Given that the stereo pairs were taken from different angles, because the satellite may have changed its trajectory at any time, it was necessary to adjust the angle from which the stereo image was viewed on the tablet to obtain the best stereo effect during the field survey.

Fig. 8

a Stereo image loaded into Map Plus and b the default satellite image provided by Map Plus

For instance, the images used to construct the optical stereo model of the Liangzhu site were captured in the northwest to southeast orientation from an angle of about 6° north by east. Therefore, when producing a stereo image using the overlapping parts of the stereo pair taken by the front and rear cameras, the best stereo effect was achieved by using a parallax of 6° in the northwest to southeast orientation. After importing the red/blue stereo image into Map Plus, the mobile device was rotated clockwise by about 6° to obtain the best stereo effect.

Results and discussion

After determining that the above method can play an important role in the rapid discovery of sites, we applied this method to two areas, Liangzhu and Shaoxing, for further validation.

Survey and discovery of dams in the Liangzhu site

In March 2021, we used tablet computers that supported GPS navigation while conducting a field survey of the Liangzhu site. Wearing red/blue glasses, we could clearly see the surface features in the 1960s images and the current features at the same time. Four members of the remote sensing archaeology research team of Zhejiang Province participated in the investigation and verification. They not only understand the characteristics of the various features presented on different types of images but also have rich experience in field investigation and site excavation, and they are very familiar with the distribution of the sites in the Liangzhu area. Considering that most of the sites in the Liangzhu area are artificially piled high platforms, the researchers were able to correspond the image features and landforms on the site and to quickly make a judgment through combination with the local exploration situation on the site.

Through comparisons, we found that the landforms in many areas where ruins were located, as shown in the 1960s images, had been damaged, and thus, it would be almost impossible to identify the landforms using only modern satellite images. Very little of the ancient dam information shown on the CORONA stereo pairs survives in the modern satellite imagery. In the course of the investigation, we also found sites that looked highly suspicious on the CORONA images, and when the red and blue glasses were removed to see the current scene, we found that some of these sites had been destroyed beyond recognition. A typical example is shown in Fig. 9: wearing red/blue glasses and using the stereo image created in this study, we identified one suspected dam in the western part of the low dam area and three suspected dams on the northeast side of the Tangshan embankment of the large hydrological system around the ancient city on the Liangzhu site. These suspicious dams cannot be identified using only modern satellite imagery. Archaeological exploration has confirmed the existence of the dam in the western part of the low dam area. Of the three suspected dams identified on the northeast side of the Tangshan embankment, the existence of the two within the red ovals in Fig. 9 has been confirmed through archaeological exploration and the one circled in yellow remains to be verified via archaeological fieldwork and can be further confirmed by subsequent collection and dating of samples.

Fig. 9

a Photo showing the locations of the photos in b and c. b One suspected dam in the western part of the low dam. c Three suspected dams in the northeast part of the Tangshan embankment (lower right) in the Liangzhu site identified using CORONA stereo images

Further analytical studies conducted on the above identified and confirmed sites indicate that the dam identified in the western part of the low dam area seems to be able to enclose the already discovered high and low dams to form a larger reservoir area. The two dams identified on the north side of the long embankment of the Tangshan Mountain are located at the mouth of a narrow valley and are connected to the mountains on both sides. This structure is consistent with that of the published Liangzhu dams and can presumably be used as a part of the water conservancy system as a whole. It should also serve the functions of stopping and storing water. These findings prove that the complexity of the hydraulic system within the Liangzhu site exceeds our current knowledge, which provides important clues for further understanding the structure of the hydraulic system within the site and for the study of the function of the dam.

Survey and discoveries of Southern Song dynasty mausoleums

After verifying the accuracy of our research method by applying it to the identification of dams in the Liangzhu site, we also applied the method to the Mausoleums of the six emperors of the Southern Song dynasty (referred to as the mausoleums) to further verify its feasibility and reliability. Located in Shaoxing City, Zhejiang Province, the Southern Song dynasty mausoleums are royal mausoleums from the Southern Song dynasty. Compared with the Liangzhu site, the location and spatial layout of the Southern Song dynasty mausoleums also make full use of the natural terrain, but the difference is that the Southern Song dynasty mausoleums were built in a different historical period, and the cultural background and the nature of the site are also significantly different. These six mausoleums were constructed about 900 years ago, and the site is mainly a large-scale royal burial complex. The Southern Song dynasty mausoleums were subjected to devastating excavations during the Yuan dynasty. By the Ming and Qing dynasties, people could no longer distinguish the distribution of the royal tombs, and by the early twenty-first century, these tombs had been completely buried between the smoke and grass. Compared to the Liangzhu culture, there are more records about the Southern Song dynasty mausoleums, and after 2018, on the basis of document research, archaeologists determined the locations and structures of some of the imperial tombs through surveys and excavations.

Based on the acquired CORONA stereo pairs of the site of the mausoleums (DS1108-1070DA094 and DS1108-1070DF087), we obtained red/blue stereo images of the area using the aforementioned methods and imported them into Map Plus. Unlike the site survey in the Liangzhu area, the members involved in this survey and validation were mainly individuals who specialized in field archaeological excavation and research, except for a few who were engaged in remote sensing archaeology. They have rich experience in field investigation and site excavation, but they are not familiar enough with the feature characteristics on the remote sensing images, so the image interpretation process included some difficulties.

Through on-site identification, we used the optical stereo model to identify the ranges of the suspected imperial tombs (orange boxes labeled 1 and 2 in Fig. 10) that had not yet been identified. The site marked box 2 has been verified to have archaeological ruins via archaeological exploration, and the site marked box 1 remains to be verified via archaeological surveys. In fact, after several subsequent comprehensive interpretations of historical images of the area and site surveys, a total of 11 boxes with highly suspicious locations and structures were identified; however, most of these locations were covered by dense vegetation and were therefore difficult to identify, either by means of site surveys or observations of modern images. Upon comparison, accurate locations and extents of both the excavated and currently researched mausoleums were identified. This further proves the feasibility of our method of using CORONA stereo image pairs to identify ancient sites.

Fig. 10

Results of the identification of the mausoleums

Discussion

This paper proposes a new method of using CORONA stereo image pairs for ancient site identification. Images of different areas were decoded and investigated on-site by different personnel in this study. In terms of the empirical research, the study of the Liangzhu area provides more clues for determining the distribution of the dams in the Liangzhu period and also contributes to a more in-depth understanding of the urban planning of Liangzhu, an ancient civilization. Studying the six Song dynasty mausoleums area is conducive to gaining a macroscopic understanding of the structure of the royal mausoleums constructed during the Southern Song dynasty while providing a more comprehensive perspective for understanding the history and culture of this period.

The results of our study prove that the new method proposed in this paper is practical and feasible, takes full advantage of CORONA stereo image pairs, and can provide new perspectives and evidence for site identification and the analysis of human-land relationships. However, we also found that it is ideal to involve researchers who are familiar with the image features and who have a better understanding of the conditions of the study area to identify the sites and study the relationship between the sites and the environment through image interpretation. There are times when identification without knowledge of the image features in the study area and without reference to existing data can be effective, but it is much more effective if targeted identification is conducted based on combination with existing results or knowledge. As for the prehistoric period, for which sufficient information is not available for reference, or for which no clear written records are available, it can only be directly deciphered and recognized. Then, the results of fieldwork and verification can be fed back to the interpreters so as to improve the identification methods and ideas. This will take a longer period of time, but we also think that this is a necessary step in scientific research.

Conclusions

In this study, we did not follow the conventional practice of constructing photogrammetric DEMs based on stereo pairs taken by CORONA satellites. Instead, we utilized previously rarely used red/blue stereo images. Compared with the conventional approach, the method developed in this study is more convenient and easier to implement because it involves lightweight data that are easier to interpret, have more efficient processing, and are compatible with widely used devices, software, and platforms. In particular, our methodology has been refined to enhance the precision, and we have also pioneered the integration of results into mobile applications.This method can synchronize real-time locating using modern and historical satellite images, making it easier to identify archaeological ruins by comparing ancient and modern landforms. Archaeological researchers can directly import the stereo pairs into mobile devices (such as mobile phones and tablet computers) that support real-time locating. In the survey area, they can synchronously observe the landforms from the same orthophoto angle and compare them with those from the 1960s to identify archaeological ruins. In fact, we created a digital sand table that overlays the images of the landforms taken in the 1960s and the images of the current landforms as a new and more economical way to search for and accurately locate archaeological ruins in field surveys.

Our results of the identification of archaeological ruins at the Liangzhu site and the mausoleums of the six emperors of the Southern Song dynasty fully prove the feasibility of the research design and method developed in this study. Furthermore, we found that the surface features are greatly exaggerated in the optical stereo model constructed using CORONA stereo pairs, making it easier for field researchers to identify subtle terrain differences and thus providing more reliable evidence for ruin identification. This finding further validates the advantages of the proposed method.

Our research method has the potential for wider use because it can simplify the processing and reduce the difficulty of image generation and interpretation. It also enables direct observation of stereo pairs during field surveys, which enhances the efficiency of locating the range of archaeological ruins and identifying the landforms of archaeological sites. This method can provide important clues and evidence for archaeological field surveys, exploration, excavation, archaeological research, and cultural heritage protection. Databases of historical base images for some regions in China have been established in recent years. For example, databases of province-wide registered base images from the 1960s have been created for Jiangsu and Zhejiang. The use of these images and the research method proposed in this paper can greatly improve the efficiency of province-wide systematic field surveys and exploration.

In summary, the advantages of this method are obvious, such as the small amount of data that needs to be processed, the uncomplicated process of data processing, the ability to observe some landscapes that have disappeared over time, the convenience of using the results for field archaeological investigations, and the low threshold for operation in the field. However, it is undeniable that the method also has shortcomings. For researchers engaged in remote sensing archaeology, micro-geomorphic differences and site features can be easily identified using this method, but for those who are not familiar with satellite imagery, especially this single-band historical remote sensing imagery, it is not so easy to identify sites. Moreover, the destruction of cultural heritage sites often occurs in a very short period of time, which may lead to passive and lagging research that should be forward-looking. Another shortcoming of our method is that it is impossible to further determine the time of the formation of a suspicious site after it has been identified through remote sensing technology, so a suspicious site identified using imagery may have been formed thousands of years ago, or it may have existed for a short period of time before the imagery was taken. This is a problem that archaeological work using remote sensing technology faces and needs to be further verified in conjunction with archaeological work and dating.