Home > Streamlined Archaeo-Geophysical Data Processing and Integration for DoD Field Use > Technical Description
Abstract | Problem Statement | Technical Description | Implementation Plan | References
Technical Objectives | Data Processing Details | Processing Steps
The ESTCP project will formalize and demonstrate the methodology developed in the Department of Defense Strategic Environmental Research and Development (SERDP) project CS-1263 New Approaches to the Use and Integration of Multi-Sensor Remote Sensing for Historic Resource Identification and Evaluation and consolidate these proven methods in an integrated software package “ArchaeoMapper”. This software will combine the best features of existing software with the new techniques resulting from SERDP research and will be designed for use by DoD and other CRM professionals. The product will then be subject to rigorous demonstration and validation by DoD and civilian agency archaeological remote sensing field users and professional practitioners. CS-1263 has developed and validated an effective multi-sensor approach to the acquisition, processing and fusion of multiple sensor data sets acquired for the characterization and assessment of subsurface archaeological resources. CS-1263 has clearly demonstrated that multi-sensor data, when used in this manner, can accurately and cost-effectively provide critical information needed by installations to meet the requirements of the National Historic Preservation Act (NHPA), the Native American Graves Protection and Repatriation Act (NAGPRA) and other national and state heritage preservation laws and implementing regulations (esp. 36CFR800).
The use of multi-instrument data fusion algorithms for archaeological remote sensing was pioneered in the SERDP project, including the successful implementation of color compositing, principle components analysis, logistic regression, decision tree, image segmentation and related image classification methods that have been developed in CS-1263 to fuse geophysical data (Kvamme 2005). In order to fuse these data from different sensors, however, a number of initial processing steps are required. “Raw” remote sensing data are not reliably interpretable in terms of the presence/absence and nature of cultural deposits. A relatively complex sequence of data processing steps is required to produce an optimal (or even usable) image (Clark 2000). Data processing includes numerous preprocessing routines necessary to make raw data suitable for integration, display, fusion and interpretation as shown below. Data integration or fusion implies the use of various graphical, mathematical, and statistical algorithms to fuse multiple images into one product that portrays the pertinent information from each layer.
Processing sequence for Pueblo Escondido magnetic susceptibility acquired and processed for the SERDP project CS-1263: (a) unprocessed, (b) basic processing applied, and (c) image enhancement.
A number of sensor types (e.g., ground-penetrating radar, magnetometry, electrical resistance, induced electrical conductivity, magnetic susceptibility, multispectral scanning, panchromatic photography, and thermography) are useful in remote sensing investigations of archaeological sites because the various sensors respond to different physical properties of the archaeological record. It is frequently difficult to predict which single sensor will provide the best results at a particular site. Therefore optimization of a remote sensing investigation typically requires use of at least several different instruments. In addition to the variability introduced by the instrument suite selected, the existing approach to processing each sensor’s data involves a variety of software packages; each with its own strengths, limitations, data formats, and idiosyncrasies. Additional software packages are then required to bring all the processed data sets together into one environment for fusion and interpretation.
SERDP project CS-1263 exemplifies the benefits of using multiple sensors and processing strategies at a site. At this time however, the processing of each instrument’s data requires the use of software that was uniquely designed for the specific instrument but not designed, in most instances, for archaeological data requirements or to allow integration or fusion of multi-sensor data. The multi-sensor field surveys, each involving coverage of 1 to 1.5 hectares and utilizing the most widely applicable sensors, required some 500 hours of data processing and integration per site. The GPR processing alone required the use of one program for preprocessing of profiles (RADAN) and a second program for time slicing (GPR Process or IDRISI). Each slice map then had to be manually formatted, mosaicked, and enhanced using a combination of three image processing programs (Excel, Surfer and Geoplot). To integrate the GPR slice maps with the other sensor data, four different COTS software products (IDRISI, ArcGIS, Surfer, and SYSTAT) were needed to (1) co-register and re-project the data layers to a common coordinate system, (2) interpolate data layers to a common resolution, (3) generate statistical summaries, (4) perform statistical and graphical fusions, and (5) display the resultant images. In each step where data are converted from one format to another it is challenging, and sometimes impossible, to retain the processing sequence metadata and the original dynamic range (such as conversion from 32- or 16-bit to 8-bit data).
These problems affect each of the sensor data formats but are particularly invidious for GPR data. GPR software, like the hardware, is principally designed for non-archaeological applications such as detection of pipes, utilities, tanks, and drums, and inspection of concrete and rebar. Only in the past decade or so have archaeologists begun to utilize GPR in a wide range of applications and realize its potential, and only in the past few years has it been used to cover the same large expanses as other methods (Ernenwein and Kvamme 2005). The expertise needed to bring GPR data processing up to the same level of efficiency as other archaeo-geophysical methods now exists, but has not yet been implemented in software.
Given the situation outlined here, multi-sensor surveys can currently only be competently undertaken by specialists that have mastered a large number of highly disparate software and data processing protocols. Labor undertaken by such specialists is understandably expensive, yet their work requires hundreds of hours of repetitive processing and data management. Transformations of data from one software package to another present numerous opportunities for error, and the fact that many of the current systems require conversion of data to alternative formats means that important spectral information can be needlessly lost. Remedying this situation will substantially reduce the high cost of archaeological remote sensing.
