4.2 Multi-Spectral Scanner Imagery

Modern multi-spectral sensors use a diffraction grating to split electro-magnetic radiation into a number of discrete channels that are recorded with photosensitive detectors to produce multi-channel digital images. This technology offers the potential of imaging at wavelengths beyond the visible spectrum and beyond that of photographic infrared film. Narrow band spectral imaging can often help to enhance or distinguish different vegetation types according to its particular absorption and reflectance properties. It also offers the potential to image emitted radiation at the thermal infrared part of the electro-magnetic spectrum. Digital image data can also be enhanced and classified by computer-based image processing. PCI EASI/PACE software was used as the primary image processing system running under Windows on 90 and 133mhz. Pentium-based PCs.

4.2.1 Daedalus 1268 airborne multi-spectral imagery

The Daedalus 1268 Airborne Multi-Spectral Scanner, used to collect the data for this project, images at the visible, infrared and thermal parts of the electro-magnetic spectrum, split into eleven discrete bands. In all, 12 channels of data were collected, and these are shown in Table 1. Channels 11 and 12 image at the same wavelength, but channel 12 is collected using a half the gain (exposure) setting, thus providing a different radiometric sensitivity to avoid problems with over-exposure. The ground resolution of the sensor is dependent upon aircraft altitude and airspeed. However, the majority of imagery was obtained with a nominal ground resolution of 2m.

Band numberWavelength (um)Spectrum component
One 0.435-0.45 blue
Two 0.45-0.52 blue-green
Three 0.52-0.605 green
Four 0.605-0.625 red
Five 0.63-0.69 red
Six 0.695-0.75 near infrared
Seven 0.76-0.9 near infrared
Eight 0.91-1.05 near infrared
Nine 1.55-1.75 short wave infrared
Ten 2.08-2.35 short wave infrared
Eleven 8.5-13 mid-infrared or "thermal"
Twelve 8.5-13 mid-infrared or "thermal"
Table 1: Daedalus sensor characteristics

4.2.2 Image analysis, feature extraction and recording

The raw data (440 megabytes) was copied from the tapes provided by the NERC, and the 8-bit image files were 'byte-swapped' to a standard Intel format to facilitate image processing on a PC. The image files were enhanced using various contrast stretching, density slicing and channel combination techniques.

Finally, images were geometrically rectified to the OS National Grid. EASI/PACE software was used for image to image rectification using Ordnance Survey 1:10000 raster maps to gather the registration data. However, this process was not begun until analysis of the relevant bands on the image was complete, due to the possible loss of data resulting from the resampling of the image pixels during rectification. The correction process is complex as a result of the distortion caused by the plane's movement in three dimensions (height above ground level, side to side motion and tilting of the wings or 'yawing') during the flight. The very high ground resolution of the data, coupled with the fixed position of the recording instrument and therefore sensitivity to yawing in particular, meant that a large number of control points were required to rectify each image. Once corrected, the images were draped onto a digital elevation model (DEM) of the project area.

The area covered by the multi-spectral image is 48.33 square kilometres and so the analysis of all 12 bands of imagery is a major undertaking, and is not yet fully complete. However, the data processed so far show that certain bands have only limited relevance to archaeological feature detection, most notably Band 1 (wavelength 0.42-0.45 micrometres, colour 'blue' in the visible spectrum). Band 9 (wavelength 1.55-1.75 micrometres, short-wave infrared) also proved to be of limited value for the purposes of this project. Of the remaining 10 bands, Bands 4, 5, 7, 10, and 11-12 have shown the most potential for feature recognition, with Bands 2, 3, 6 and 8 providing less definition. In addition, various combinations of bands were investigated, with a combination of bands 7 and 12 proving particularly useful.

Analysis of multi-spectral image data has much in common with analysis of aerial photographs, in that experience is necessary in order to separate the genuine archaeological features out from those features which are caused by natural action or modern human activities. The multi-spectral images are also rich in geological information, for instance clearly showing the dry valleys and relict stream channels running from the foot of the Wolds out into the Vale of Pickering and the complex of relict channels found in the centre of the valley. The sensitivity to relative water content within the crops and the soil, reflected in the thermal channel in particular, has important potential for the recovery of detailed geomorphological data and the isolation, for instance, of relict channels, potential peat deposits which could assist in identifying areas for palaeo-environmental research and gravel islands which may have supported human habitation or ritual sites. One aspect of the MSS data that proved more problematic was that of image classification. Classification of low-resolution digital MSS data, such as that gathered by LANDSAT or SPOT satellites, can be used to identify crop and soil types with relative ease and to a large extent automatically; the very high resolution of the recovered data coupled with the variable response within a single field and crop makes such automatic classification of these data impossible. Slight variations in the spectrum recovered across a single crop were such that simple classification, even at the crop level, was impossible. Any idea that the archaeological component could be quickly extracted through a classification scheme had to be quickly discounted.

An interactive and iterative process was required in which each field had to be examined in detail with a variety of different filters applied to build up a composite picture of the features recorded. The immediate benefit of the MSS image over the established database of oblique air-photographs was the potential to follow features for several kilometres; fragments of features identified from the air could be joined to allow for the assessment of the whole landscape. Where features hardly showed in the vertical photographs and not at all in the oblique record they could be enhanced within the digital image, demonstrating the presence of previously unidentified trackways across the valley, enclosure systems and extensive barrow cemeteries which have forced a re-appraisal of the landscape, particularly towards the centre of the valley. Areas that had previously been identified as small islands now appear to be more extensive, very slight, linear ridges running parallel to the valley edge. In one case a gravel ridge forms the setting for a cemetery of round and square barrows extending over more than a kilometre of ground that is linked to the dry land to the south by a ditched trackway.

To research which bands in the multi-spectral image would provide the most potential archaeological information, an area with identified crop-mark coverage and apparently blank areas was selected for detailed study. Sections were extracted from the image and all 12 bands examined in detail both individually and in combination with other bands. Histogram equalization and contrast stretch functions were applied to each band to improve definition of possible archaeological features. Of particular interest was an area of 'Ladder Settlement', a sequence of overlying enclosures following a trackway (of late Iron Age to Romano-British date) which follows the 27 to 31 metre contour lines along both sides of the Vale of Pickering. Such 'Ladder Settlements' have a widespread distribution in the region and are found extensively in the valleys of the Yorkshire Wolds as well as on both the northern and southern margins of the Vale of Pickering. That on the northern side of the Vale of Pickering is less well understood as it lies beyond the focus of the project area and has received less aerial photographic attention.

Although the 'Ladder Settlement' is believed to be a continuous landscape feature running along the southern margin of the Vale, aerial photographic plots had left a number of apparent 'gaps' in the line of the settlement, particularly in areas of pasture, and it was hoped that these 'gaps' could be filled in with new data derived from the interpretation of the multi-spectral image. Two areas in particular were investigated (see Figs. 8 - 11 ), where the line of the Ladder Settlement was believed to exist, but where there was no evidence of the settlement from either the oblique or the vertical colour photographs. In both cases, evidence of the continuous nature of the settlement was visible; in Area 1 Bands 7, 10, 11 and 12 all showed clear indications of the typical linear features associated with the settlement. In Area 2, Band 11 gave the clearest definition of these linear features. In addition, the known line of the Ladder Settlement was also visible in the multi-spectral image, in many cases equalling the definition obtained by the combined aerial photographic plots.

During the analysis of the imagery, many previously unrecognised features were noted. One example, a potential hengiform enclosure towards the centre of the valley, with an internal circular feature, is clearly visible only in the thermal bands (11 and 12), although Band 7 (near infrared) also shows a faint trace. Following this discovery, the aerial photographs were scrutinized for any indications, and although there was an indication of one part of the western half of the feature in the vertical photographs, it was ill-defined, and would certainly not have been picked up during a normal scan of the aerial photographs. We hope to carry out further geophysical surveying later in 1997 in order to verify the existence of the potential enclosure.


© Internet Archaeology
Last updated: Wed Apr 23 1997