Creating Digital Versions of Historic Ordnance Survey Maps

Introduction
Source Material
The Digitisation Methodology
Conversion of Historical County Series Maps into National Grid tiles

Introduction

The following text is adapted from support documents provided by Landmark Information Group, with kind permission.

Ordnance Survey has one of the largest collections of historical mapping in the United Kingdom and until recently this was only held as a paper archive. As the result of a joint venture between Ordnance Survey and Landmark Information Group Ltd the paper archive has been converted into digital format. The digital archive consists of 1:10 560 or 1:10 000 and 1:2500 scale County Series and National Grid mapping.

Back to top

Source Material

The source mapping in the scanning process was the original mapping published at the time of survey and held in the OS archives. These maps are working editions and have therefore been in constant use over the last century. The maps are stored on racking and were only bagged within the last decade; this, coupled with constant handling, has resulted in some maps being torn, folded, and affected by a certain amount of dust. Maps used in the daily business of OS have also been drawn / written on by those who have used them, often obscuring the data originally published (i.e. boundary changes).

Finally, because storage has not been in a controlled environment, the paper of some maps have become slightly warped, obviously affecting the accuracy of that map; this is especially the case with Full Sheet 1:10,560 maps.

A number of factors affect the quality of the scanned image. In particular dust on the maps has resulted in dark areas - other imperfections result from folds and tears which make it impossible to scan an even image; in some extreme cases a section of the map is missing resulting in a loss of that data. It is important to note, however, that damaged maps are the exception and that most of the historical mapping is in very good condition especially considering its age.

Although the collection of paper maps in the OS archive is approximately 93% complete there are maps missing. It is the intention of the joint venture (OS & Landmark Information Group) to locate and scan the missing maps and include them in the archive.

Back to top

The Digitisation Methodology

The production process is composed of a number of stages which are described below:-

Scanning

Scanning was commenced by Landmark staff during May 1995. Industrial roller scanners were chosen to meet operational needs, as over 400,000 maps had to be scanned. The resolution chosen, 300 dpi, matched scanning speed, with file size and line detail. The results are clear images and manageable file sizes, although on poorer quality maps a higher dpi would have resulted in cleaner, crisper images this would have meant considerably larger files, making them less manageable, costing more to store, and would have slowed the scanning team considerably. The data format chosen for the raster images was Binary (black and white pixels) tiled TIFF with group 4 compression.

Owing to the use of multicamera roller scanners a minor distortion can be introduced into the raster image. This is due in part to the camera optics not being 100% linear and the constant stretch or compression error caused by the rollers as the map is fed into the scanner.

The acid test for the scanning process was to ensure that the images were of an acceptable standard for inclusion in the range of products developed by Landmark Information Group and OS.

Raw Audit

Before processing could commence, each raster image, i.e. scanned images in their original format, had to be opened, and its date and map name checked to ensure that the scanning was successful. An Excel file was created listing all maps available and was run through a database comparing it back to an actual DOS listing of the raster images. Any mistakes were corrected and the database check was re-run until no errors occurred. The raster images and Excel file were then backed up to CD-ROM creating an accurate raw record of these historical maps.

Processing

In order to tile the data and create a mosaic of Historical Maps covering the whole country, each scanned image had to be processed, removing all information outside the neat line leaving just the map tile itself. Image Centre (supplied by S.S.I.) was chosen as the processing software as it was the best available at that time which could perform the specialised tasks required.

The process involved two phases. The first ensured that each image was orthogonal by placing two points along one edge of the map (top or bottom), Image Centre levelled out this line leaving the map horizontal. The second process ensured that the image was cropped from one corner to the diagonally opposite corner removing all data outside the crop marks in order to obtain a 'tile'.

This process produces accurate historical tiles. Unfortunately a small number of the maps have become warped with age; the limitations of Image Centre meant that it could deskew to only 0.004 of a degree, and scanning, no matter how accurate, will not always produce perfect copies. As a result the maps were not perfect rectangles and once deskewed, although they were flat on the edge chosen, the opposite edge was still at an angle so that when cropped by selecting diagonally opposite corners, areas outside of the detail were left. These had to be removed which meant a second crop using the other two corners as reference points. Because of this limitation some data could be lost, resulting in a less than perfect join when the maps are lined up.

Processed Audit

The QA process involves opening every processed map and ensuring that the following two phases had been completed correctly: one, checking that there were no black / white lines; two, making sure that the maps were readable. Note that the actual 'paper size' of each map varies and so the number of pixels in the .TIF images will vary in pixel extent in X and Y. The cropping process (see above) merely involves removing unwanted data, it does not involve re-sizing the image. A DOS listing of the processed maps was then compared against the raw Excel file to ensure no errors occurred. Once audited, the data was backed up to CD-ROM.

Summary

This process results in an Excel file listing all maps (including dates) and two CD-ROMs, one with raw files and one with processed files which all match. Once a county revision was completed the task was then repeated with the next county revision.

Index

The shape of each county grid was derived from the Ordnance Survey book of indexes for England and Wales and Ordnance Survey book of indexes for Scotland. Each county index was then transferred into a GIS and projected on the original county origin. Each rectangle was named with a unique identifier made up of the county identifier and the map sheet name.

Production of Co-ordinates

The County Series index maps have been used to calculate the national grid co-ordinates for each map sheet. The unique identifier (as mentioned above) is listed with the four pairs of co-ordinates that define the corners of the map sheet.

Back to top

Conversion of historical County Series Maps into National Grid tiles

The County Series Mapping was surveyed using the Cassini Projection, which was replaced in the 1950's when Ordnance Survey brought in the National Grid system.

The National Grid Projection differs to the one used for the County Series maps. In order to overlay National Grid on to County Series one set of data needs to be re-projected. Many GIS software products cannot perform this task, so RMSI India with the assistance of Landmark Information Group Ltd and the Ordnance Survey have developed a process where by 1:10,560 or 1:2,500 scale County Series maps can be converted into actual National Grid tiles.

Each raster image (Banded Tiff) is reprojected to fit the National Grid projection and cut out into its National Grid blocks. The white background data is then made transparent so that other Raster images can be imported, warped and joined to this file. This creates a single National Grid file which can form a mosaic of up to eight Cassini Maps.

Each file is named using Ordnance Survey's naming conventions with a two character prefix relating to which historical County the data originated from.

For example: 05SJ3965

Note that because the resulting National Grid Tiles are cut from a mosaic of County Series data, white areas will occur in the tiles either where there are County Boundaries or because the historical data is not available.

Back to top