WWII Bomb Damage


Case Study

Extraction of Map Symbols from World War II Bomb Damage in Hand-drawn 1940's Age Maps

View Imagery

Background

Maps illustrating the scale of damage to London by bombs and V1 and V2 missile atacks in World War II were hand drawn using color coded symbols to show damage to buildings. The first dataset is annotated with color circles that range from black, to show total destruction, and purple illustrating damage beyond repair, to yellow for minor damage. Open circles were used to delineate areas hit by V1 bombs and V2 long range rockets. A second set uses similar color codes but identifies individual buildings. These maps were created to show the severity of damage to individual buildings across approximately 117 square miles of London to aid in the rebuiliding effort after the war ended. (Source: BBC News, 06July2005)

Old and Faded Maps

The original base maps (reproduced from the1914 Ordnance Survey map) were used as base maps for the annotated 1940's age maps, Though scanned at a high resolution, they remain visually difficult to discern colors and symbols (Images 1 and 2). The desire was to produce shapefile content of each damage classification. With over 2000 maps, one with approximately 950 and the other over 1100, hand digitizing the information would be both a tedious and time consuming process considering the volume of imformation and the quality of the data.

Genie Pro was used to examine one image of each type and output shapefiles for each class. The results in the first set were then tested against hand counts. Genie Pro produced excellent results. Image 2 is a close up of a portion of the image showing the faded quality of the colors that include black, red, yellow, green and purple color markers. Image 3 shows the results image produced by Genie Pro with the background in white and each individual damage assessment class colorized according to the orginal annotation scheme.

Results were compared to hand counts in the first dataset. Using Genie Pro was very successful with 100% detection rate of purple, yellow and blue shapes, 97% for the green, 99% for the red and 89% for the black. The black count was less successful due to the overwhelming black content in the image. An overall across class accuracy was 95.8%.

The second dataset (Image 5) is based on individual building damage rather than the generalized area designations of the first dataset. The results layer shows that Genie Pro performs extremely well in separating out the individual buildings in each damage classification. Images 5 and 6 show Genie Pro results for the second dataset. Truth data was not determined for accuracy evaluation.

Images

Image 1 - Image Dataset 1 - An original image shows the poor quality of the maps and the colors of hand drawn symbols and markers.

Image 2 - A portion of the image showing the original symbols and colors.

Image 3- A solution was derived using the 7 classes of the original annotation scheme and an additional background class. The same area as seen in image 2 is shown here with white as the translucent background of the results layer. Each symbol is color coded and represented as either a solid circle or lines on the image.

Image 4 - The same portion of the results image as in Image 3 with the original base image removed to better view the classes. Each class was then vectorized and exported as shapefiles.

Image 5 - Image Dataset 2 - This original map uses colors and symbols to assess individual building damage and bombed areas.

Image 6 - The results layer shows that Genie Pro performs extremely well in separating out the individual buildings in each damage classification. With the many maps that need to have information layers derived from them, it is again the quality of the images and the large number of individual buildings that need to be quantified that makes a Genie Pro based solution extremely valuable.

Image 7 - A portion of the results layer. Genie Pro Results layer with brown as the background color and 8 additional classes using the color designations as in the original annotation scheme.

Source: BBC News, 06July2005, http://news.bbc.co.uk/2/hi/uk_news/england/london/4655437.stm, accessed 06June2012.
Source: © Crown Copyright 1914