Applications in Archaeology

Dorel Micle, Department of History, UVT Daniela Zaharie, Department Computer Science, UVT

Motivation There is need for tools which assist the archaeologists to: „

Identify the traces of ancient human activity and locate potential archaeological sites

„

Analyze the spatial distribution of archaeological sites (e.g. burial mounds) and get information about their morphometrical properties (size, distances etc)

GiSHEO Training - 27 September, 2009

2

Context „

The available data are usually: ‰

„

Satellite images (preferable of high resolution)

The tool should offer the possibility to: ‰

‰

Enhance the image characteristics in order to make easier the identification of elements of interest by visual inspection (semi)automatically detect potential location of structures os interest: „ Linear structures „ Circular structures

GiSHEO Training - 27 September, 2009

3

Searching for linear fortifications and medieval roads Example 1: find the orientation of a fortification located near Vladimirescu (Arad) Remark: The image contains both obvious shapes (e.g. recent roads) and less obvious ones (these one are of interest)

GiSHEO Training - 27 September, 2009

4

Searching for linear fortifications and medieval roads Step 1: Grayscale conversion

GiSHEO Training - 27 September, 2009

5

Searching for linear fortifications and medieval roads Step 2: Histogram equalization

GiSHEO Training - 27 September, 2009

6

Searching for linear fortifications and medieval roads Step 3: Smoothing the equalized image

GiSHEO Training - 27 September, 2009

7

Searching for linear fortifications and medieval roads Step 4: Edge detection: Canny filter Parameters: Minimal Threshold: 80, Maximal Threshold: 160 Aperture size: 3

GiSHEO Training - 27 September, 2009

8

Searching for linear fortifications and medieval roads Step 5: Line detection: Hough transform Parameters: Accumulator Threshold: 125, Minimal Line Length: 180 Maximal Gap between Lines: 80

GiSHEO Training - 27 September, 2009

9

Searching for linear fortifications and medieval roads Step 6: Postprocessing: keep only the lines having the slope in a specified range Example: keep only the lines which have an angle between 40 and 50 degrees

GiSHEO Training - 27 September, 2009

10

Searching for linear fortifications and medieval roads Remark: the false lines can be avoided by carefully choosing the parameters of the Hough transform Left: Accumulator Threshold: 125, Minimal Line Length: 180, Maximal Gap: 80 Right: Accumulator Threshold: 125, Minimal Line Length: 500, Maximal Gap: 100

GiSHEO Training - 27 September, 2009

11

Searching for linear fortifications and medieval roads Exercise 1: Identify the linear fortification (“roman wave”) located near Alios (Timis)

Steps: 1. 2. 3.

Search for the image corresponding to the region Process the image Interpret the results

GiSHEO Training - 27 September, 2009

12

Searching for linear fortifications and medieval roads Step 1: Search for the image corresponding to the region Variants: • Search by the name vendor:UVT place:”Alios/Timis/Romania”

•

Search by the coordinates of a bounding box vendor:UVT bbox:(21.4625,46.0407,21.4962,46.0642) (21.4962,46.0642)

(21.4625,46.0407) GiSHEO Training - 27 September, 2009

13

Searching for linear fortifications and medieval roads Step 1: Load the image corresponding to the region

GiSHEO Training - 27 September, 2009

14

Searching for linear fortifications and medieval roads Step 1: Load the image corresponding to the region

GiSHEO Training - 27 September, 2009

15

Searching for linear fortifications and medieval roads Step 2: Convert the image to grayscale

GiSHEO Training - 27 September, 2009

16

Searching for linear fortifications and medieval roads Step 3a: Histogram equalization

GiSHEO Training - 27 September, 2009

17

Searching for linear fortifications and medieval roads Step 3b: Linear stretching

GiSHEO Training - 27 September, 2009

18

Searching for linear fortifications and medieval roads Step 4: Edge detection (Canny filter)

GiSHEO Training - 27 September, 2009

19

Searching for linear fortifications and medieval roads Step 5: Lines detection (Hough transform)

GiSHEO Training - 27 September, 2009

20

Searching for linear fortifications and medieval roads Results: Lines detection (Hough transform)

GiSHEO Training - 27 September, 2009

21

Searching for linear fortifications and medieval roads Results: Lines detection (Hough transform) Canny param.:tmin=30,tmax=60,aperture size=3

GiSHEO Training - 27 September, 2009

22

Searching for linear fortifications and medieval roads Results: Lines detection (Hough transform); after post-processing Parameters: acc_threshold=175, min length=800, max gap=100

GiSHEO Training - 27 September, 2009

23

Searching for burial mounds Aim: Identify the burial mounds located near Nerau

Remark: the land marks of the burial mounds look like some disks with lighter or darker interior

GiSHEO Training - 27 September, 2009

24

Searching for burial mounds Aim: Identify the burial mounds located near Nerau

Remark: 1. this image contain three burial mounds; their location was confirmed on the terrain 2. There could be other spots in the image which do not really correspond to burial mounds

GiSHEO Training - 27 September, 2009

25

Searching for burial mounds Step 1: Gray level conversion

GiSHEO Training - 27 September, 2009

26

Searching for burial mounds Step 2: Histogram equalization

GiSHEO Training - 27 September, 2009

27

Searching for burial mounds Step 3: Edge detection by Canny filter Parameters: minimal threshold =35, maximal threshold=70, aperture size=3

GiSHEO Training - 27 September, 2009

28

Searching for burial mounds Step 4: Circles detection by Hough transform Parameters: minimal distance between circles=20, Canny maximal threshold=70, Accumulator threshold=35, Minimal radius=2, Maximal radius=15

GiSHEO Training - 27 September, 2009

29

Searching for burial mounds Step 5: Postprocessing: eliminate some of the false circles (keep the circles with homogeneous interior)

GiSHEO Training - 27 September, 2009

30

Searching for burial mounds Exercise 2: Identify other burial mounds located near Nerau Step 1: Load the image corresponding to the bounding box: (20.5759, 45.9819, 20.6102, 46.0058) Step 2: Convert the image to grayscale Step 3: Apply histogram equalization

GiSHEO Training - 27 September, 2009

31

Flows of basic operations Enhancement of shapes corresponding to ancient settlements

Enhancement of linear shapes corresponding to wave-like fortifications

‰ Gray level converting

‰ Gray level converting

‰ Histogram equalization

‰ Emboss (convolution operation)

‰ Quantization and/or binarization

‰ Histogram equalization

‰ Layers combination

‰ Layers combination

Tool used in tests:

GiSHEO Training - 27 September, 2009

32

Identified challenges ‰ The buried archaeological remains are expressed by weak signals which are obscured by stronger signals emitted by actual structures (roads, buildings) ‰ Data fusion (panchromatic and multispectral images)

‰ It is difficult to discriminate between ancient and actual structures ‰ use of GIS information (current road networks) ‰ use of expert knowledge (e.g. ancient settlements were usually on northern part of hills, near river etc.)

GiSHEO Training - 27 September, 2009

33

Conclusions ‰ Even flows of basic operations on images allow to enhance the image in order to identify easier by visual inspection the elements of interest ‰ Useful in training of students in landscape archaeology ‰ More sophisticated operations are needed to develop a semiautomated tool able to identify the location of elements of interest and to assist the user in obtaining morphometrical information

GiSHEO Training - 27 September, 2009

34