automatic detection and characteristics extraction of ... · this presentation deals with the...

Automatic detection and characteristics extraction of archaeological structures using coupled Geographic

Information System software and Scilab

J-P. Toumazet, F. Vautier, E; Roussel, M. Flores & B. Dousteyssier

Automatic feature extraction has been initially developed for industrial applications.

A new domain for this application appeared a few years ago : the use of Airborne laser scanning (ALS) of archaeological regions of interest allowing accurate topographic and micro topographic description.

This method allows the reconstruction of reliable digital terrain models (DTM) of forested areas where traditional prospection methods are inefficient, time-consuming and non-exhaustive.

The ALS technology provides the opportunity to discover new archaeological features hidden by vegetation and allows a comprehensive survey of archaeological sites within their environmental context.

Introduction

But the obtained point clouds produce a huge quantity of data. They are still generally analyzed by a human operator. This process is time-consuming, subjective and may be non-exhaustive.

It’s why a method of automatic detection and characterization of archaeological structures, based on Scilab algorithm has been developed. The obtained results will be presented and discussed in the next slides.

Example of antique structure, discovered with a LIDAR

This presentation deals with the process of automatic detection of archaeological structures.

It is applied to former agricultural constructions, built from the medieval to the modern period.

They can be found in very high densities in some places in Auvergne. These structures have been chosen to test the process of automatic detection because they are particularly delicate to treat : they are indeed very variable in forms, appearing sometimes isolated, sometimes in group and they are characterized by small relief variations very difficult to detect among the global relief.

Introduction

What is a LIDAR ?

LIDAR for archaeological applications

From 3D point cloud to Local Relief Model

Automatic detection applied to archaeological structures

Conclusions

What is a LIDAR ?

Acronym for Light Detection And Ranging

Lidar instrumentation

�Laser Source

�Laser Detector

�Scanning mechanism & controller

�Electronics for timing emissions & reflections

�Airborne GPS (position, speed, direction)

�Inertial Measurement Unit (orientation angles)

�High Performance Computing Support

�� High Capacity Data Recorders

Courtesy of Imao Aerial data acquisition

What is a LIDAR ?

Acronym for Light Detection And Ranging

Principle

1) A pulse of light is emitted and the precise time is recorded.

2) The reflection of that pulse is detected and the precise time is recorded.

3) Using the constant speed of light, the delay can be converted into a distance.

4) Knowing the position and orientation of the sensor, the XYZ coordinate of the reflective surface can be calculated.

Courtesy of Ohio departement of transportation

zx

y

GPS + INSControlled

1st Return

2nd Return

3rd Return

Multiple ReturnIntensity of Return

POSITION OFINSTRUMENT

DAR HEIGHT ANDER DETERMINATION

What is a LIDAR ?Generally, there is not only one but several returns, depending on the vegetation cover

Magnitude

time

Emitted signal

Received signal

Example of obtained results

Allier river section : 10 kmsLiDAR cloud points : 104 732 500Average density : 215 pts/m²0.01 % shown

Surface : 8500 m²Nb of points : 1 835 375100% shown

Water areas

Data are then classified : The point cloud is filtered to obtain a Digital Elevation Model (DEM) that represents the ground.

It allows also the extraction of « above ground objects » - i.e. Trees, Buildings, etc.


Water areas


It allows also the extraction of « above ground objects » - i.e. Trees, Buildings, etc.

Ground and vegetation

CROSS SECTION (profil depht : 4.5 m)

Ground


Water areas


It allows also the extraction of « above ground objects » and features - i.e. Trees, Buildings, etc.

CROSS SECTION (profil depht : 4.5 m)

LIDAR for archaeological applicationsView of the investigated zone located near the Puy de Dôme volcano. It is covered by a dense forest of deciduous trees. In winter, when no more leaf remain, the laser beam of the LIDAR penetrates the canopy and allows the calculation of relief models.

LIDAR for archaeological applications

Digital Elevation Model (DEM) obtained from the LIDAR results.

Process used to calculate the DEM

Point cloud generated by LIDARPoint cloud generated by LIDAR

Classification of ground pointsClassification of ground points

General point cloudGeneral point cloud

Ground points Vegetation points

Process used to calculate the DEM

Point cloud generated byLIDARPoint cloud generated byLIDAR

Classification of ground pointsClassification of ground points

Generation of a mesh based on classified ground point

Generation of a mesh based on classified ground point

Interpolation of the mesh to obtain a matrix of elevation values

Interpolation of the mesh to obtain a matrix of elevation values

Automation of archaeological structures detection

In our case, we will focus on a smaller zone, in order to test the automatic detection process.

• Not efficient for structure detection, because the elevation variation due to the archaeological structures are very small compared to natural altitude variations in the studied area.

• Necessity to find another representation of the data.


• The hillshade function is the most commonly used in GIS software. The position of an artificial illumination source is arbitrary chosen by the operator, and the illumination values are calculated for each cell or the DEM.

• Archaeological structured shadows become then visible, but it introduces an error of position and the shape of each structures is distorted.


Creation of a Local Relief Model

Creation of the DEMCreation of the DEM



Creation of a 2DMoving average filter


This filter has to be adapted to the size of the element that we want to filter.

We chose a circular filter, with a diameter of 6 m, corresponding to the dimension of the elementary elements (hollows) of thesestructures

6 m





Filtering of the DEM by the filterFiltering of the DEM by the filter

All the elements corresponding to sharp relief variations have been removed, especially anthropogenic elements (roads, constructions, etc.)





Filtering of the DEM by the filterFiltering of the DEM by the filter

Calculation of the LRM by subtraction of the 2 previous elements

Calculation of the LRM by subtraction of the 2 previous elements


The obtained LRM represents the local relief variation compared to the global relief.

The black elements correspond to negative relative elevation, and the white elements to positive elevations.

This approach highlights elements characterized by sharp relief variations. These elements can be natural, but they are generally due to human activity

Gray level representation

3D representation

The difficulty is to indentify structures with different typologies(composed of a variable number of elements) and being more or less eroded.

Automatic detection and characteristic extraction

Definition, by the archaeologists,of the searched item





Automatic extraction of the item main properties


Dimension, ratio L/l, morphology

L

l






LRM binarizationLRM binarization






LRM binarizationLRM binarization

Detection and suppression of linearelements (roads)

Detection and suppression of linearelements (roads)


Detection of hollow parts of archaeological structures


Coupled detection using both :

Z level detectionImage correlation




Detection of all the elements corresponding to a negativealtitude ( hollow + corridor)


It includes hollow parts and corridors






Detection of all the elements corresponding to a positive

altitude (ridges)


altitude (ridges)







altitude (ridges)


altitude (ridges)

Fusion of all the elements to definethe complete structure

Fusion of all the elements to definethe complete structure


Extraction of sub images foreach complete structure





Creation of a database withall geometric characteristicsCreation of a database withall geometric characteristics

Conclusion – Further researchQuantitative interpretation(only for the hollow part : elementary structure)

Expert detection : 225 structures (Results obtained by E. Roussel)

� The efficiency of the automatic detection has been demonstrated.

� Only 9.3 % of the structure have been undetected, and 76 % of them correspond to eroded elements (some elements, very eroded and close to each other, have been detected as just one element).

� 94.7 % of erroneous detections correspond to other archaeological or anthropological features.

� Improvement of the algorithm efficiency (less erroneous detection).

� Application for other archaeological structures (example of charcoal kiln, easier to identify).

� Development of automatic detection and morphologic extraction directly from the 3D point cloud, and so with no modifications due to filtering.

Conclusion

Further research

Thank you very much for your patience !

Bibliography

� M. FLÓREZ, B. DOUSTEYSSIER, E. ROUSSEL, F. VAUTIER, J-P. TOUMAZET, M. LEFEUVRE, Y. MIRAS, L. RIEUTORT, 2013, El fenomeno de las «STRUCTURES EN PEIGNE» en el Macizo Central (Francia). Aportaciones a la caracterzacion de un paisajecultural de montana, 18th International Congress of Classical Archaeology - Merida - May 2013

� E. ROUSSEL, J-P. TOUMAZET, M. FLOREZ, F. VAUTIER, B. DOUSTEYSSIER, Using airborne LiDAR in geoarchaeologicalcontexts: Assessment of an automatic tool for the detection and the morphometric analysis of grazing archaeological structures (French Massif Central), European Geosciences Union, General Assembly 2014, Vienne (Austria) – 27 April – 2 May 2014

� M. FLOREZ, Y. MIRAS, Y. LLEROGO, A. BEAUGER, R. JULIA, S. RIERA MORA ; E. ROUSSEL, F. VAUTIER and J-P TOUMAZET, 2012, Genèse des paysages culturels de montagne sur la longue durée : Apport d’une démarche interdisciplinaire. Cas d’étude pyrénéens et auvergnats ,Congrès du réseau national des MSH– Caen – 6 et 7 décembre 2012.

� M. FLOREZ, J-P.TOUMAZET, E. ROUSSEL, F. VAUTIER, B. DOUSTEYSSIER, M. ABADI and L. RIEUTORT, 2013, Reconstructing the ancient Chaîne de Puys cultural landscape (Auvergne, France) through Landscape Archaeologie and airborne LIDAR data : the LIDARCHEO Projet ,XIXe colloque « Archeometrie » du GMPCA– Caen-2013

� M. LEFEUVRE, M. FLORÈZ, Y MIRAS, F VAUTIER., B. DOUSTEYSSIER, E. ROUSSEL, J.-P TOUMAZET. Genesis of a cultural landscape : the Chaîne des puys (Massif central, France)Colloque MINaH – Managing Inhabited Natural Heritage – Clermont-Ferrand – 11 – 13 septembre 2013

� F. VAUTIER, J-P. TOUMAZET, M. FLORES, E. ROUSSEL, M. FAURE, M. ABADI, B. DOUSTEYSSIER, Détection automatisée de vestiges archéologiques à partir d’images dérivées de nuages de points LiDAR, Rencontre internationale TRAIL 2014 : étude des données LiDAR (Light Detection and Ranging) sous couvert boisé – Frasne – 26 – 28 Mars 2014.

automatic detection and characteristics extraction of ... · this presentation deals with the...

Documents