Show bigger map

The entire site was scanned at ±5mm accuracy using a Leica HDS 3000 laser scanner accompanied by a Leica TCR407 Power reflectorless total-station (Figures 1 and 2). Although laser scanners and total-stations were providing 3D data, the regulations in Turkey required architects to generate 2D drawings out of them at a scale of 1/100.

Seeing and Doing
Cyclone software and Cloudworx is used for post-processing of point-clouds. During the processing stage architects encountered some difficulties in evaluating the 3D laser-scanning data in 2D. The most time-consuming task was drawing the site stone by stone. Looking at point-cloud data at a suitable level of zoom one can easily recognise objects on the site with the naked eye. So that the challenge arose of somehow mimicking the human eye and making the software automatically do the same. In order to do this we had first to figure out the way in which eyes recognise objects. There are two indicators here, one of which is change of colour. If change of colour occurs in a textured point-cloud it shows there is either another object or part of the same object in a different dimension, or a shadow. 

Another indicator for recognition of objects is change in the perception of depth relative to the plane of an object. Again, change in depth implies another object or part of the object present in another dimension.

Detecting Stones
A tool for the extraction of stones was developed by analysing differences in depth for each point, and we named it ‘Stone Wall Tool' (SWT). A mathematical model was created for digitally sensing, in other words automatically extracting, stones from the point-cloud data. SWT analyses the distance of each point to a specified plane relative to the object. It then detects and stores information about the point if this distance is less than for the preceding and subsequent points. These points make up the area of stones. A basic representation of this model is shown in Figure 3.

Setting Parameters
In order to use SWT some settings have to be applied to the data. Point-cloud data is first grouped ith respect to the orientation, using Cyclone software. For each group an average plane is then defined. Three points representing the average plane are selected from the point-cloud and the coordinate system set to a local coordinate system using these three points. The reference plane is then set to YZ plane, so that finally setting is achieved of an average reference plane parallel to the object. The three points are recorded to geometrically define the plane. The input and output data format of SWT is ‘.pts' format. SWT implementation of auto-detection is shown in Figure 4.

A tolerance value has to be put nto the equation in SWT to ensure that the user can select points in defined depths. This tolerance value is defined by the user according o the stonewall structure and scanning density of the point-loud data. Different parts of the point-cloud data may require he definition of differing tolerances.

In Conclusion
SWT is not a fully automatic solution for such problems. It is, however, a good start. And it certainly cuts down on a lot of working time.

Acknowledgments
Our thanks are due to System Computer and Technical Services Ltd, equipment providers, and all contributing members of KaleTakimi: Seddülbahir Fortresses Team.