2. MODULES OF LANDSAFE 
2.1 LandSAfe Architecture 
LandSAfe is based on a web-based client/server model: the 
system providing the Product Generation Modules and Data 
Store is running on a server which is accessible through a 
graphical user interface (UIF) running in a web browser. The 
UIF provides the Commander functionality to execute the 
Product Generation Modules and to provide the input 
parameters required for processing. Furthermore, the UIF 
provides viewing and browsing capabilities for mapping and 
derived products. 
The LandSAfe UIF Commander functionality is based on a 
Service Oriented Architecture (SOA) using web services. The 
LandSAfe core is embedded in a Web Processing Service 
(WPS) using the open-source PyWPS solution which is an 
implementation of the Web Processing Service standard from 
the Open Geospatial Consortium (OGC). The LandSAfe UIF 
Commander generates processing requests and sends those 
requests over the internet (http/soap) to the WPS service which 
executes the request by running Python scripts from the 
LandSAfe core. 
The LandSAfe UIF Viewer is based on open-source JavaScript 
libraries like OpenLayers, GeoExt and ExtJS. Those libraries 
are also based on SOA architecture and allow the client side, 
i.e. the web browser, to access Web Map Services (WMS), Web 
Feature Services (WFS) and Web Coverage Services (WCS). 
Such services are implemented using the open-source 
GeoServer which is the reference implementation of the Open 
Geospatial Consortium (OGC) for the WFS, WCS and WMS. 
On the server side the LandSAfe core is comprised of the 
Product Generation Modules which generate the DTMs and 
derived products like boulder and crater detection, risk maps 
and illumination maps. The Product Generation Modules are 
described in the following sections. Moreover, the Data Store 
provides access to the internally generated mapping and derived 
products. Additionally, the Data Store also provides access to 
external product databases like NASA PDS (Planetary Data 
System) (McMahon 1996) and ESA PSA (Planetary Science 
Archive) which contain for example the input images and the 
navigation data. 
2.2 DTM Generation 
The DTM generation module is responsible for the automatic 
derivation of high resolution DTMs by means of stereoscopic 
digital image matching techniques. This module is based on the 
ISIS system (Anderson et al. 2004) of the USGS and the Ames 
Stereo Pipeline (Moratto et al. 2010) developed by NASA. ISIS 
provides image processing tools especially developed for 
planetary imagery. It also includes a collection of camera 
models of many NASA planetary missions and a database with 
navigation data (spacecraft position and camera pointing) for 
those missions via SPICE kernels (Acton 1996). The Ames 
Stereo Pipeline (ASP) complements ISIS by providing 
algorithms for dense digital image matching. 
Python scripts implement the automatic processing chain 
making use of ISIS tools and interfaces and the ASP for dense 
matching. The processing chain is comprised of three steps: 
image ingestion, bundle adjustment, DTM matching. Image 
ingestion includes data import from PDS, radiometric 
calibration and SPICE kernel initialization. Bundle adjustment 
aims at improving the SPICE values for spacecraft position and 
camera pointing and includes tie point matching, measuring 
ground control points (GCPs) and calculating the bundle 
adjustment. The tie points improve the relative accuracy of the 
image strips, pairs or image blocks comprised of several image 
strips and the GCPs tie the images to the body fixed coordinate 
system. The latest and most accurate control net on the Moon is 
the ULCN 2005 (Archinal et al. 2006) which is not applicable 
in our case as its point density is too low. Therefore, we tie our 
image blocks to the GLD100 (Scholten et al. 2012) dataset and 
to the LOLA dataset at the poles. While the matching results 
from LROC NAC imagery yield high resolution height values 
the LOLA dataset can fill holes in low texture or shadowy 
areas. A detailed example of a result of the processing chain is 
described in section 3. 
2.3 Boulder detection 
The boulder detection module allows the automatic detection of 
boulders and the estimation of their abundance on the landing 
site area using size/frequency distribution functions. The 
detection is made on high resolution LROC NAC imagery 
applying a maximum entropy thresholding (MET) method 
(Kapur et al. 1985) (s. Fig. 1). Then, boulders are measured and 
counted to produce a boulder size-frequency histogram for the 
target zone. 
The algorithm for boulder detection has to be adapted to the 
low sun elevation angles typical of polar regions (like on 
Shackleton Rim). A zonal application of the MET algorithm 
and image texture processing methods are also foreseen to 
improve the procedure. If the availability of data allows it, the 
detection is repeated on several images of the same area taken 
at different illumination conditions. 
For the size measurement of boulders, the module also makes 
use of the high resolution DTM in order to extract the height of 
the biggest boulders. If the slope (derivable from this DTM) is 
also known this height can also be extracted from the length of 
shadows measured on the images. 
2.4 Crater Detection 
Craters are automatically detected, measured and counted on 
the same imagery used for boulder detection. The Canny edge- 
detector is firstly applied (Troglio et al. 2010) in order to obtain 
edge images on which the Hough Transform (Sawabe et al. 
2006) can detect the circular shapes of craters. The edge image 
is first cleaned up by deleting isolated edges (i.e. edges not 
linked to crater outlines) using a directional gradient based 
technique. Successive ranges of crater size are then detected 
using the Hough Transform and the edge image is updated at 
each step to remove edges corresponding to craters already 
detected. As described for boulders, crater detection is also 
repeated on several images covering the same area if available. 
This crater detection on images is coupled with a DTM-based 
approach and all the detected craters are compared in position 
and radius in order to eliminate duplicates. Craters are counted 
in each range of radius to produce a crater size-frequency 
histogram for the target zone. 
474