Xth Anniversary Racurs Conference
From Imagery to Map

Italy was the stage for this year's Racurs Conference. Gaeta, an ancient village near Naples was the location where participants could enjoy the international scientific and technical conference organized by Racurs. This year's conference was visited by 100 specialists in the field of photogrammetry and remote sensing, from 19 different countries, making this a truly international event. A three-day program of presentations and workshops was complemented with social events, such as an excursion to Naples and Pompeii, as well as a sports event and gala dinner.

by Eric van Rees

The first two days of the conference were meant for presentations of different sorts, grouped by themes such as regional and corporate projects, digital cameras and aerial equipment. This report features a selection of topics discussed during the presentations.

Spatial Data Infrastructure in Italy
Renzo Carlucci (editor of GEOmedia, an Italian publication) made an interesting comparison between the national infrastructure of geospatial data in both Italy and the US. Although this comparison is not equal for a number of reasons, it became clear that Italy lacks a central body such as the FGDC (Federal Geographic Data Committee) in the US. Whereas the US does not yet have a national GSDI, Italy is involved in the INSPIRE program, but the situation between state and regions has suffered a lot in the past. Italy's current situation can be described as a transition from the traditional cartographic school to the new digital spatial infrastructure. Carlucci noticed a lack of interest among politicians in establishing a master’s degree program in geomatics at Italian universities, with the result that the job market is suffering as there are too few young employees to fill the positions available.

10 Years of Racurs Conference
Continuing this series of contemplative presentations was an address by Victor Adrov, Managing Director at Racurs, who focused on the past Racurs Conferences. His recollections held many good memories of previous locations and particular events. His speech showed that this is a special sort of conference, since a lot of visitors are regulars or 'family', as one visitor told me during a coffee break. Adrov mentioned the increasing number of specialists that participate in the conference, as well as the still growing user base of PHOTOMOD, now at around 5000.

Adrov's presentation followed a similar one by Gottfried Konecny on 100 years of ISPRS, which was in fact something of a university course on the scientific development of photogrammetry. Not only was the technology covered but also its practitioners, and even its political context.

Topographic Map Update Issues
Konecny's second presentation on the first conference day focused on topographic map update issues. He compared the costs of orthophoto mapping and line mapping by outlining all the various costs involved in producing an orthophoto map: aerial photography, optional scanning of aerial photos, aerial triangulation, creating a digital elevation model via image correlation, digital orthophoto production and vectorization in both 2D and 3D. A comparison between orthophoto mapping and line mapping showed that orthophoto mapping is at least three to five times cheaper than line mapping, and as such can serve as a map substitute. But, since analysis capacities by a GIS are required, one tends to opt for line mapping. Therefore, attempts are made towards automatic feature extraction, but, since this is not possible to 100% accuracy, statistically controlled partial automation is desired. According to the speaker, at this stage it can save 2/3 of the manual effort.

After this statement, Konecny analyzed different scenarios in different parts of the world where Land Management takes place: whether or not a base map is used and/or updated and the technology that is used to do this. For instance, in African countries, satellite imagery is used to map and update slums that are not on cadastral maps since they are not legal and thus not on any official maps.

3D/4D City Modeling
Armin Gruen is a new face among the Racurs conference participants and he delivered two presentations on the first conference day. '3D/4D City Modeling: progress and prospects' was the title of the first one, which opened with a comment on the name of the conference ('From imagery to map:digital photogremmetric technologies'), questioning the use of the map. This has everything to do with the topic of modeling, which nowadays can be seen as a substitute for a map, at least in terms of an end product.


Armin Gruen during his presentation

Gruen is associated with the Institute of Conservation and Building Research of the ETH Zurich, where a lot of research on city modeling takes place. Known from the CAD world is the concept of BIM (Building Information Models), which comprises geometry, topology, semantics, the appearance of individual buildings and also whole cities. These are used for the analysis of life cycles of building stocks and flows, which means the change of models over time, by updating the models.

Gruen devoted some time in his presentation to CyberCity-Modeler, a methodology and software for 'semi-automated object extraction and modeling of built-up environments using imagery from satellite, aerial and terrestrial platforms'. Nowadays these city models are used by a wide range of industry professionals (such as city planners, for environmental studies) and Gruen states that interesting markets are expected in the entertainment and infotainment industries (video games, movies).

Reality-based 3D city modeling involves the following data acquisition techniques: scanning of maps, laserscanning from airborne or terrestrial platforms and photogrammetry (satellite, aerial and terrestrial). Combinations of all three are also possible. Depending on a number of criteria, a 3D city model may be done 'quick and dirty' or can have a high quality level in terms of accuracy, topological correctness or completeness. It's easy to cheat, especially with textures.

Texturing Matters
Gruen is critical to the notion of 'people who claim you can do automated 3D city models with high quality'. A new initiative is Google Building Maker for creating buildings online, but it cannot do what professionals expect from it. Some examples of new initiatives show his point: from a distance, a lot of 3D buildings look okay, but when you zoom in, you see that details are wrong, for instance facades that are copied incorrectly to cover the whole of the outside of a building. 

Since 1989, photogrammetry has been used for automatic building extraction. This process consists of three steps: detection (recognition of a building), reconstruction (generation of a 3D description at a desired level of detail) and attribution (assigning of descriptive elements, like the type and usage of a building).  As for the current state of research, Gruen stated that there are many automatic procedures, but they are not usable in practice. As for semi-automatic procedures, there are not very many of these and system manufacturers are waiting in line for them. The main problems concerning automated reconstruction have to do with image interpretation, the automated control of the level of detail and the correct topology. Since there is currently no progress in automated image interpretation, a way out of this situation may be a multiple sensor/data approach, by developing semi-automated methods. An example of this is semi-automated facade texturing by using oblique aerial images. In concluding his presentation, Gruen hinted at his second presentation that day, namely by mentioning helicopters and unmanned vehicles for photogrammetric recordings.

PHOTOMOD
As to be expected, a large part of the conference was devoted to PHOTOMOD, the flagship product developed by Racurs. The latest version 5.0 was discussed in detail having been originally presented last year at the 2009 conference. A number of new features were given particular attention, as well as the company's road map for future releases.

Dmitry Kochergin of Racurs spoke about PHOTOMOD today and tomorrow: looking back at the release of version 5.0, he stated several of the premises for this release at that time: requirements for project format optimization for tens or hundreds of thousands of images per block, a need for more open resource system and data formats, optimized for network distributed processing and a need for workflow change, eliminating rigid sequence of processing stages.

The first release of version 5.0 included a number of highlights, such as no limits for image size, full 16-bit raster support, 'dynamic' rebuilding of TIN while editing, a new user-friendly interface of tie point manual measurements and direct working with the source images.

The main changes from last year are automatic DTM creation algorithm improvements, new powerful DTM filters ('removing' buildings and vegetation), and an increase (2x) in the speed of the automatic AT process. In addition, high quality smooth contour lines, more distributed processing (adding images, orthomasaicking, DTM, AAT) and VisionMap A3 sensor support have also seen improvements. Smaller changes include self-calibration procedure enhancements, more start approaches for bundle adjustments, speeding up raster visualization and RapidEye sensor support, among others.

In a couple of months, version 5.1 is expected, which will include 'dense DTM' creation and filtering, a new autoseams algorithm in PHOTOMOD Mosaic, Lidar data conversion to the DEM of unlimited size and 3-DAS-1 sensor support. For the future, the company promises final integration of all systems modules in one environment, a full 64-bit system version, feature-based correlator modifications, DSM and True-Ortho improvements, Lidar-oriented module developing, full multi-sensor triangulation and more. 
After this look into the future of PHOTOMOD, more user-oriented presentations followed, such as 'Effective techniques for working in PHOTOMOD by Alexandra Kiseleva of Racurs. 

The second conference day started with a series of project presentations by PHOTOMOD users. An example was a project on the mapping of biodiversity in the Aspromonte National Park (Italy) where PHOTOMOD was used. In Greece, photogrammetry is taught with PHOTOMOD in the National Technical University of Athens.

Other presentations during the second day were held by Andrey Shumakov (GeoEye USA), Pawel Ziemba (DigitalGlobe UK), Olga Gershenzon (ScanEx), Milana Elerdova and Olga Kolesnikova (both from Sovzond) and Yury Baranov (Gazprom VNIIGAZ), among others. Topics ranged from development of on-line services based on space imagery data and web technology, creating regional geoportals using earth remote sensing data, and the generation of digital elevation models and detection of surface movements via space radar interferometry.


All participants in this year's Racurs Conference

The third conference day consisted of three master classes and business meetings. Participants were shown new algorithms for automatic DTM creation in PHOTOMOD 5.0, how to set up data storages, adding imagery and distributed processing, as well as what's new in the software package PHOTOMOD radar.

Internet: www.racurs.ru