Would you like to brief us what all geospatial activities are undertaken at Future Cities Laboratory?
The Future Cities Laboratory consists of 10 Research Modules plus 3 Assistant Professorships. All groups need spatially-referenced data of various kinds and in various formats. It is very obvious that good and actual geospatial data plays a major role in the planning, design, management and further development of the smart city concept. Our job is, besides focusing on our own research agenda, to generate this data which may not be available on the market and to manage a GIS platform. There is a clear trend towards the need of even more data. The Big Data issue is a point of discussion also among city planners.
Smart cities/green cities/sustainable development are becoming trends worldwide. Do you think photogrammetry can play a big role to support this trend?
Of course not every kind of data can be generated photogrammetrically, but many can. Whenever it comes to actual, complete and accurate/reliable data, photogrammetry may play a significant role. With the ever increasing number of platforms and sensors, like on satellites, aerial platforms including UAVs, static terrestrial and Mobile Mapping Systems our capabilities for raw data acquisition have increased lately tremendously. The main problem today, is in data processing, where automation is still only possible at a very rudimentary level. In this context there is the danger that incompetent personnel create results of low quality, which is set to ruin the good reputation of photogrammetry. Photogrammetry is neither black box nor is it a secret technology. But there are many possible pitfalls.
Your research activities include generation of 4D city models for smart city application. Tell us more.
Naturally 3D/4D city modelling plays a major role in our project, with raw data coming from satellite images, UAV and terrestrial images and laser-scanners (Mobile Mapping Systems). Singapore is a special case in the sense that standard aerial imagery is classified and not available for city modelling. So we were using IKONOS and WorldView-2 stereo models to generate 3D models of Rochor (“Little India”) and Punggol, a new quarter built for a population of 360,000.
We were permitted to perform a UAV flight over the campus of NUS (National University of Singapore) at a very high image resolution of 5 cm. From those ca. 850 images we derived a textured 3D model, including DTM, buildings and other man-made objects, like for instance 900 street lights. We also measured and modelled 2,000 trees.
In order to get under the tropical tree canopy for DTM data and to record the facades we employed also a Mobile Mapping System with 2 laser-scanners (resolution 1-5 cm) and video image sequences. In addition we took terrestrial images. This creates a unique hybrid dataset, which is being used for research and development, testing and applications as well.
Singapore was badly-affected by haze a few months back. Haze, cloud cover, and bad weather are very common in Southeast Asia. Would you agree that these factors also make remote sensing a little bit unpopular in this region?
No. Actually satellite remote sensing is used to detect the sources of haze (forest fires) and contributes to the search for the responsible polluter. A problem in the tropics is the many clouds, obstructing very often the view of optical systems onto the ground. This is a reason why radar systems are becoming very popular in Southeast Asia. But with the many optical systems planned for the near future the chances of getting a better view onto the ground are also improving.
Based on your experience working in Southeast Asia, which country in this region do you think has made optimal use of remote sensing technology?
The three major forces in Asian remote sensing: China, Japan and India do not belong to Southeast Asia. But Southeast Asian countries like Indonesia, Thailand and Vietnam make big efforts to catch up. All the others are a bit behind the international standard.
Anything you would like to see more in Southeast Asia remote sensing scene?
There are already many and will be even more sensors deployed in space, with better performance and ever increasing resolution. But from the raw image data to value-added products is a long and sometimes thorny way. Here it needs more engagement, more R&D activities.
Especially in photogrammetry I see a lack of well-educated experts. Asian students seem to lean nowadays more towards remote sensing and GIS. This may prove a big problem in the near future, because the capabilities and skills to derive 3D models from raw image and point cloud data will be required even more in the future.
I understand you also lecture in various universities in Asia. Do you think the geospatial resources produced in this region are internationally competitive?
When we talk about resources we must also address hard- and software products. It is obvious that only very few Asian products are successful on the international market. China for instance has a strong national market, but no impact internationally.
On the human resources side, the sheer number of students, for instance in China, has a positive impact. But despite those large numbers, I am told that Chinese graduates do not have any problems in finding good jobs. In India it is said that there is a lack of graduates with background in Geomatics.
In general the level of education and research in Geomatics is still trailing behind in Asia compared to Europe. Europe has a longstanding tradition in the field which will not so quickly be reached or overtaken. There are reasons for that, e.g. the lack of a truly international perspective in hiring faculty and recruiting the best possible students in Asia.
There are places where a Geomatics curriculum does not exist at all (e.g. in Singapore), or where Geomatics is just part of other curricula, like Civil Engineering. In such cases there will be a lack of competitiveness.
There has been explosion of different sensors and platforms available in market today. And we see a lot of new companies emerged to exploit these markets (i.e. UAV). Do you think this is a healthy development for the industry in a long run?
If we look back at the last 40-50 years, the development and employment of sensors has always been faster than the related processing algorithms and software. Our field is clearly sensor-driven (compare radar, laser, CCD/CMOS cameras, etc.). New applications will result from new sensor and processing capabilities. This is very obvious in the earth observation area where the number of satellite missions are almost too numerous to be counted. The processing capabilities on the other hand are trailing behind. The increase in platforms and sensors is surely a good trend for the user. He will have more options in selecting a specific sensor or system, and, hopefully, the prices for raw data will drop. This applies in particular to the UAV market, which is dominated by SMEs.
On the other hand, UAV photogrammetry opens up a whole new world of possibilities for the geospatial industry that we didn”t have before. What do you visualise for the future of this technology?
Yes, in my understanding UAVs are not here to replace existing tasks, like standard mapping projects, but they open up possibilities for totally new applications in areas where we could not work before. In Singapore we are discussing currently the use of UAVs at the Port Singapore to locate corrosion and cracks on cranes and make deformation measurements, to detect stagnant water pools on building roofs where Dengue fever mosquitos reside, to track oil spills at sea, to model a Botanical Garden including recording of the quality of the leafs of trees, to produce 3D temperature models of buildings, etc.
Technically, the UAVs will develop further. We should note that most existing UAVs were originally not built for photogrammetric applications. They were meant to go up, take a few pictures of an object and come back again. In photogrammetry we have many additional requirements which most existing UAVs do not fulfil. There is much room for improvement. Just a few items: Longer flying times, to produce large image sets efficiently; integration of image- and range-based sensors; calibration of sensors and system; image-based navigation; more accurate direct geo-referencing; and of course the large area of improved data processing functions.
Also, current UAVs are usually dumb. We need much more intelligence on-board to make UAVs safer, more flexible and powerful. Some of this work is already done in the robotics domain, but mostly only under laboratory conditions.
In general, system components in hard- and software are not robust enough and system failures result too often.
However, the key issue remains: Getting permissions to fly. In many countries this is a big problem of its own. Whenever a new technology becomes available, misconceptions and exaggerations are part of the business. This can currently be observed in the UAV domain and in city modelling. Both UAVs and 3D city models are around for quite a long time. Model helicopters and airplanes equipped with cameras have been used in the late 70s already, R&D in 3D building and city modelling goes back to the late 60s. Enough time has passed to acquire a realistic attitude towards these technologies.
There is some competitive vibe between LiDAR and photogrammetry. What are your views on this?
LiDAR and Photogrammetry are clearly complementary. In terms of accuracy photogrammetry is usually better in planimetry than in height. With LiDAR it is the other way around. Photogrammetry has the distinct advantage that images include much more information than a point cloud. But this information is only implicitly existent and has to be extracted by suitable algorithms. The attraction of LiDAR is that the point clouds are readily available. But one should realise that point clouds are usually not the final result and the generation of complex 3D models from point clouds can also be a demanding and time-consuming task. In simple words, photogrammetry is better on edges while LIDAR is better on smooth surfaces. Through the use of a wide range of possible focal lenses and flying heights, photogrammetry is more flexible in terms of varying resolution. Also, image data acquisition is considered less expensive than the use of laser-scanners. This is very obvious in terrestrial applications, especially when work in remote and inaccessible areas is to be done.
So, again, if the financial situation allows for it, the simultaneous use of both sensors is preferable. However, there is still a lack of concepts and algorithm for truly integrated data processing.