Projects / Programmes
Morphological operators for pattern recognition in large point-clouds
| Code |
Science |
Field |
Subfield |
| 2.07.03 |
Engineering sciences and technologies |
Computer science and informatics |
Programming technologies - software |
| Code |
Science |
Field |
| T120 |
Technological sciences |
Systems engineering, computer technology |
| Code |
Science |
Field |
| 1.02 |
Natural Sciences |
Computer and information sciences |
mathematical morphology, pattern recognition, remote sensing, event recognition
Researchers (12)
| no. |
Code |
Name and surname |
Research area |
Role |
Period |
No. of publicationsNo. of publications |
| 1. |
37956 |
PhD Marko Bizjak |
Computer science and informatics |
Researcher |
2015 - 2016 |
0 |
| 2. |
33285 |
PhD Simon Gangl |
Computer science and informatics |
Junior researcher |
2013 - 2014 |
0 |
| 3. |
36815 |
Denis Kolednik |
Computer science and informatics |
Researcher |
2014 - 2016 |
0 |
| 4. |
33709 |
PhD Niko Lukač |
Computer science and informatics |
Junior researcher |
2013 - 2016 |
0 |
| 5. |
29243 |
PhD Domen Mongus |
Computer science and informatics |
Researcher |
2013 - 2016 |
0 |
| 6. |
05892 |
PhD Dalibor Radovan |
Geodesy |
Researcher |
2013 - 2016 |
0 |
| 7. |
28150 |
Blaž Repnik |
Computer science and informatics |
Researcher |
2013 - 2016 |
0 |
| 8. |
08638 |
PhD Krista Rizman Žalik |
Computer science and informatics |
Researcher |
2013 - 2016 |
0 |
| 9. |
26035 |
PhD Denis Špelič |
Computer science and informatics |
Researcher |
2013 - 2016 |
0 |
| 10. |
23564 |
PhD Mihaela Triglav Čekada |
Geodesy |
Researcher |
2013 - 2016 |
0 |
| 11. |
24314 |
PhD Tomaž Žagar |
Geodesy |
Researcher |
2013 - 2016 |
0 |
| 12. |
06671 |
PhD Borut Žalik |
Computer science and informatics |
Head |
2013 - 2016 |
0 |
Organisations (2)
Abstract
The advanced technologies of laser scanning with their accuracy, speed and resolution, have revolutionized the field of Earth observation. The amount of information contained within 3D point clouds has introduced the recognition of geometric structure as the most important computational challenge of this decade. Developing new solutions requires coping with irregular point distribution, the lack of topology and their sheer size that often exceeds the capabilities of modern computer systems. Using the known concepts that were developed for pattern recognition in raster data has often proven to lead to inefficient algorithms that require intensive user interaction and additional information about the geographical areas.
The proposed project aims to research a new methodology for recognizing 3D geometrical structures, monitoring their kinematics, and detecting events within large point clouds by applying contemporary findings of mathematical morphology. Although, mathematical morphology is considered to be a relatively young mathematical theory, its quantitative definition of arithmetic and description of shapes offers great expressional strength. Morphological operators are derived from the set theory and extended by using the concepts of geometry, topology, probability, and statistics, and are completely adapted for digital and parallel processing. The recently developed algebraic formalization of scanning morphology offer a spatially-dependent, selective, and completely automatic adaptation to the geometrical structures of input data. These theoretical foundations offer the possibility of developing an efficient pattern recognition methodology, where adaptation to the temporal domain would allow a quantitative presentation of events and a description of their kinematics. The efficiency of the developed method shall be demonstrated with by two scenarios: (i) recognition of geomorphological process kinematics and (ii) monitoring tree development in Slovenia. For the purpose of recognizing the kinematics of geomorphological changes (such as landslides) it is intended to develop an automatic method for ground recognition within 3D point clouds, and the construction of a digital terrain model that would be more accurate and time efficient by eliminating the need for the users to set parameters. Such a procedure would allow for the detection of changes in the terrain and evaluate the volume, mass and speed of moving earth masses over large geographical areas (whole of Slovenia) with high resolution (under 0.5m) and accuracy (over 90%). Similar accuracy can be expected regarding (ii) monitoring tree development, where a new method for recognizing single trees shall be developed. This method would estimate the number of trees within a respected area and provide the geographical positions, heights and volumes of tree-crowns. It would measure growth of a single tree, forest biomass growth by multi-temporal data acquisition, and develop a predictive simulation of their development.
The precision of the proposed uses would be tested by field measurements, while the construction of a digital terrain model of Slovenia will demonstrate their computational efficiencies. In this way, the national project for surface scanning of Slovenia with LiDAR technology would be supported directly. The results of our research will be published in the most distinguished international journals, and regularly presented to the Slovenian public by organizing symposiums and workshops. The projects results shall also promoted abroad by attending international conferences.
Significance for science
In the scope of the project we developed a new methodology for processing LiDAR data, where we extracted geomorphological features of recognized terrain, vegetation and other objects. This was performed using new concepts of mathematical morphology (i.e. mathematical differential profiles) and context-dependent analysis. We introduced a new methodology for local fitting of surfaces, which particularly improved the accuracy of classification within point clouds. The methodology was successfully applied to two use cases: monitoring geomorphological variations (simulation and landslide prediction) and vegetation growth in time-varying point clouds. The developed methodologies enable new opportunities in the fields of computer science, remote sensing data processing, geoinformatics and environmental sciences. The presented research was published in prestigious journals with impact factor (three publications categorised as A'' by SRA), which brought us international attention and reputation. Papers, that are a direct result of the project, have been cited over 70 times already before the end of the project. Furthermore, a US patent was granted to us. We were awarded for these scientific achievements as well (Information Society Slovenia award and Danubis young scientist), which improved the international reputation of the presented research further.
Significance for the country
In the scope of the project we built the software gLiDAR, which includes all scientifically-supported algorithms that were developed during the project. This software was used to classify the entire point cloud of Slovenia (approximately 45 TB) as buildings, terrain or vegetation. Furthermore, a digital relief model was created with a spatial resolution of 1 m2. With the products that were obtained using the point cloud of Slovenia and mostly made with gLiDAR, it is possible to acquire river systems, road infrastructure, real estate cadastre, various ecosystems, etc. This demonstrates that the social impact of our research affects many other fields (i.e. agriculture, forestry, environmental protection, transport, archaeology, etc.). The given data and related products are available on SEA portal (http://gis.arso.gov.si/evode/profile.aspx?id=atlas_voda_Lidar@Arso). The results of high accuracy, according to ISPRS, led to a notable media attention to a researchers of the project. This brought additional promotion of the project in the media (translated): RTV SLO: Let’s bite science – Laser scanning (24.10.2013), 24UR: LiDAR data are significantly more accurate, there are plenty of potential users (21.01.2015), Slovenian Press Agency: In Maribor, the most accurate algorithm for terrain recognition in the world was developed (08.01.2015). Val 202: There is no boundary between researching and life! (11.01.2015), 24UR: At Maribor’s faculty, the most accurate algorithms for terrain detection were developed (09.01.2015), 24UR Videoteka: Maribor’s innovators created an incredibly accurate algorithm, TV Maribor: Algorithm for terrain recognition (12.01.2015). The workshops for extensive training of professional LiDAR data users and products that are a result of the methods, developed in the scope of the project, were carried out (translated): Geodetic institute of Slovenia: Workshop 06.05.2015), Geodetic institute of Slovenia: LiDAR workshop (21.04.2015), Geodetic institute of Slovenia: Workshop (18.06.2015).
Most important scientific results
Annual report
2013,
2014,
2015,
final report
Most important socioeconomically and culturally relevant results
Annual report
2013,
2014,
2015,
final report
