The Image Processing and Pattern Recognition Laboratory is used extensively by faculty and graduate students who do research in the area. Example studies include processing of remote sensing imagery, 3D object modelling, document image processing, content-based image retrieval and bio-medical pattern recognition applications.
Research projects are supported by METU, TUBITAK and various other resources (DPT, State-Planning Agency, Secretariat of Defense Industries), while the international collaborations are also granted by various agencies (NSF, Paris University and others).
Current Projects
HASAT
In this project, remote sensing imagery is analyzed and automatic detection/classification of objects/lands is aimed. There are several groups whose target objects vary in the project. Advisors of the groups are from different departments (Computer Eng., Electric Electronic Eng., GGIT, Informatic Institute). Fatos Yarman-Vural's groups have developed techniques to detect airports, buildings, harbours etc during last year. Nese Yalabik and Sebnem Duzgun's group developed a land use-land cover classification process. HASAT project is conducted in collaboration with HAVELSAN.
Previous Projects
Incorporating Prior and Contextual Knowledge into Computer Vision Modules
Sibel Tari's group is investigating contextual influences in early and mid level vision. Within the last two years, the group developed systematic methods to integrate prior and domain knowledge into perceptual modules.
YAHTS
The objective of this project to create a software system for physicians for supplying accurate and practical ways to diagnose and further classify a musculoskeletal disease using gait analysis methods. Gait analysis can be defined as the numerical and graphical representation of the mechanical measurements of human walking patterns and is manually performed by expert physicians for the non-automated and automated diagnosis of various abnormalities and diseases. The data used in this project are collected in Ankara University Faculty of Medicine Gait Laboratory .
Interactive Classification of Satelite Image Content by Active Learning
Interactive classification is an attractive alternative and complementary for to automatic classification of satellite image content,since the subject is visual and there are no yet powerful computational features corresponding to the sought visual features yet. In this project, we are building a software system with capabilities for interactive classification of the content of satellite images. The system allows the user to indicate a few image regions that contains a specific geographical object, for example, a bridge, and to retrieve similar objects on the same satellite images. Retrieval process is iterative in the sense that the user guides the classification procedure by interaction and visual observation of the results. The classification procedure is based on a one-class classifier.
Multimedia Retrieval and Querying Systems for E-Government
Fatos Yarman-Vural's group has recently developed a retrieval system that uses data as web pages from all over the Internet. This retrieval system is based on both visual and textual information in the web pages. In general sense, the group built up an image search engine that works on the semantic information in the images (human faces). The application of this system is to search similar images of a query image from images of web pages.
Recent Publications
-
Barış Yüksel, Çağlar Şenaras, Mete Özay, Fatoş Yarman-Vural, “Yığılmış Genelleme Yöntemiyle Uydu Görüntülerinden Otomatik Bina Bulunması,” SIU 2011, Antalya, Turkey.
-
Erşan Demircioğlu, Fatoş Tünay Yarman Vural, “Otomatik Görüntü Etiketleme Sistemleri için Yeni Bir İyileştirme Yöntemi,” SIU 2011, Antalya, Turkey.
-
Ulya Bayram, Gülcan Can, Barış Yüksel, Şebnem Düzgün, Neşe Yalabık, “Multispektral Uydu Görüntülerinde Eğitmensiz Arazi Örtüsü ve Kullanımı Sınıflandırması,” SIU 2011, Antalya, Turkey.
-
Ulya Bayram, Gülcan Can, Şebnem Düzgün, Neşe Yalabık, ” Evaluation of Textural Features for Multispectral Images,” SPIE Remote Sensing Conference, 2011.
Selected Publications
-
Emre Akbas, Fatos T. Yarman-Vural, “Design of a Feature Set for Face Recognition Problem,” ISCIS, 2006.
-
Nafiz Arica, Fatos T. Yarman-Vural, “Shape Similarity Measurement for Boundary Based Features,” ICIAR, 2005.
-
Mutlu Uysal, Fatos T. Yarman-Vural: ORF-NT: An Object-Based Image Retrieval Framework Using Neighborhood Trees. ISCIS, 2005.
-
E. Erdem, A. Sancar, S. Tari: Mumford-Shah Regularizer with Spatial Coherence, Scale Space and Variational Level Set Methods, 2007.
-
C. Aslan and S. Tari, An Axis Based Representation for Recognition, ICCV, 2005.
-
N. S. Koktas, N. Yalabik, G. Yavuzer, “Combining Neural Networks for Gait Classification”, The 11th Iberoamerican Congress on Pattern Recognition (CIARP 2006), Mexico, Cancun, 2006.
-
N. S. Koktas, N. Yalabik, G. Yavuzer, “Ensemble Classifiers for Medical Diagnosis of Knee Osteoarthritis Using Gait Data”, The Fifth International Conference on Machine Learning and Applications (ICMLA 2006), Orlando, Florida, USA, 2006.
-
O. Dalay and V. Atalay, “Interactive classification of satellite image content based on query by example”, ISPRS Spatial Data Mining Workshop, Ankara, Turkey, 24-25 November, 2005.
-
R. Hassanpour and V. Atalay, “An Experimental Study on the Sensitivity of Auto-Calibration to Projective Camera Models Parameters”, Optical Engineering, Vol. 45, 047002, 2006.
-
M. Musa, D. de Ridder, R. Duin and V. Atalay, “Almost Autonomous Training of Mixtures of Principal Component Analyzers”, Pattern Recognition Letters, vol.25, No.9, pp.1085-1095, 2004.