Akademik Veri Yönetim Sistemi
Tezin Türü: Yüksek Lisans
Tezin Yürütüldüğü Kurum: Eskişehir Osmangazi Üniversitesi, FEN BİLİMLERİ ENSTİTÜSÜ, ELEKTRİK-ELEKTRONİK MÜHENDİSLİĞİ, Türkiye
Tezin Onay Tarihi: 2007
Tezin Dili: Türkçe
Öğrenci: NAZMİ ALPER KALE
Danışman: Nihat Adar
Özet:
In this work, the process of detecting human upper body parts in an image and assembling them into human figure is presented. In the process, human upper body is represented by six body parts: face, torso, upper-arms and hands. Possible locations and scales of human body parts in the image are detected by support vector machine based part detectors, and assemblies, which constitute candidates for human upper body, are constructed. The most likely candidate for human body configuration is determined via hidden Markov model and selected to represent human upper body figure. In regard to occluded cases of body parts, there are five hidden Markov models composed. Models are congregated in a decision process and body parts that represent human body figure in occluded cases can also be determined successfully with the definition of the occluded case. The detection method shows promising results when tested on images from MIT pedestrian database and additional pictures that contain pedestrians. The choice of using hidden Markov model in assembling body parts into human figure is verified by comparing the results with those of obtained from support vector machines and multivariate Gaussian distribution. Keywords: Human body-part detection, posture recognition, Hidden Markov model