Kinect verisi ile 3-boyutlu slam uygulamasında 3-boyutlu yüzey kavis özellikleri kullanımı analizi.


Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2014

Tezin Dili: İngilizce

Öğrenci: Ömer Faruk Adil

Danışman: İLKAY ULUSOY

Özet:

The introduction of affordable and sufficiently accurate range sensors such as TOF cameras, laser scanners and RGBD cameras has significantly contributed to the robotic applications like SLAM. Although range sensors are used extensively and successfully in SLAM applications, there is still room for improvement in the sensor data utilization process. In this thesis a novel approach for the feature extraction part of 3D SLAM is introduced. The use of compact surface curvature features in the SLAM algorithm is proposed which will appear for the first time in the literature. The proposed method uses mean and Gaussian curvature calculations to extract curvedness features from RGBD output of Microsoft Kinect sensor. The extracted features are then fed to the SLAM algorithm to be used in the global data association process. The results are compared to the state-of-the art feature extraction techniques for SLAM, namely plane features, SURF features and corner features on real Kinect dataset sequences which are conventionally used as benchmark for SLAM algorithms.