Yazar "Sen, Mehmed Oguz" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • Küçük Resim Yok
    Öğe
    Artificial Immune System Optimization Based Duplex Kinect Skeleton Fusion
    (Ieee, 2017) Gunduz, Ali Fatih; Sen, Mehmed Oguz; Karci, Ali; Yeroglu, Celaleddin
    Human motion tracking, which requires both motion sensing hardware and algorithms based on computer vision, is an enjoyable and active research area with diverse applications. As a depth sensor device Kinect is a famous hardware component for this task. In this work, we studied using more than one Kinect camera to obtain better motion tracking which is applicable for motion capture. We synthetically created two camera data from one and then focused on de-noising and fusing these data in order to obtain more realistic skeleton joint coordinates. Artificial Immune System (AIS) optimization algorithm is suggested and used for this task. As a result we obtained 30% better fusion results from noisy synthetic data. Our results showed that AIS is a promising algorithm for obtaining optimal joint coordinates in the fusion of multiple Kinect skeleton data.
  • Küçük Resim Yok
    Öğe
    Comparison of Trajectories Formed With Kinect Motion Capture Data
    (Ieee, 2015) Sen, Mehmed Oguz; Erturkler, Metin
    Motion capture is the process of recording people or object movements. Generally, marker based systems are used for this process. In these systems, motion data is acquired by placing markers on certain points on the object whose motion is going to be captured. However preserving markers' positions on the object, dependency on the object, high cost and requirement of specially designed workspace causes difficulties in usage. In recent years, development of motion sensing camera systems like Microsoft Kinect became an alternative of marker based systems. In this study, Frechet distance algorithm is suggested for similarity measurement of movement trajectories that are formed by motion capture data acquired using Microsoft Kinect cameras. Measurement of movement similarity is crucial for certain applications. (activity recognition, object classification, path finding for humanoid robots etc.)