作者: Raghu Raj P. Kumar , Suresh S. Muknahallipatna , John E. McInroy
关键词:
摘要: In our previous work, a novel algorithm to perform robust pose estimation was presented. The estimated using points on the object regions image correspondence. laboratory experiments conducted in work showed that accuracy of over 99% for position and 84% orientation estimations respectively. However, larger objects, requires high number achieve same accuracy. requirement higher makes algorithm, computationally intensive resulting infeasible real-time computer vision applications. this paper, is parallelized run NVIDIA GPUs. results indicate even objects having more than 2000 points, can estimate real time each frame high-resolution videos.