作者: Ho-Sik Park , Cheol-Soo Bae
DOI: 10.7840/KICS.2011.36C.7.421
关键词:
摘要: The importance of the real-time object recognition and tracking field has been growing steadily due to rapid advancement in computer vision applications industry. As is well known, mean-shift algorithm widely used robust systems. Since mentioned easy implement efficient computation, many say it suitable be applied However, one major drawbacks this that always converges a local mode, failing perform cluttered environment. In paper, an Optical Flow-based which fits for multiple moving objects proposed. Also tests, newly proposed method contributed raising similarity objects, was as high 0.96, up 13.4% over algorithm. Meanwhile, level pixel errors from using new keenly decreased by more than 50% applying If data processing speed video surveillance systems can reduced further, owing improved algorithms faster functions, we will able expect much intelligent industrial arena.