Omni-vision Tracking Implementation Based On FPGA

Fisheye lens with its advantage of Wide-angle and panoramic sight has had significant impact on a variety of real-world systems, such as video surveillance systems, intelligent transportation systems, group video conferencing systems, mobile robot visual navigation systems, and so on. However, the fisheye lens brings a strong unavoidable inherent distortion while it provides hemispherical field of view, so an image restoration algorithm is always required. Many researchers have done a very good theoretical research, but these algorithms are always very complex, computationally intensive, and generally are implemented on PC. In many applications, it is difficult to meet the reality system for real-time requirements.Therefore it is a very meaningful issue to find more efficient hardware implementation of the correction algorithms.Mean Shift algorithm was proposed by Fukunaga in 1975. But it had not been obtaining people's attention until 1995.As a non-parameter statistical iterative algorithms, Li Xiang-Ru has been made a rigorous derivation of the algorithm'convergence. This job has laid a theoretical basis of the algorithm'application. Architecture of the embedded systems to achieve Mean Shift tracking algorithm is alos a very practical issues.This paper was supported by the State Ministry of Science and international scientific cooperation project. In this paper, we have carried out extensive study on the imaging principle of the fisheye lens. We also collect many correction algorithms of the fisheye lens.And we have implemented one correciton algorithm based on FPGA.In this paper, Mean Shift algorithms researches are also carrid out. We have implemented the algorithm in an embedded system and achieved good tracking performance. At last, the corrected target coordinates is corrected by using FPGA-based fish-eye lens distortion correction device. The realization of the image tracking system based on embedded architecture is a meaningful job.