Study Of Information Extracting Based On Panoramic Annular Lens (PAL)

Stereo vision is an important branch of computer vision, has also been a focus in the study of computer vision. The research achievements of stereo vision are widely used in robot vision, object recognition, precision measurement, virtual reality and other industrial and military fields. The traditional optical system imaging field is usually small, the depth information obtained by using stereo vision techniques is limited, which can cause inconvenience in practical application field. If can obtain wide field of three-dimensional information real-time, stereo vision technology will be more widely applied.This paper applied panoramic imaging system in stereo vision technology, presents a kind of information extraction method based on panoramic stereo imaging system with double coaxial panoramic annular lenses (PAL) units. Its imaging circles are concentric so that depth information of360-degree horizontal field-of-view (FOV) object can be extracted real-time. Based on imaging principle, the system is calibrated with a model that can accommodate the characteristics of non-single viewpoint system, and the calibration results are verified. Making the original omni-directional stereo image into rectangle images by projective transformation in order to make the use of the method of corresponding match in existence, Simplify the matching complexity. Combined the SIFT algorithm and region-growing algorithm applied to the corresponding point matching, obtain dense matching points and make full use of collinear constraint of the system to improve the matching speed and accuracy. Effective depth information is extracted from the captured image with applying triangulation principle. Depth information extraction experiments is showed, gives the results and error analysis. This system can restore depth information effectively, error rate of3meters distance range is in the plus or minus5.2%, proved the feasibility of this system.