A new strategy for improving vision based tracking accuracy based on utilization of camera calibration information

摘要

Abstract— Camera calibration is one of the essentialudcomponents of a vision based tracking system where the objectiveudis to extract three dimensional information from a set of twouddimensional frames. The information extracted from theudcalibration process is significant for examining the accuracy of theudvision sensor, and thus further for estimating its effectiveness as audtracking system in real applications. This paper introducesudanother use for this information in which the proper location ofudthe camera can be predicted. Anew mathematical formula basedudon utilizing the extracted calibration information was used forudfinding the optimum location for the camera, which provides theudbest detection accuracy. Moreover, the calibration informationudwas also used for selecting the proper image Denoising filter. Theudresults obtained proved the validity of the proposed formula inudfinding the desired camera location where the smallest detectionuderrors can be produced. Also, results showed that the properudselection of the filter parameters led to a considerableudenhancement in the overall accuracy of the camera, reducing theudoverall detection error by 0.2 mm.