Calibration of Small and Low-Cost UAV Video System for Real-Time Planimetric Mapping

High-resolution planimetric mapping generated from unmanned aerial vehicle video greatly attract many civilian users. To make such a mapping of natural disaster, exterior of parameters (EOPs) of video frames have to be exactly determined. To this end, one from coarse to fine calibration method is developed in this paper, which include: (1) coarsely calculating camera IOP by using vanish point (VP) geometry; (2) Initially calculating Boresight matrix by simply selecting one pair of stereo frame to estimate photogrametry orientation parameters and compared to ones from the on-board GPS/INS; (3) A chain of high-overlapped video frames and valid tie points are rapidly generated based on developed data flow processing technology; (4) any EOP of generated video frames is solved based on cubic spline interpolating from all boresight aligned GPS/INS derived orientation parameters; (5) Taking tie points generated in Step 3 as observation, all EOPs solved in Step 4 and the camera's IOPs solved in Step 1 as unknown parameters, camera radial distortion as addition parameters, and a few non-traditional "ground control points" measured from registered USGS DEM and reference image, self-calibration bundle adjustment is applied to self-calibrate UAV video system. Using the calibrated EOP, IOP and USGS DEM, 2D planimetric mapping (orthoimage) is generated for each video frame individually, which are finally mosaicked automatically. The experimental results demonstrates that the planimetric accuracy of orthoimage can achieve 1~2 pixels. Some recommendations in application of UAV system for disaster management, e.g., forest fire, are made.