Calibration and Stitching Methods of the Around View Monitor System of Articulated Multi-Carriage Road Vehicle for Intelligent Transportation 2019-01-0873
The Around View Monitor (AVM) system of the long-body road vehicle with multiple articulated carriages usually suffers from the incomplete distortion correction of fisheye cameras on the side of the coach and the irregular image stitching area caused by the change of relative position of the cameras on different carriages while the vehicle is in motion. In response to these problems, a set of calibration and stitching methods of AVM are proposed. When the system is in the calibration mode, first a two-step optimization method is adopted to solve the polynomial radial distortion-based fisheye camera model. Then, a robust corner detection technique is proposed to extract all the chessboard corners and square vertexes in the system calibration scene. With the calibrated camera model and geometric information of the mounted system, the initial look-up table from the fisheye images to a top view of the vehicle is extrapolated. When the system is in working mode, in order to obtain the real-time relative position of cameras on different carriages, the angles between adjunct carriages are measured by sensors in hinge plate. Then according the relative position of the cameras, an offset look-up table are obtained, which is combined with the initial look-up table to generate the real-time look-up table. After that, a color space conversion-based approach is utilized to reduce the variance of brightness and color between images captured by different cameras, and an angle-based weighted fusion technique is leveraged to blend the images of adjacent cameras and remove the stitching seam. Finally, a real-time top view is generated. The results of field experiments show that the proposed algorithms achieve a competitive performance with fast executing speed on the embedded system in comparison with other relevant methods.