Development of Auto-Calibration for In-Vehicle Multi-Camera Monitor System: High-Precise Position Detector of Target Marker

摘要

We developed a novel image processing algorithm to detect with high-precision the positions of target markers in the fisheye camera images used in the camera calibration process of an in-vehicle multi-camera monitor system. The key point of the algorithm is a newly developed region extraction process that utilizes shape features to efficiently reduce the many potential regions to promising marker candidates. The final detection results are obtained by template matching on the candidates and the detected positions are refined through a filtering process. Based on experiments using video data under several illumination conditions, we confirmed achievement of a 100% detection rate in which the positional accuracy is within one pixel. Our method is expected to save manual labor in the calibration process and ensure stable quality on the mass-production line of the synthesized image produced by the multi-camera monitor system.