The tunnel image has simple texture,few match-ing features,therefore the stitching result based on feature matching is not ideal. At the same time,due to the existence of camera distortion,the image taken by each CCD camera has the greatest deformation in the edge areawhere the stitch-ing place is also there. As a result,there will be serious stitch-ing errors with the entire transformed image. To this end,Li-DAR is involved to simulate the tunnel surface according to the distance information provided by it. Then the camera's in-ternal and external parameters are used to map the image to the simulated tunnel surface,thus image stitching on the ob-ject side is completed. With the help of Bayesian optimization (BO) algorithm,with the goal of maximizing the similarity of the images of adjacent camera object overlaps,the coordinate system conversion relationship between Lidar and CCD cam-eras is iteratively optimized to realize the laser-assisted stitch-ing of tunnel images. The experimental results demonstrate the effectiveness of the method.
维普
万方数据