Fusion of laser point clouds and color images with post-calibration

Abstract

Fusion of laser point clouds and color images has a great advantage in the photogrammetry, computer vision, and computer graphics communities. Most of existing methods estimate the projection matrix for mapping laser points to image pixels based on pre-calibration using checkboard patterns. We propose a method with post-calibration based on reference objects. We first generate an ideal projection matrix based on the pinhole camera model and the positional relationship between the laser scanner and the camera. We then correct the projection matrix based on the reference objects having longer vertical edges, as it is easy to detect the edge of the laser points. The horizontal coordinate offset is computed using the laser points and pixels at the vertical edges, and this is finally added to the projection matrix. Our method reduces the complexity of the data fusion. Experimental results verify the correctness of our method.