A method of extrinsic calibration between a four-layer laser range finder and a camera

This paper presents a novel extrinsic calibration method based on four-layer laser and a camera, In the case of camera intrinsic parameters are known, we can obtain the corresponding points on a set of laser data and image feature points by using the same feature that both captured by laser and camera, which followed by acquiring the extrinsic parameters of sensor sysems according to the projective relation between the laser coordinate and pixel coordinate system. In order to make full use of advantages of four-layer laser, we designed a special calibration board, which allows us to solve unique solution by using P6P algorithm with only three different postures. The precision of solutioin depends on accuracy of feature points, which can be affected by following reasons. 1) Lack of optimization on extracted features points. Because the interval between adjacent scan points, which may not just fall on the edge of slit, is about 20mm, so deviation range of the feature points in Y axis direction of the radar coordinate is about 0~10mm. 2) The extracted feature points are approximate. Because the ideal feature point should be located in the middle of the edge of slit, of which the width is 5mm, so deviation range of the feature points in Z axis direction of the radar coordinate is about 0~3mm. The experimental results show that the average projection error is about 30pixels, while that of which in the state of the art achieve 6pixels^[1], which is mainly occured by no optimal processing. The advantage of this method is simple and reliable, and the cost of the sensors is low as well, which allows it can be used in scenes such as off-road vehicle autonomous navigation.