In order to determine the position parameters between an imaging laser radar and a camera and to establish the mapping relation between them, an improved calibration method was proposed based on Singular Value Decomposition (SVD) and present calibration arithmetics. Firstly, the calibration principle was introduced, then the SVD was used to solve overdetermined equations to obtain the projective transformation matrixes of laser radar coordinates and camera coordinates. To reduce the influence of noise, the initial result was optimized with the Levenberg-Marquardt algorithm. Then, the calibration experiments were carried out with a 3D calibration object and the image fusion experiment was performed according to the calibration result to validate the improved calibration arithmetic. Experimental results indicate that the average calibration precision of the arithmetic is 13.33 μm and the standard error is 7.49 μm,which is doubled that of the original arithmetic. The precision of the improved calibration arithmetic can meet the requirement of image fusion.