Abstract: In the field of industrial inspection, if two depth cameras exist overlapping fields of view, a chessboard can be used as a common calibration tool to establish an association between the coordinate systems of the two cameras. However, when two depth cameras lack overlapping fields of view, the traditional chessboard calibration method is inapplicable. Instead, this paper introduces a calibration approach termed the “one rod, two balls” scheme. By recording 7 to 20 positions of the calibration object, the transformation matrix between the two camera coordinate systems is determined. This method involves seven variables accounting for translation, rotation, and rod length, with the “constant rod length” being a critical factor in linking the camera coordinate systems. The centers of the balls are consistently determined using a Voronoi diagram, and an objective function addresses deviations in rod lengths. The L-BFGS algorithm is employed as the solver. Through extensive simulation experiments, the effectiveness of the proposed method has been verified. Experimental results on a self-built platform show that the method is simple to operate, efficient in computation, with an average error of 0.003 1 mm and a standard deviation of 0.001 0 mm, which meets the accuracy requirements for various industrial inspection applications.