Camera calibration is a crucial key in image-based optical measurement. Calibration methods based on different targets are popular. The camera calibration method based on traditional sphere targets can calibrate cameras that are arranged in various angles at the same time, but the calibration precision is relatively low. The camera calibration method based on checkerboard-based targets proposed by Zhang offers a simple calibration process and high accuracy, but it is not suitable for simultaneous calibration of multiple cameras at different angles. In this paper, we propose a new form of target, namely, grid sphere target, which combines the advantages of sphere target and planar target. The grid sphere target is a sphere target with circles of latitude and longitude, similar to the globe. The proposed camera calibration method can not only calibrate cameras that are arranged in many different angles at the same time, but also obtain high accuracy similar to that of Zhang’s method. In the proposed method, the initial solutions of the intrinsic and extrinsic parameters of the camera are solved according to elliptical images of the latitude and longitude circles in the target. The intersection points of longitude and latitude circles are then used as the image feature points, and the optimal solutions of the intrinsic parameters of the camera are obtained by nonlinear optimization. Experimental results validate the effectiveness of the proposed method.