Calibrating a small field camera is a challenging task because the traditional target with visible feature points that fit the limited space is difficult and costly to manufacture. We demonstrate a novel combined target used in camera calibration. The tangent points supplied by one circle located at the center of a square are used as invisible features, and the perspective projection invariance is proved. Both visible and invisible features extracted by the proposed feature extraction algorithm are used to solve the calibration. The target supplies a sufficient number of feature points to satisfy the requirements of calibration within a limited space. Experiments show that the approach can achieve high robustness and considerable accuracy. This approach has potential for computer vision applications particularly in small fields of view.