Since the American Mars Exploration Rover Opportunity landed on Mars in 2004, it has travelled more than 40 km, and heading-determination technology based on its sun sensor has played an important role in safe driving of the rover. A high-precision heading-determination method will always play a significant role in the rover’s autonomous navigation system, and the precision of the measured heading strongly affects the navigation results. In order to improve the heading precision to the 1-arcminute level, this paper puts forward a novel calibration algorithm for solving the comparable distortion of large-field sun sensor by introducing an antisymmetric matrix. The sun sensor and inclinometer alignment model are then described in detail to maintain a high-precision horizon datum, and a strict sun image centroid-extraction algorithm combining subpixel edge detection with circle or ellipse fitting is presented. A prototype comprising a sun sensor, electronic inclinometer, and chip-scale atomic clock is developed for testing the algorithms, models, and methods presented in this paper. Three field tests were conducted in different months during 2017. The results show that the precision of the heading determination reaches 0.28–0.97′ (1σ) and the centroid error of the sun image and the sun elevation are major factors that affect the heading precision.