The technology for high precision attitude determination when using low precision sensors is a key requirement of a modern small satellite. This paper presents a new attitude determination algorithm termed preprocess EKF(PP-EKF) based on preprocessing of the sensor data. It can enhance the overall modeling accuracy by using the quadratic penalty function to correct the dynamic model error and angular velocity error in realtime, based on the information fusion of the current and the past measurement information. The measurement model of the EKF is linearized by introducing the q method. The solution error of this process is also corrected to further improve the accuracy of the measurement model and make better use of measurement data from the low precision sensors, to obtain good attitude determination results. Finally, the simulation results demonstrate the high reliability and other advantages of the proposed algorithm. The technology for high precision attitude determination when using low precision sensors is a key requirement of a modern small satellite. This paper presents a new attitude determination algorithm termed preprocess EKF (PP-EKF) based on preprocessing of the sensor data. It can enhance the overall modeling accuracy by using the quadratic penalty function to correct the dynamic model error and angular velocity error in realtime, based on the information fusion of the current and the past measurement information. The measurement model of the EKF is linearized by introducing the q method. The solution error of this process is also corrected to further improve the accuracy of the measurement model and make better use of measurement data from the low precision sensors, to obtain good attitude determination results. Finally, the simulation results demonstrate the high reliability and other advantages of the proposed algorithm.