Triaxial accelerometers are used in various applications, such as inertial navigation systems (INSs) and inclinometers. Such accelerometers must be calibrated as accurately as possible because accelerometers with even small biases could result in a very fast position drift when they are used for INS applications and could result in inaccurate tilt angle measurements. Since the calibration parameters of a triaxial accelerometer vary under different environmental conditions, including temperature change, the gain factors and biases change whenever the sensor is switched on. Thus, an accelerometer must be calibrated before use or when the temperature is significantly changed. This paper presents a triaxial accelerometer calibration method using a mathematical model of nine calibration parameters: three gain factors, three biases and three non orthogonality factors. The fundamental principle of the proposed calibration method is that the sum of the triaxial accelerometer outputs is equal to the gravity vector when the accelerometer is stationary. The proposed method requires the triaxial accelerometer to be placed in forty eight tilt angles to estimate the nine calibration parameters. Since the mathematical model of the calibration parameters is nonlinear, an iterative method is used Levenberg-marquardt algorithm (LMA). The results are verified by comparing the estimated data and corrected data with how far or close from surface of sphere that have radius 1g and center (0,0,0) by calculating the standard deviation.