Updating attitude precisely in time is the primary task of strapdown inertial navigation system(SINS) algorithms. This paper mainly studied the application of rotation vector in three different methods of data fusion respectively named linear interpolation, gradient descent and complementary filter for attitude-updating, using low-cost MEMS inertial sensors in SINS. Meanwhile, an idea that the quaternion attitude could be updated by constructing micro-rotation quaternion from rotation vector in the sampling interval is proposed. The idea is based on geometric interpretation of space rotation transformation, while the general method is the differential equations of quaternion about rotation vector. Therefore the new method is an approximation method within enough short update interval, but its best superiority is the higher speed of attitude-updating than general method with little loss of accuracy because of no necessary to solve differential equations. The experimental results also show the effectiveness and accuracy of three improved algorithms with the new idea.