Precise Orbit Determination (POD) of Low Earth Orbiter satellites (LEOs) based on Point Positioning (PPP) technique utilizing dual-frequency spaceborne GPS observations has become one of the best POD methods at present. Quality control of raw spaceborne GPS observations is very complex but critical for achieving high orbiting accuracy. Among the various existing methods for detecting outliers, the majority-voting algorithm used in Bernese 5.0 is a very efficient one. However, the performance of this algorithm may be affected by choice of the input parameters such as standard deviation for arranging the observations into groups, standard deviation for setting the rejection threshold and factors α and β whose values are often manually selected by experience. If the threshold is set too high, the relatively small outliers might not be found; on the contrary, more observations might be excluded and no solution could be computed for a particular epoch if the number of satellites per epoch is set to be smaller than 4. To overcome these limitations, this paper presents an improved majority-voting algorithm, which determines the some options by iteration instead of manual selection, and utilizes QUasi-Accurate Detection of outliers (QUAD) to correct the marked observations by this algorithm. The determined orbit of LEO using this new algorithm is continuous and smooth. Therefore, the improved majority voting is feasible and efficient.