Improving GNSS data accuracy using DBSCAN, moving averages, and Hampel identifier

Abstract

There is a common issue in GNSS (Global Navigation Satellite System) data, which is the presence of outliers that can affect the accuracy of positional measurements. In this study, three methods for outlier detection and removal in GNSS data were compared: Hampel filter, moving average, and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm. Both synthetic and real GNSS datasets were used to test these methods. The Hampel filter and moving average were first applied to clean the data, and then the DBSCAN algorithm was used to detect outliers. The results were evaluated using the RMS error criterion. The study found that DBSCAN was effective with appropriate parameter settings, but the combination of Hampel filter and moving average was the most successful method. The Hampel filter was particularly efficient in filtering outliers in low-quality GNSS data. These findings suggest that the combination of multiple methods can result in more accurate and reliable outlier detection and removal in GNSS data.