Identification of The Survey Points from Network RTK Trajectory with Improved DBSCAN Clustering, Case Study on HNCORS

The data mining on GNSS network RTK trajectories could be used to reveal the user behavior pattern, assess the local economy development and even improve the work plan. While the identification of the survey points by stationary and moving states information, is one of the most critical step of data preparation. The performance of traditional DBSCAN approach is not desired since it only supports the horizontal geographic distance information. In this paper an improved DBSCAN clustering is proposed. A novel set of parameters as the change of heights, number of satellites, fixing statues, are taken into consideration as well as the traditional ones as time interval, horizontal distance and minimal neighbors. After the clustering, the centroid analysis is applied to determine the survey points from each cluster. Both the simulated and real data experiments from HNCORS GNSS network RTK services are carried out to test the performance. The results show that the identification rate could reach 86.3%, which is increased by 15.2% compared to the traditional approach. Meanwhile, the best identification recall rate 90.6% occurs for road style surveying activities. Moreover, this approach could also be able to find the invalid or abnormal surveys.

Automated summary

Data mining on GNSS network RTK trajectories can reveal user behavior patterns, assess local economic development, and improve work plans. An improved DBSCAN clustering algorithm considering new parameters such as height change and satellite number is proposed. Centroid analysis is used to determine survey points after clustering. Experiments using simulated and real data from HNCORS GNSS network RTK services show that the identification rate can reach 86.3%, an increase of 15.2% compared to traditional methods.