GIScience & Remote Sensing (Oct 2020)
On the accuracy of low-cost dual-frequency GNSS network receivers and reference data
Abstract
The trend in spatial data accuracy has gone from meters to decimeters and even centimeter levels in the last two decades. In large part, the centimeter(s) level spatial accuracies of geospatial products result from the use of LiDAR data and both low altitude manned and unmanned aerial systems for imagery and for topographic and surface models. All these sources are dependent on high precision/accuracy GNSS technologies to achieve such accuracies. The cost of L1/L2 receivers capable of centimeter(s) level position accuracy has rapidly, in the last year, decreased from over $15k – $20k to (in early 2019) to about $300 (in 2020). Such recent low costs provide an economically affordable revolution in the use of centimeter(s) level accuracies in aerial remote sensing, ground support, field data collections, and classroom instruction with GNSS RTK technologies. Except for the marketing literature little is known about their performance in the typical applications of the remote sensing and GIS communities. How accurate are the new low-cost dual-frequency multi-constellation receivers? What is the reliability in typical landscapes and mountainous landscapes? Answers to these questions are important to control the bundle-adjustment in SfM approaches using ground control points, setting confidence limits in topographic change detection, collecting ground reference data, and evaluating orthoimagery. The problem of assessing performance in typical applications is the lack of a reference source of higher precision and accuracy than the low-cost GNSS receivers themselves. In this study, a low-cost receiver/antenna was evaluated using height modernization monuments in the National Geodetic Survey network. These monuments are typically of sub-centimeter positional accuracy. Thirty-six monuments in four piedmont counties of South Carolina were surveyed (in RTK FIX mode). Resulting accuracies for these typical environments were 2.2-cm RMSE in both horizontal and height dimensions. The 95% confidence level accuracies for the horizontal and height dimensions were 3.7-cm and 4.2-cm (95%), respectively. Performance tests in the South Carolina mountains revealed numerous issues with the low cost survey grade GNSS receiver (cellular connections, availability of reference sites, and satellite signal occlusion from mountains) that also plagues more expensive receivers.
Keywords
