36 research outputs found
Detecting offsets in GPS time series: First results from the detection of offsets in GPS experiment
No full textpeer reviewedThe accuracy of Global Positioning System (GPS) time series is degraded by the presence of offsets. If these are not detected and adjusted correctly they bias velocities, and hence geophysical estimates, and degrade the terrestrial reference frame. They also alter apparent time series noise characteristics as undetected offsets resemble a random walk process. As such, offsets are a substantial problem. A number of offset detection methods have been developed across a range of fields, and some of these are now being tested in geodetic time series.
The DOGEx (Detection of Offsets in GPS Experiment) project aims to test the effectiveness of automated and manual offset detection approaches and the subsequent effect on GPS-derived velocities. To do this, simulated time series were first generated that mimicked realistic GPS data consisting of a velocity component, offsets, white and flicker noises (1/f spectrum noises) composed in an additive model. We focus on offset detection and together with velocity biases induced by incorrect offset detection. We show that, at present, manual methods (where offsets are hand -picked by GPS time series experts) almost always give better results than automated or semi-automated methods (two automated methods give quite similar velocity bias as the best
manual solutions). For instance, the 5th percentile ranges (5% to 95%) in velocity bias for automated approaches is equal to 4.2mm/year,whereas it is equal to 1.8mm/yr for the manual solutions. However the True Positive detection rate of automated solutions is significantly higher than those for the manual solutions, being around 37% for the best automated, and 42% for the best manual solution. The amplitude of offsets detectable by automated solutions is greater than for hand picked solutions, with the smallest detectable offset for the two best manual solutions equal to 5mm and 7mm and to 8mm and 10mm for the two best automated solutions. The best manual solutions yielded velocity biases from the truth commonly in the range ±0.2mm/yr,
whereas the best automated solutions produced biases no better than double this range. Assuming the simulated time series noise levels continue to be representative of real GPS time series, robust geophysical interpretation of individual site velocities lower than these levels is
therefore not robust. Further work is required before we can routinely interpret sub-mm/yr velocities for single GPS stations
Multi-GNSS Slant Wet Delay Retrieval Using Multipath Mitigation Maps
No full textThe conventional Global Navigation Satellite System (GNSS) processing is typically contaminated with errors due to atmospheric variabilities, such as those associated with the mesoscale phenomena. These errors are manifested in the parameter estimates, including station coordinates and atmospheric products. To enhance the accuracy of these GNSS products further, a better understanding of the local-scale atmospheric variability is necessary. As part of multi-GNSS processing, station coordinates, carrier phase ambiguities, orbits, zenith total delay (ZTD) and horizontal gradients are the main parameters of interest. Here, ZTD is estimated as the average zenith delay along the line-of-sight to every observed GNSS satellite mapped to the vertical while the horizontal gradients are estimated in NS and EW directions and provide a means to partly account for the azimuthally inhomogeneous atmosphere. However, a better atmospheric description is possible by evaluating the slant path delay (SPD) or slant wet delay (SWD) along GNSS ray paths, which are not resolved by ordinary ZTD and gradient analysis. SWD is expected to provide better information about the inhomogeneous distribution of water vapour that is disregarded when retrieving ZTD and horizontal gradients. Usually, SWD cannot be estimated directly from GNSS processing as the number of unknown parameters exceeds the number of observations. Thus, SWD is generally calculated from ZTD for each satellite and may be dominated by un-modelled atmospheric delays, clock errors, unresolved carrier-phase ambiguities and near-surface multipath scattering. In this work, we have computed multipath maps by stacking individual post-fit carrier residuals incorporating the signals from four GNSS constellations, i.e. BeiDou, Galileo, Glonass and GPS. We have selected a subset of global International GNSS Service (IGS) stations capable of multi-GNSS observables located in different climatic zones. The multipath effects are reduced by subtracting the stacked multipath maps from the raw post-fit carrier phase residuals. We demonstrate that the multipath stacking technique results in significantly reduced variations in the one-way post-fit carrier phase residuals. This is particularly evident for lower elevation angles, thus, producing a retrieval method for SWD that is less affected by site-specific multipath effects. We show a positive impact on SWD estimation using our multipath maps during increased atmospheric inhomogeneity as induced by severe weather events.VAPOU
AN EFFICIENT DEEP LEARNING APPROACH FOR GROUND POINT FILTERING IN AERIAL LASER SCANNING POINT CLOUDS
No full textGround surface extraction is one of the classic tasks in airborne laser scanning (ALS) point cloud processing that is used for three-dimensional (3D) city modelling, infrastructure health monitoring, and disaster management. Many methods have been developed over the last three decades. Recently, Deep Learning (DL) has become the most dominant technique for 3D point cloud classification. DL methods used for classification can be categorized into end-to-end and non end-to-end approaches. One of the main challenges of using supervised DL approaches is getting a sufficient amount of training data. The main advantage of using a supervised non end-to-end approach is that it requires less training data. This paper introduces a novel local feature-based non end-to-end DL algorithm that generates a binary classifier for ground point filtering. It studies feature relevance, and investigates three models that are different combinations of features. This method is free from the limitations of point clouds’ irregular data structure and varying data density, which is the biggest challenge for using the elegant convolutional neural network. The new algorithm does not require transforming data into regular 3D voxel grids or any rasterization. The performance of the new method has been demonstrated through two ALS datasets covering urban environments. The method successfully labels ground and non-ground points in the presence of steep slopes and height discontinuity in the terrain. Experiments in this paper show that the algorithm achieves around 97% in both F1-score and model accuracy for ground point labelling
Robust approach for urban road surface extraction using mobile laser scanning 3D point clouds
No full textRoad surface extraction is crucial for 3D city analysis. Mobile laser scanning (MLS) is the most appropriate data acquisition system for the road environment because of its efficient vehicle-based on-road scanning opportunity. Many methods are available for road pavement, curb and roadside way extraction. Most of them use classical approaches that do not mitigate problems caused by the presence of noise and outliers. In practice, however, laser scanning point clouds are not free from noise and outliers, and it is apparent that the presence of a very small portion of outliers and noise can produce unreliable and non-robust results. A road surface usually consists of three key parts: road pavement, curb and roadside way. This paper investigates the problem of road surface extraction in the presence of noise and outliers, and proposes a robust algorithm for road pavement, curb, road divider/islands, and roadside way extraction using MLS point clouds. The proposed algorithm employs robust statistical approaches to remove the consequences of the presence of noise and outliers. It consists of five sequential steps for road ground and non-ground surface separation, and road related components determination. Demonstration on two different MLS data sets shows that the new algorithm is efficient for road surface extraction and for classifying road pavement, curb, road divider/island and roadside way. The success can be rated in one experiment in this paper, where we extract curb points; the results achieve 97.28%, 100% and 0.986 of precision, recall and Matthews correlation coefficient, respectively. Optical and Laser Remote Sensin
Impact of Antenna Phase Centre Calibrations on Position Time Series: Preliminary Results
No full textpeer reviewedAdvances in GPS error modelling and the continued effort of re-processing have considerably decreased the scatter in position estimates over the last decade. The associated reduction of noise in derived position time series has revealed the presence of previously undetected periodic signals. It has been shown that these signals have frequencies related to the orbits of the GPS satellites. A number of potential sources for these periodicities at the draconitic frequency and its harmonics have already been suggested in the literature and include, e.g., errors in the sub-daily tidal models, multipath and unresolved integer ambiguities.
Due to the geometrical relationship between the observing site and the orbiting satellite, deficiencies in the modelling of electromagnetic phase centres of receiving antennas have the potential to also contribute to the discovered periodic signals. The change from relative to absolute type mean antenna/radome calibrations within the International GNSS Service (IGS) led to a significant improvement, but the use of individual calibrations could possibly add further refinements to computed solutions. However, at this stage providing individual calibrations for all IGS stations is not feasible. Furthermore, antenna near-field electromagnetic effects might outweigh the benefits of individual calibrations once an antenna is permanently installed.
In this study, we investigate the differences between position estimates obtained using individual and type mean antenna/radome calibrations as used by the IGS community. We employ position time series derived from precise point positioning (PPP) as implemented in two scientific GNSS software packages. Our results suggest that the calibration differences propagate directly into the position estimates, affecting both sub-daily and daily results and yielding periodic variations.
The sub-daily variations have periods close to half a sidereal day and one sidereal day with peak-to-peak amplitudes of up to 10~mm in all position components. The stacked power spectra of the daily difference time series reveal peaks at the GPS draconitic frequency and its harmonics with peak-to-peak amplitudes of up to 1~mm. Although these results are still preliminary, they confirm that small differences between individual and type mean antenna/radome calibrations propagate into position time series
and may be partly responsible for the spurious signals with draconitic frequency and its harmonics
A Two-Step Feature Extraction Algorithm: Application to deep learning for point cloud classification
No full textMost deep learning (DL) methods that are not end-to-end use several multi-scale and multi-type hand-crafted features that make the network challenging, more computationally intensive and vulnerable to overfitting. Furthermore, reliance on empirically-based feature dimensionality reduction may lead to misclassification. In contrast, efficient feature management can reduce storage and computational complexities, builds better classifiers, and improves overall performance. Principal Component Analysis (PCA) is a well-known dimension reduction technique that has been used for feature extraction. This paper presents a two-step PCA based feature extraction algorithm that employs a variant of feature-based PointNet (Qi et al., 2017a) for point cloud classification. This paper extends the PointNet framework for use on large-scale aerial LiDAR data, and contributes by (i) developing a new feature extraction algorithm, (ii) exploring the impact of dimensionality reduction in feature extraction, and (iii) introducing a non-end-to-end PointNet variant for per point classification in point clouds. This is demonstrated on aerial laser scanning (ALS) point clouds. The algorithm successfully reduces the dimension of the feature space without sacrificing performance, as benchmarked against the original PointNet algorithm. When tested on the well-known Vaihingen data set, the proposed algorithm achieves an Overall Accuracy (OA) of 74.64% by using 9 input vectors and 14 shape features, whereas with the same 9 input vectors and only 5PCs (principal components built by the 14 shape features) it actually achieves a higher OA of 75.36% which demonstrates the effect of efficient dimensionality reduction. Optical and Laser Remote Sensin
INVESTIGATION OF POINTNET FOR SEMANTIC SEGMENTATION OF LARGE-SCALE OUTDOOR POINT CLOUDS
No full textpeer reviewedSemantic segmentation of point clouds is indispensable for 3D scene understanding. Point clouds have credibility for capturing geometry of objects including shape, size, and orientation. Deep learning (DL) has been recognized as the most successful approach for image semantic segmentation. Applied to point clouds, performance of the many DL algorithms degrades, because point clouds are often sparse and have irregular data format. As a result, point clouds are regularly first transformed into voxel grids or image collections. PointNet was the first promising algorithm that feeds point clouds directly into the DL architecture. Although PointNet achieved remarkable performance on indoor point clouds, its performance has not been extensively studied in large-scale outdoor point clouds. So far, we know, no study on large-scale aerial point clouds investigates the sensitivity of the hyper-parameters used in the PointNet. This paper evaluates PointNet’s performance for semantic segmentation through three large-scale Airborne Laser Scanning (ALS) point clouds of urban environments. Reported results show that PointNet has potential in large-scale outdoor scene semantic segmentation. A remarkable limitation of PointNet is that it does not consider local structure induced by the metric space made by its local neighbors. Experiments exhibit PointNet is expressively sensitive to the hyper-parameters like batch-size, block partition and the number of points in a block. For an ALS dataset, we get significant difference between overall accuracies of 67.5% and 72.8%, for the block sizes of 5m×5m and 10m×10m, respectively. Results also discover that the performance of PointNet depends on the selection of input vectors.SOLSTIC
The status of GNSS data processing systems to estimate integrated water vapour for use in numerical weather prediction models
No full textModern Numerical Weather Prediction (NWP) models make use of the GNSS-derived Zenith Total Delay (ZTD) or Integrated Water Vapour (IWV) estimates to enhance the quality of their forecasts. Usually, the ZTD is assimilated into the NWP models on 3- hourly to 6-hourly intervals but with the advancement of NWP models towards higher update rates e.g. 1-hourly cycling in the Rapid Update Cycle (RUC) NWP, it has become of high interest to estimate ZTD on sub-hourly intervals. In turn, this imposes requirements related to the timeliness and accuracy of the ZTD estimates and has lead to a development of various strategies to process GNSS observations to obtain ZTD with different latencies and accuracies. Using present GNSS products and tools, ZTD can be estimated in real-time (RT), near real-time (NRT) and post-processing (PP) modes. The aim of this study is to provide an overview and accuracy assessment of various RT, NRT, and PP IWV estimation systems and comparing their achieved accuracy with the user requirements for GNSS meteorology. The NRT systems are based on Bernese GPS Software 5.0 and use a double-differencing strategy whereas the PP system is based on the Bernese GNSS Software 5.2 using the precise point positioning (PPP) strategy. The RT systems are based on the BKG Ntrip Client 2.7 and the PPPWizard both using PPP. The PPP-Wizard allows integer ambiguity resolution at a single station and therefore the effect of fixing integer ambiguities on ZTD estimates will also be presented
Status of TIGA activities at the British Isles continuous GNSS Facility and the University of Luxembourg
No full textpeer reviewedIn 2013 the InternationalGNSS Service (IGS) Tide Gauge Benchmark Monitoring (TIGA) Working Group started their reprocessing campaign which proposes to reanalyse all relevant GPS observations from 1995 to the end of 2012 in order to provide high quality estimates of vertical land motion for monitoring of sea level changes. The TIGA Working Group will also produce a combined solution from the individual TIGA Analysis Centres (TAC) contributions. The consortium of British Isles continuous GNSS Facility (BIGF) and the University of Luxembourg
TAC (BLT) will contribute weekly minimally constrained SINEX solutions from its reprocessing using the Bernese GNSS Software (BSW) version 5.2 and the University of Luxembourg will also act as a TIGA Combination Centre (TCC). The BLT will generate two solutions, one based on
BSW5.2 using a network double difference (DD) strategy and a second one based on BSW5.2 using a Precise Point Positioning (PPP) strategy. In the DD strategy we have included all IGb08 core stations in order to achieve a consistent reference frame implementation. As an initial test for the TIGA combination, all TACs agreed to provide weekly SINEX solutions for a four-week
period in December 2011. Taking these individual TAC solutions the TCC has computed a first combination using two independent combination software packages: CATREF and GLOBK. In this study we will present preliminary results from the BLT reprocessing and from the combination test
An assessment of Bernese GPS software precise point positioning using IGS final products for global site velocities
No full text
