Search

Most 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 Sensing

Road 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 Sensing

Semantic 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., Optical and Laser Remote Sensing

Ground 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., Optical and Laser Remote Sensing

Two independent continuous global positioning system (CGPS) processing strategies, based on a double-difference regional network and a globally transformed precise point positioning solution, provide horizontal and vertical crustal motion estimates for Great Britain. Absolute gravity and geological information from late Holocene sea level data further constrain the vertical motion estimates. For 40 CGPS stations we estimate station velocities and associated uncertainties using maximum likelihood estimation, assuming the presence of white and coloured noise. Horizontal station velocity estimates agree to <1 mm yr(-1) between the two CGPS processing strategies and closely follow predicted plate motions. Residual velocities, generally < 1 mm yr(-1), follow no regular pattern, that is, there is no discernible internal deformation, nor any dependence on station monumentation or time-series length. Vertical station velocity estimates for the two CGPS processing strategies agree to similar to 1 mm yr(-1), but show an offset of similar to 1 mm yr(-1) with respect to the absolute gravity (AG) estimates. We attribute this offset to a bias related to known issues in current CGPS results and correct for it by AG-alignment of our CGPS estimates of vertical station velocity. Both CGPS estimates and AG-aligned CGPS estimates of present-day vertical crustal motions confirm the pattern of subsidence and uplift in Great Britain derived from Holocene sea level data for the last few thousand years: ongoing subsidence on Shetland, uplift in most areas of Scotland, and subsidence in large areas of England and Wales.

CGPS coordinate time series analysis strategy was evaluated to determine highly accurate vertical station velocity estimates with realistic uncertainties. This strategy uses a combination of techniques to (1) obtain the most accurate parameter estimates of the station motion model, (2) infer the stochastic properties of the time series in order to compute more realistic error bounds for all parameter estimates, and (3) improve the understanding of apparent common systematic variations in the CGPS coordinate time series, which are believed to be of geophysical and/or technical origin. The strategy provided a pre-processing of the coordinate time series in which outliers and discontinuities were identified. Subsequent parameterization included a mean value, a constant rate, periodic terms with annual and semi-annual frequencies, and offset magnitudes for identified discontinuities. All parameters plus the magnitudes of different stochastic noise were determined using maximum likelihood estimation (MLE). Empirical orthogonal function (EOF) analysis was used to study both the temporal and spatial variability of the common modes determined by this technique. After outlining the CGPS coordinate time series analysis strategy this paper shows initial results for coordinate time series for a four year (2000–2003) period from a selection of CGPS stations in Europe that are part of the European Sea Level Service (ESEAS) CGPS network

The current terrestrial reference frame, current global GPS products and current precise GPS processing techniques, limit the determination of accurate, long-term, vertical station velocities from continuous GPS measurements on a global scale. Several authors have reported biases in their vertical station velocities determined from continuous GPS when compared to alternative geodetic methods. It has been argued that until these problems have been resolved, the study of relative land and sea level rates on regional scales is the only way to investigate vertical land movements at tide gauges co-located with continuous GPS. In the UK, we have been operating a network of continuous GPS and absolute gravimetry stations for the purpose of determining vertical land movements at tide gauges for almost ten years. This network consists often continuous GPS stations and three absolute gravimetry stations, all of which are either co-located or close to tide gauges. In this paper, we compare vertical land movements obtained from both geodetic methods with estimates of vertical land movements from high quality, independent geological and geophysical evidence, and derive a GPS-specific bias for which the estimates of vertical land movements from all continuous GPS stations are corrected. Based on recently published mean sea level trends by the Permanent Service for Mean Sea Level, we estimate a change in sea level, de-coupled from vertical land movements, for the British Isles

Researchers investigating climate change have used historical tide-gauge measurements from all over the world to investigate the changes in sea-level that have occurred over the last century or so. However, such estimates are a combination of any true sea-level variations and any vertical movements of the land at the specific tide-gauge. For a tide-gauge record to be used to determine the climate related component of changes in sea-level, it is therefore necessary to correct for the vertical land movement component of the observed change in sea-level. In 1990, the Institute of Engineering Surveying and Space Geodesy and Proudman Oceanographic Laboratory started developing techniques based oil the Global Positioning System (GPS) for measuring vertical land movements (VLM) at tide-gauges in the UK. This paper provides brief details of these early developments and shows how they led to the establishment of continuous GPS (CGPS) stations at a number of tide-gauges. The paper then goes on to discuss the use of absolute gravity (AC), as an independent technique for measuring VLM at tide-gauges. The most recent results, from CGPS time-series dating back to 1997 and AG time-series dating back to 1995/1996, are then used to demonstrate the complementarity of these two techniques and their potential for providing site-specific estimates of VLM at tide-gauges in the UK.