Extracting accumulation and erosion from near-continuous 3D observation of a natural scene is an important step in many geoscientific analyses. Change forms are typically detected and quantified via pairwise 3D surface changes. Surface increase or decrease with a duration over multiple acquisitions may not be detected if corresponding changes are small between successive epochs. Analyzing longer timespans performs poorly where coinciding change processes lead to superimposed change signals. We examine how spatiotemporal segmentation improves the extraction of change volumes from near-continuous 3D time series by using the full temporal information of surface changes. Synthetic changes and manually derived reference changes from an hourly terrestrial laser scanning time series of snow cover monitoring are detected in the temporal domain and delineated accurately (area intersection over union of 0.86 for snow cover changes). The accuracy of change volumes ( $\mu = -25\%$ , $\sigma = 20\%$ deviation to the reference) can be improved in the future by refining the detected start and end times in the automatic approach. The baseline methods only achieve high quantification accuracies if area and timespans of changes are known a priori . Incorporating the surface change history in change extraction is therefore essential for fully automatic change analysis of near-continuous 3D time series as acquired in geographic monitoring settings.
Near-continuously acquired terrestrial laser scanning (TLS) data contains valuable information on natural surface dynamics. An important step in geographic analyses is to detect different types of changes that can be observed in a scene. For this, spatiotemporal segmentation is a time series-based method of surface change analysis that removes the need to select analysis periods, providing so-called 4D objects-by-change (4D-OBCs). This involves higher computational effort than pairwise change detection, and efforts scale with (i) the temporal density of input data and (ii) the (variable) spatial extent of delineated changes. These two factors determine the cost and number of Dynamic Time Warping distance calculations to be performed for deriving the metric of time series similarity. We investigate how a reduction of the spatial and temporal resolution of input data influences the delineation of twelve erosion and accumulation forms, using an hourly five-month TLS time series of a sandy beach. We compare the spatial extent of 4D-OBCs obtained at reduced spatial (1.0 m to 15.0 m with 0.5 m steps) and temporal (2 h to 96 h with 2 h steps) resolution to the result from highest-resolution data. Many change delineations achieve acceptable performance with ranges of ±10 % to ±100 % in delineated object area, depending on the spatial extent of the respective change form. We suggest a locally adaptive approach to identify poor performance at certain resolution levels for the integration in a hierarchical approach. Consequently, the spatial delineation could be performed at high accuracy for specific target changes in a second iteration. This will allow more efficient 3D change analysis towards near-realtime, online TLS-based observation of natural surface changes.
Geographic observation benefits from the increasing availability of time series of 3D geospatial data, which allow analysis of change processes at high temporal detail and over extensive periods. In this context, the demand for advanced methods to detect and extract topographic surface changes from these 4D geospatial data emerges. Changes in natural scenes occur with varying magnitude, duration, spatial extent, and change rate, and the timing of their occurrence is not known. Standard pairwise change detection requires the selection of fixed analysis periods and the specification of magnitude thresholds to determine accumulation or erosion forms. In settings with continuous surface morphology and dynamic changes to the surface due to material transport, such change forms are typically temporary and may be missed or aggregated if they occur with spatial and/or temporal overlap. This is overcome with the extraction of 4D objects-by-change (4D-OBCs). These objects are obtained by firstly detecting surface changes in the temporal domain at locations in the scene. Subsequently, they are spatially delineated by considering the full history of surface change during region growing from the seed location of a detected change. To perform this spatiotemporal segmentation systematically for entire 3D time series, we develop a fully automatic approach of seed detection and selection, combined with locally adaptive thresholding for region growing of individual objects with varying change properties. We apply our workflow to a five-months hourly time series of around 3,000 terrestrial laser scanning point clouds acquired for coastal monitoring at a sandy beach in The Netherlands. This provides 2,021 4D-OBCs as extracted accumulation or erosion forms. Results are validated through majority agreement of six expert analysts, who evaluate the seg-mentation performance at sample locations throughout the scene. Accordingly, our method extracts surface changes with an error of omission of 4.7% and an error of commission of 16.6%. We examine the results and provide considerations how postprocessing of segments can further improve the change analysis workflow. The developed approach thereby provides a powerful tool for automatic change analysis in 4D geospatial data, namely to detect and delineate natural surface changes across space and time.
Abstract: High-resolution 3D time series data of the surface can be used to analyze spatiotemporal morphologic changes of the snow cover. For this, a time series of the snow cover below the UFS was acquired hourly over several days with a terrestrial laser scanner. Using this dataset, a statistical method is developed for morphological change analysis to automatically derive information on spatial and temporal dynamics of the snow surface. This contributes to a novel approach of autonomous long-term monitoring of topographic change. Zusammenfassung: Mit einer hochaufgelösten 3D-Zeitserie der Oberfläche kann die morphologische Schneedeckenveränderung in ihrer raumzeitlichen Variabilität untersucht werden. Dazu wurde die Schneedecke unterhalb der UFS über mehrere Tage stündlich mit einem terrestrischen Laserscanner aufgenommen. Mit den Daten wird eine statistische Methode entwickelt, um automatisiert Informationen über die Dynamik der Schneedecke abzuleiten. Dies trägt zu einem neuen Ansatz für die autonome Langzeit-Beobachtung von topographischen Veränderungen bei.
Zeitserien von 3D-Punktwolken werden zunehmend für die Beobachtung geomorphologischer Phänomene genutzt. Dieser Beitrag untersucht, wie eine veränderte zeit-liche Auflösung die raumzeitliche Abgrenzung von Oberflächenprozessen beeinflusst. Dazu wird eine Zeitserien-basierte Region Growing Segmentierung verwendet. Die Untersuchung erfolgt am Beispiel einer Sandbank an einem Strand in den Niederlanden, der über fünf Mo-nate stündlich mit terrestrischem Laserscanning erfasst wurde. Die Ergebnisse zeigen, dass die zeitliche Prozessabgrenzung maßgeblich vom gewählten Zeitintervall abhängt. Auf Basis dieser Prozessabgrenzung kann die anschließende raumzeitliche Segmentierung auf ausge-dünnten Zeitserien und somit mit reduziertem Berechnungsaufwand erfolgen.
Time series of topographic data are becoming increasingly widespread for monitoring geomorphic activity. Dense 3D time series are now obtained by near-continuous terrestrial laser scanning (TLS) installations, which acquire data at high frequency (e.g. hourly) and over long periods. Such datasets contain valuable information on topographic evolution over varying spatial and temporal scales. Current analyses however are mostly conducted based on pairwise surface or object-based change, which typically require the selection of thresholds and intervals to identify the processes involved and fail to account for the full history of change. Detected change may therefore be difficult to attribute to one or more underlying geomorphic processes causing the surface alteration. We present an automatic method for 4D change analysis that includes the temporal domain by using the history of surface change to extract the period and spatial extent of changes. A 3D space-time array of surface change values is derived from an hourly TLS time series acquired at a sandy beach over five months (2967 point clouds). Change point detection is performed in the time series at individual locations and used to identify change processes, such as the appearance and disappearance of an accumulation form. These provide the seed to spatially segment '4D objects-by-change' using a metric of time series similarity in a region growing approach. Results are compared to pairwise surface change for three selected cases of anthropogenic and natural processes on the beach. The obtained information reflects the evolution of a change process and its spatial extent over the change period, thereby improving upon the results of pairwise analysis. The method allows the detection and spatiotemporal delineation of even subtle changes induced by sand transport on the surface. 4D objects-by-change can therefore provide new insights on spatiotemporal characteristics of geomorphic activity, particularly in settings of continuous surfaces with dynamic morphologies.
Geomorphic processes occur spatially variable and at varying magnitudes, frequencies and velocities, which poses a great challenge to current methods of topographic change analysis. For the quantification of surface change, permanent terrestrial laser scanning (TLS) can generate time series of 3D point clouds at high temporal and spatial resolution. We investigate how the temporal interval influences volume change observed on a sandy beach regarding the temporal detail of the change process and the total volume budget, on which accretion and erosion counteract. We use an hourly time series of TLS point clouds acquired over six weeks in Kijkduin, the Netherlands. A raster-based approach of elevation differencing provides the volume change over time per square meter. We compare the hourly analysis to results of a three- and six-week observation period. For the larger period, a volume increase of 0.3 m³/ m² is missed on a forming sand bar before it disappears, which corresponds to half its volume. Generally, a strong relationship is shown between observation interval and observed volume change. An increase from weekly to daily observations leads to a five times larger volume change quantified in total. Another important finding is a temporally variable measurement uncertainty in the 3D time series, which follows the daily course of air temperature. Further experiments are required to fully understand the effect of atmospheric conditions on high-frequency TLS acquisition in beach environments. Continued research of 4D geospatial analysis methods will enable automatic identification of dynamic change and improve the understanding of geomorphic processes.
3D point clouds acquired by laser scanning and other techniques are difficult to interpret because of their irregular structure. To make sense of this data and to allow for the derivation of useful information, a segmentation of the points in groups, units, or classes fit for the specific use case is required. In this paper, we present a non-end-to-end deep learning classifier for 3D point clouds using multiple sets of input features and compare it with an implementation of the state-of-the-art deep learning framework PointNet++. We first start by extracting features derived from the local normal vector (normal vectors, eigenvalues, and eigenvectors) from the point cloud, and study the result of classification for different local search radii. We extract additional features related to spatial point distribution and use them together with the normal vector-based features. We find that the classification accuracy improves by up to 33% as we include normal vector features with multiple search radii and features related to spatial point distribution. Our method achieves a mean Intersection over Union (mIoU) of 94% outperforming PointNet++'s Multi Scale Grouping by up to 12%. The study presents the importance of multiple search radii for different point cloud features for classification in an urban 3D point cloud scene acquired by terrestrial laser scanning.