103 research outputs found
Methods For Feature Extraction of an Outcrop Using Terrestrial LiDAR: Methods For Feature Extraction of an Outcrop Using Terrestrial LiDAR
No full textThe following report brings to light how LiDAR data of an outcrop can be acquired and subsequently processed in order to extract the geometry of the outcrop and its pertaining parameters. It makes the use of a sandstone outcrop in the Trooz quarry, Belgium, near Liege. The outcrop was scanned using a terrestrial laser scanner, after which the point cloud data recorded information, pertaining to position and intensity was then processed and subject to interpretation. The report aims to systematically and chronologically take the reader through the entire workflow used by the author, its drawbacks and advantages. Whilst providing a specific analysis of the outcrop in question, creating a general guideline is also an aim. Prior to the processing the significance of the geologic history of the area is touched upon along with the practical acquisition of that data, concerning equipment and procedure. The physical principles of LiDAR and other possible options of acquiring data such as photogrammetry are explained, each with their benefits and shortcomings. The practicality of LiDAR is especially analysed followed by explanation and usage of processing methods such as registration, geo-referencing and computation of geometrical parameters, specifically with the help of software’s such as CloudCompare and MATLAB. Once the results of such parameters are derived, their integrity is discussed by comparing MATLAB oriented methods to those of CloudCompare. These quantified results include values of dip angle and direction over the entire outcrop and the corresponding normal vectors. Varying roughness over the surface of the outcrop and analysis of the intensity distribution present. Whilst also validating these results by comparing them with manual recorded results of dip angle and direction taken at the site. The significance of these parameters is discussed and their application in the characterization of layering. Finally, suggestions regarding the methodology both during acquisition and processing such as during geo-referencing are made by displaying obstacles encountered and flaws realized in the author’s own work.Applied Earth Science
Classifying Mangroves in Vietnam using Radar and Optical Satellite Remote Sensing: Processing Sentinel-1 and Sentinel-2 Imagery in Google Earth Engine
No full textMangroves are forest ecosystems growing in (sub)tropical saline coastal environments. With their unique root structure they serve as important natural coastal protection and provide habitats with excellent conditions for cultivating fish, shrimp and crab species. Despite all benefits mangrove forests are disappearing at alarming rates around the world but especially in Asia such as the Mekong Delta coast. Therefore, this research focusses on the Ca Mau Province in Vietnam. The Ca Mau province is the southernmost province of Vietnam with mangroves present along the coastlines, the Mui Ca Mau National Park and in mixed mangrove aquaculture farms. Remote sensing has been widely proven to be essential in mapping mangrove ecosystems. Previous research used either expensive optical and radar data sources or free but lower resolution systems. This study is the first that uses the new Copernicus Sentinel-1 radar and Sentinel-2 multispectral satellite missions that provide free available data with high spatial (10-20 meter) and temporal (10-12 days) resolution. Since optical data is prone to cloud effects and radar data is hard to interpret, both data sets are combined to investigate improvements for classifying mangroves. The data is processed in the new online Google Earth Engine platform providing a powerful tool for big data applications such as land cover classification. Optical data is found to separate mangroves by their spectral reflectance mainly in the near-infrared wavelength domain. The dominant mangrove species in the Ca Mau province, Rhizophora Apiculata and Avicennia Alba, are found to be separable from comparing unsupervised clustering results with ground truth locations. The C-band radar signal is dominated by volume scattering, indicating the density of the canopy. Especially VV-polarization has good correlation with canopy parameters. To improve information from the radar signal a temporal analysis is executed. Seasonal variations are quantified and show an increase according to the spatial succession of mangroves. Pioneer species, such as Avicennia genus, show less seasonal variations than mature species, such as Rhizophora genus. With the previous information five classes are defined: urban area, water and three mangrove classes: Rhizophora Apiculata species in extensive shrimps, Rhizophora Apiculata species in natural environment and Avicennia Alba species. A classification method is set-up in the Google Earth Engine with a Random Forest classifier using the satellite data inputs and ground truth training input of the five classes. A combination of the optical data with the temporal information of the radar data is found to be the best data input for separating those five classes. Classification results are obtained for discriminating mangrove types up to an overall accuracy of 87\%. The classification gets less reliable when mangrove species are mixed or at locations where the ground truth training input was scarce. With the resulting yearly land cover maps land cover changes can be detected. Comparing the land cover map of 2017 with a mangrove cover product of 2000 shows a regression along the southern coastline. No significant changes inside the shrimp farms are found between 2016 and 2017 but with the future availability of a long time series of Sentinel-1 and 2 data those can be detected with the method that is resulted from this study.Applied Earth Science
To order or not to order: Predicting customer grocery shopping behaviour using multi-label classification techniques
No full textResearch and Objective: In the recent years the online grocery sector experienced an enormous uplift and evolved to a highly competitive business sector. Within this demanding environment, the need for strategic information has become extremely important, as it greatly enhances decision-making processes and the optimisation of the supply chain. In this research, a novel approach is proposed that is aimed at predicting customers’ daily purchase probabilities, with the goal to improve short-term forecasting accuracy. Besides the well-acknowledged importance of forecasting practices and customer relationship management, this research is motivated by three main observations in online grocery retail; short interpurchase times, consistent shopping patterns and loyal customers. Methodology: The approach involves the application of binary classification methods to analyse and predict online shopping behaviour. Within this context, two non-parametric learning algorithms, namely stochastic gradient boosting and random forest, are compared to traditional logistic regression. Both stochastic gradient boosting and logistic regression are extended using classifier chains (CC) to handle multiple outputs. Subsequently, the obtained purchase probabilities are aggregated and compared to the predictions of a univariate Seasonal Autoregressive Integrated Moving Average Exogenous (SARIMAX) time series model. Results: The boosted tree CC model was able to achieve an improvement of 1.77% in mean-absolute-percentage error (MAPE) and 20.95% in mean-squared logarithm of the accuracy ratio (MSLAR) compared to the predictions of the random forest and an improvement of 1.15% in MAPE and 16.81% in MSLAR compared to the SARIMAX time series model. The model acquired consistent results for customer groups of different sizes, with prediction errors that exhibited the lowest bias as well as variance of all models. The analysis of the explanatory variables indicate that behavioural attributes and variables, that concern interpurchase times in particular, were most significant of the target variables. Eventually, the application of calibration methods led to a decrease in forecasting performance rather than improving it. Conclusion: This research proposes a novel approach for short-term customer demand prediction within the online grocery retail market, which can provide an alternative to conventional time series forecasting techniques. The obtained results are satisfactory and of value for management and decision makers.Mechanical Engineering | Systems and Contro
Indentifying glacial features with sentinel-2 data
No full textThe Tibetan Plateau is a vast elevated plateau in Central and East Asia. It contains thousands of glaciers and other geographical features. Through this area rivers like the Brahmaputra is flowing and making a basin providing about millions of people a home. The last years global warming has been a focus of public and scientific debate. Not knowing what to expect and what the changes are result in ruling uncertainties which are of major concern because it could cause serious implications for water resources. During this study the area located in the Upper Brahmaputra in the South-East of the Tibetan Plateau called the Yiong Zangbu catchment will be investigated. To understand what glacial changed have occurred in the past few years data from different years were collected, processed and compared. The used datasets are ASTER GDEM, HydroSHEDS, GLIMS glacier mask and Sentinel-2 . Sentinel-2 data has been processed and different images are made like true colour and false colour images. Next to that a combination of datasets are used to see whether there is an accuracy issue and to understand what kind of features can be found on which height and what spectral reflectance belongs to it. After processing all the data of two following years differences can be seen. There are lakes that are frozen out within three months. Also glaciers which expanded downwards the hills. This can be a result of strong winters. On the higher parts there was more precipitation, which is helpful while remaining the current glaciers. <br/
Road Detection from Remote Sensing Imagery
No full textRoad network maps facilitate a great number of applications in our everyday life. However, their automatic creation is a difficult task, and so far, published methodologies cannot provide reliable solutions. The common and most recent approach is to design a road detection algorithm from remote sensing imagery based on a Convolutional Neural Network, followed by a result refinement post-processing step. In this project I proposed a deep learning model that utilized the Multi-Task Learning technique to improve the performance of the road detection task by incorporating prior knowledge constraints. Multi-Task Learning is a mechanism whose objective is to improve a model's generalization performance by exploiting information retrieved from the training signals of related tasks as an inductive bias, and, as its name suggests, solve multiple tasks simultaneously. Carefully selecting which tasks will be jointly solved favors the preservation of specific properties of the target object, in this case, the road network. My proposed model is a Multi-Task Learning U-Net with a ResNet34 encoder, pre-trained on the ImageNet dataset, that solves for the tasks of Road Detection Learning, Road Orientation Learning, and Road Intersection Learning. Combining the capabilities of the U-Net model, the ResNet encoder and the constrained Multi-Task Learning mechanism, my model achieved better performance both in terms of image segmentation and topology preservation against the baseline single-task solving model. The project was based on the publicly available SpaceNet Roads Dataset.Geomatic
Core sample characterisation using 3D terrestrial laser scanning | TU Delft
No full textThe main purpose of this thesis is to conclude whether terrestrial laser scanning (TLS) can be used as an automated method to improve the classification of soils. It was in collaboration with Fugro N.V., a large Dutch consultancy and engineering firm that is active in onshore and offshore services in the field of geological and geotechnical investigations for a wide field of clients in the petroleum and gas industry but also for other infrastructure. Current methods for the classification of soils comprise of laboratory and in-situ measurements. Fugro is searching for a more accurate and time efficient method to classify soil and provided five soil samples on which a classification had to be performed. The Leica C10 laser scanner, provided by the department of Geoscience and Remote Sensing (TU Delft), was used to scan the soil samples. TLS is based on LIDAR (light detection and ranging) which is known for emitting pulses in a very narrow beam of monochromatic light (one wavelength) in the ultraviolet, visible or near-infrared range of the electromagnetic spectrum. After scanning the soil samples, the following features extracted from the 3D point cloud data were used to perform classification on each soil sample: intensity (backscattered energy), colour (an RGB-image), surface height variations (roughness). The chosen method was iso cluster unsupervised classification. After classification, interpretations were made based on the descriptions of the soil sample and knowledge of bare soil reflectance. The main factors that influenced the soil reflectance were related to characteristics of the samples: glauconite content (an iron-bearing mineral) and surface roughness. Both factors were responsible for a decrease in soil reflectance. Unsupervised classification was applicable on three out of five samples. However, only one sample provided good results. The other two samples were less promising, although some features could clearly be identified after comparing the classified image with ground truth data.Applied Earth Science
Random Forest Classification of three different species of trees in Delft, based on AHN point clouds: Additional Thesis
No full textTrees are an important aspect of the world around us, and play a sufficient role in our daily lives. They contribute to human health and well-being in various ways. Tree inventory and monitoring are of great interest for biomass estimations and changes in the purifying effect on the air. It is a very time consuming and cost inefficient way to check every tree in and around a city or town, therefore there is further research required in the use of AHN data. Together with the “tree information data set” formthemunicipality ofDelft, the location and the corresponding point cloud of tree different species of trees are selected. For the species of interest, Aesculus Hippocastanum, Acer Saccharinum and Platanus x Hispanica, different characteristics are determined. In this research six different characteristics are estimated; Height, Trunk Height, Normalized Trunk Height, Canopy Projected Area, Normalized Canopy Projected Area, Ratio of Diameters, Normalized Ratio of Diameter, Centre of Gravity and at least the Normalized Centre of Gravity. These characteristics are used as features for the Random Forest Classification, Consequently the Confusion Matrix is used as performance measurement. The results of a test of 30 pointclouds, per species of interest, show that the Random Forest Classification is able to classify individual trees. However, these three different species cannot by sufficiently classified using clustering
Automated photogrammetry of outcrops: with MicMac
No full textEvery year the second years of the bachelor Applied Earth Science have a geological fieldwork in Southern France. To supervise the students after a day in the field the idea is to collect a 3D model database of the outcrops in the field. This is done by photogrammetry and the open-source program MicMac.In this report an acquisition protocol and an automatic method for processing the images is developed. Students only need to collect images of their outcrop (following the given acquisition method) and upload them to the server. Some test cases are tested in this report and these results are given. The written workflow works on most test cases, except test cases made with a drone. The problems which arose were the right scale of the 3D models and sparse data at the boundaries of the outcrops. Although the workflow works in most cases, there are several adjustments to do in the future, like to make it more robust, find a method to create the right scales and to display all 3D models in one map. <br/
Remote Sensing of Japanese WWII airstrips in the Papua Province Republic of Indonesia: Classification of the area surrounding three WWII airstrips (Mongosah, Otawiri and Sagan)
No full textIn the Second World War Dutch New Guinea was a strategic battle front for both the Japanese and the Allied forces in the Pacific War. A lot of airstrips were constructed and bombed during this time, of which at least three (Mongosah, Otowari and Sagan) have never been visited after the war. This provided a great opportunity to find potential war heritage and airstrip equipment. Later this year an additional research team will go on an in-situ exploration to potentially find those objects. To do so, they needed a classification map giving information on the type and location of the vegetation. This map helps to know where to land with a helicopter, to setup base camp, to find travel ways, etc. Thus, the main objective of this thesis is to check whether it is possible to create a proper classification image with the available data. I used data obtained from the Sentinel 2 Mission (Optical data), the ALOS PALSAR Mission (L-Band Radar data) and the SRTM Mission (Digital elevation data). I pre-processed the data and used the supervised classification method, “Maximum Likelihood Classification” (MLC). I masked clouds via three different cloud masking methods, MLC Method, Threshold Method and Sen2cor (scene classification) Method. I compared the three different methods with each other and there is no significant difference between them. The classifications have been cross-validated with a reference validation dataset and the classified pixels are on average about 90% correctly classified
The influence of drone flightpath on photogrammetric model quality
No full textIn 2017 the Netherlands had 7146km of railways [Ramaekers et al., 2009], which are owned and managed by ProRail and have to be frequently surveyed, for quality inspection. Surveying is done using a variety of surveying techniques, many of which require surveyors to walk on or near the tracks, which could cause injuries. An alternative surveying technique would be photogrammetry using images acquired with a camera mounted on a drone. In this technique a 3D point cloud is constructed out of images, which is georeferenced using GPS measurements on the drone and Ground Control Points (GCP). The positions of the rail rods can then be found by modelling the rail rod using the point cloud. The quality of the point cloud, defined here as a combination of accuracy, precision, point density, completeness, and the rail rod feature detectability, is an important factor in deciding if photogrammetry is a sufficient surveying technique for this project.In this research the influence of the distance between the camera and the track, the baseline, and the number and spread of GCPs on the quality of the resulting point cloud. The distance to the track was set to 25, 30 and35m and the baseline was varied between 2 and 10m. The requirement set for the accuracy is 15mm, while the precision has to be below 10mm. For the point density an K-nearest neighbors (K-nn) of at least 10 is required, the completeness requires less than 10% missing cells and finally the feature detectability requires that at least once every 2m the location of the rail can be determined. The features are detected using a 2D model of the rail that is fitted in a flattened slice of the point cloud using a modified version of the iterative closest point algorithm.It is shown that the accuracy and precision are influenced by the distance and baseline, but no clear relation was found, while the accuracy improves when using GCPs. For the point density it is shown that increasing the distance between the camera and the track linearly decreases the point density, while the baseline has no impact and for the completeness a decreasing trend was found when increasing the baseline, while no relation could be found between the completeness and the distance between the camera and the track. For the rail rod feature detectability it was shown that the percentage of accurately found rail rod positions goes down when increasing the distance to the track and the baseline. Overall it is proven that when using a 100MP camera, 5 GCPs, a distance of 25m, and a baseline of 2.17m the rail rods could be surveyed so that the requirements set for this research were met.Civil Engineerin
- …
