Search

As urbanisation accelerates, the urban landscape reshapes at a fast pace. Consequently, the urban built environment continuously evolves horizontally as well as vertically. However, more attention in the field of spatial analysis is given to horizontal dynamics, despite the importance of geoinformation in the vertical dimension. 3D modelling methods have gained popularity due to their powerful capability of capturing and analysing geoinformation in the vertical dimension and visualising objects lifelike in the urban built environment. Various urban applications with diverse 3D modelling methods at different research scales and purposes have emerged. However, there is no systematic overview of these different modelling methods. Therefore, it is imperative to provide an up-to-date review of these advances. In this paper, we aim to review urban 3D modelling methods widely used in the prior 5-year period (2015–2020). Our analysis focuses on five attributes, i.e., basic characteristics, data requirements, technical requirements, users requirements and ethical considerations. The discussion presents the current status of 3D modelling methods – a wide range of applications yet with substantial development potential. This paper closes with insights for future work regarding the necessities of 3D data structure support as well as interdisciplinary research, specifically for big data management and integration.

This paper presents our contribution to the development of a standardized 3D input data model for solar photovoltaic potential estimation. Presently, different input data and processing steps influence the calculation for estimating the potential of solar energy in the Netherlands. The variety in characteristics of input data and issues with temporal accuracy extracted from the national registers and databases makes it challenging to obtain a consistent and reliable result. To address this issue, we created a point cloud dataset that integrated from LiDAR point cloud and dense image matching which is complete, recent and positionally accurate. Furthermore, we made a 3D building model from the integrated point cloud and identified the effect of finer resolution in the photovoltaic potential analysis.

Deep learning methods based on Fully convolution networks (FCNs) have shown an impressive progress in building outline delineation from very high resolution (VHR) remote sensing (RS) imagery. Common issues still exist in extracting precise building shapes and outlines, often resulting in irregular edges and over smoothed corners. In this paper, we use PolyMapper, a recently introduced deep-learning framework that is able to predict object outlines in a vector representation directly. We have introduced two main modifications to this baseline method. First, we introduce EffcientNet as backbone feature encoder to our network, which uses compound coefficient to scale up all dimensions of depth/width/resolution uniformly, to improve the processing speed with fewer parameters. Second, we integrate a boundary refinement block (BRB) to strengthen the boundary feature learning and to further improve the accuracy of corner prediction. The results demonstrate that the end-to-end learnable model is capable of delineating polygons of building outlines that closely approximate the structure of reference labels. Experiments on the crowdAI building instance segmentation datasets show that our model outperforms PolyMapper in all COCO metrics, for instance showing a 0.13 higher mean Average Precision (AP) value and a 0.60 higher mean Average Recall value. Also qualitative results show that our method segments building instances of various shapes more accurately.

One of the major challenges in precision viticulture in Europe is the detection and mapping of flavescence dorée (FD) grapevine disease to monitor and contain its spread. The lack of effective cures and the need for sustainable preventive measures are nowadays crucial issues. Insecticides and the plants uprooting are commonly employed to withhold disease infection, even if these solutions imply serious economic consequences and a strong environmental impact. The development of a rapid strategy to identify the disease is required to cover large portions of the crop and thus to limit damages in a time-effective way. This paper investigates the use of Unmanned Aerial Vehicles (UAVs), a cost-effective approach to early detection of diseased areas. We address this task with an object detection deep network, Faster R-CNN, instead of a traditional pixel-wise classifier. This work tests Faster R-CNN performance on this specific application through a comparative analysis with a pixel-wise classification algorithm (Random Forest). To take advantage of the full image resolution, the experimental analysis is performed using the original UAV imagery acquired in real conditions (instead of the derived orthomosaic). The first result of this paper is the definition of a new dataset for FD disease identification by UAV original imagery at the canopy scale. Moreover, we demonstrate the feasibility of applying Faster-R-CNN as a quasi-real-time alternative solution to semantic segmentation. The trained Faster-R-CNN achieved an average precision of 82% on the test set.

Dynamic Time Warping (DTW) has been successfully used for crops mapping due to its capability to achieve good classification results when a reduced number of training samples and irregular satellite image time series is available. Despite its recognized advantages, DTW does not account for the duration and seasonality of crops and local differences when assessing the similarity between two temporal sequences. In this study, we implemented a Weighted Derivative modification of DTW (WDDTW) and compared it with DTW and Time Weighted Dynamic Time Warping (TWDTW) for crops mapping. We show that WDDTW outperformed DTW achieving an overall accuracy of 67 %, whereas DTW obtained an accuracy of 57%. Yet, TWDTW performed better than both methods obtaining an accuracy of 88%.

Deforestation is a wide-reaching problem, responsible for serious environmental issues, such as biodiversity loss and global climate change. Containing approximately ten percent of all biomass on the planet and home to one tenth of the known species, the Amazon biome has faced important deforestation pressure in the last decades. Devising efficient deforestation detection methods is, therefore, key to combat illegal deforestation and to aid in the conception of public policies directed to promote sustainable development in the Amazon. In this work, we implement and evaluate a deforestation detection approach which is based on a Fully Convolutional, Deep Learning (DL) model: the DeepLabv3+. We compare the results obtained with the devised approach to those obtained with previously proposed DL-based methods (Early Fusion and Siamese Convolutional Network) using Landsat OLI-8 images acquired at different dates, covering a region of the Amazon forest. In order to evaluate the sensitivity of the methods to the amount of training data, we also evaluate them using varying training sample set sizes. The results show that all tested variants of the proposed method significantly outperform the other DL-based methods in terms of overall accuracy and F1-score. The gains in performance were even more substantial when limited amounts of samples were used in training the evaluated methods. © 2020 International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives.

Semantic segmentation models are often affected by illumination changes, and fail to predict correct labels. Although there has been a lot of research on indoor semantic segmentation, it has not been studied in low-light environments. In this paper we propose a new framework, LISU, for Low-light Indoor Scene Understanding. We first decompose the low-light images into reflectance and illumination components, and then jointly learn reflectance restoration and semantic segmentation. To train and evaluate the proposed framework, we propose a new data set, namely LLRGBD, which consists of a large synthetic low-light indoor data set (LLRGBD-synthetic) and a small real data set (LLRGBD-real). The experimental results show that the illumination-invariant features effectively improve the performance of semantic segmentation. Compared with the baseline model, the mIoU of the proposed LISU framework has increased by 11.5%. In addition, pre-training on our synthetic data set increases the mIoU by 7.2%. Our data sets and models are available on our project website.

Deep learning-based semantic segmentation models for building delineation face the challenge of producing precise and regular building outlines. Recently, a building delineation method based on frame field learning was proposed by Girard et al. (2020) to extract regular building footprints as vector polygons directly from aerial RGB images. A fully convolution network (FCN) is trained to learn simultaneously the building mask, contours, and frame field followed by a polygonization method. With the direction information of the building contours stored in the frame field, the polygonization algorithm produces regular outlines accurately detecting edges and corners. This paper investigated the contribution of elevation data from the normalized digital surface model (nDSM) to extract accurate and regular building polygons. The 3D information provided by the nDSM overcomes the aerial images’ limitations and contributes to distinguishing the buildings from the background more accurately. Experiments conducted in Enschede, the Netherlands, demonstrate that the nDSM improves building outlines’ accuracy, resulting in better-aligned building polygons and prevents false positives. The investigated deep learning approach (fusing RGB + nDSM) results in a mean intersection over union (IOU) of 0.70 in the urban area. The baseline method (using RGB only) results in an IOU of 0.58 in the same area. A qualitative analysis of the results shows that the investigated model predicts more precise and regular polygons for large and complex structures.

Updated information on urban land use allows city planners and decision makers to conduct large scale monitoring of urban areas for sustainable urban growth. Remote sensing data and classification methods offer an efficient and reliable way to update such land use maps. Features extracted from land cover maps are helpful on performing a land use classification task. Such prior information can be embedded in the design of a deep learning based land use classifier by applying a multitask learning setup—simultaneously solving a land use and a land cover classification task. In this study, we explore a fully convolutional multitask network to classify urban land use from very high resolution (VHR) imagery. We experimented with three different setups of the fully convolutional network and compared it against a baseline random forest classifier. The first setup is a standard network only predicting the land use class of each pixel in the image. The second setup is a multitask network that concatenates the land use and land cover class labels in the same output layer of the network while the other setup accept as an input the land cover predictions, predicted by a subpart of the network, concatenated to the original input image patches. The two deep multitask networks outperforms the other two classifiers by at least 30% in average F1-score.

Modern machine learning, especially deep learning, which is used in a variety of applications, requires a lot of labelled data for model training. Having an insufficient amount of training examples leads to models which do not generalize well to new input instances. This is a particular significant problem for tasks involving aerial images: often training data is only available for a limited geographical area and a narrow time window, thus leading to models which perform poorly in different regions, at different times of day, or during different seasons. Domain adaptation can mitigate this issue by using labelled source domain training examples and unlabeled target domain images to train a model which performs well on both domains. Modern adversarial domain adaptation approaches use unpaired data. We propose using pairs of semantically similar images, i.e., whose segmentations are accurate predictions of each other, for improved model performance. In this paper we show that, as an upper limit based on ground truth, using semantically paired aerial images during training almost always increases model performance with an average improvement of 4.2% accuracy and .036 mean intersection-over-union (mIoU). Using a practical estimate of semantic similarity, we still achieve improvements in more than half of all cases, with average improvements of 2.5% accuracy and .017 mIoU in those cases.

Land use (LU) is an important information source commonly stored in geospatial databases. Most current work on automatic LU classification for updating topographic databases considers only one category level (e.g. residential or agricultural) consisting of a small number of classes. However, LU databases frequently contain very detailed information, using a hierarchical object catalogue where the number of categories differs depending on the hierarchy level. This paper presents a method for the classification of LU on the basis of aerial images that differentiates a fine-grained class structure, exploiting the hierarchical relationship between categories at different levels of the class catalogue. Starting from a convolutional neural network (CNN) for classifying the categories of all levels, we propose a strategy to simultaneously learn the semantic dependencies between different category levels explicitly. The input to the CNN consists of aerial images and derived data as well as land cover information derived from semantic segmentation. Its output is the class scores at three different semantic levels, based on which predictions that are consistent with the class hierarchy are made. We evaluate our method using two test sites and show how the classification accuracy depends on the semantic category level. While at the coarsest level, an overall accuracy in the order of 90% can be achieved, at the finest level, this accuracy is reduced to around 65%. Our experiments also show which classes are particularly hard to differentiate. © 2020 Copernicus GmbH. All rights reserved.

Many countries were the target of air strikes during World War II. Numerous unexploded bombs still exist in the ground. These duds can be tracked down with the help of bomb craters, indicating areas where unexploded bombs may be located. Such areas are documented in so-called impact maps based on detected bomb craters. In this paper, a stochastic approach based on marked point processes (MPPs) for the automatic detection of bomb craters in aerial images taken during World War II is presented. As most areas are covered by multiple images, the influence of redundant image information on the object detection result is investigated: we compare the results generated based on single images with those obtained by our new approach that combines the individual detection results of multiple images covering the same location. The object model for the bomb craters is represented by circles. Our MPP approach determines the most likely configuration of objects within the scene. The goal is reached by minimizing an energy function that describes the conformity with a predefined model by Reversible Jump Markov Chain Monte Carlo sampling in combination with simulated annealing. Afterwards, a probability map is generated from the automatic detections via kernel density estimation. By setting a threshold, areas around the detections are classified as contaminated or uncontaminated sites, respectively, which results in an impact map. Our results show a significant improvement with respect to its quality when redundant image information is used.

Deep learning classification models require large amounts of labeled training data to perform properly, but the production of reference data for most Earth observation applications is a labor intensive, costly process. In that sense, transfer learning is an option to mitigate the demand for labeled data. In many remote sensing applications, however, the accuracy of a deep learning-based classification model trained with a specific dataset drops significantly when it is tested on a different dataset, even after fine-tuning. In general, this behavior can be credited to the domain shift phenomenon. In remote sensing applications, domain shift can be associated with changes in the environmental conditions during the acquisition of new data, variations of objects' appearances, geographical variability and different sensor properties, among other aspects. In recent years, deep learning-based domain adaptation techniques have been used to alleviate the domain shift problem. Recent improvements in domain adaptation technology rely on techniques based on Generative Adversarial Networks (GANs), such as the Cycle-Consistent Generative Adversarial Network (CycleGAN), which adapts images across different domains by learning nonlinear mapping functions between the domains. In this work, we exploit the CycleGAN approach for domain adaptation in a particular change detection application, namely, deforestation detection in the Amazon forest. Experimental results indicate that the proposed approach is capable of alleviating the effects associated with domain shift in the context of the target application. © 2020 International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives.

Pixel-based land cover classification of aerial images is a standard task in remote sensing, whose goal is to identify the physical material of the earth’s surface. Recently, most of the well-performing methods rely on encoder-decoder structure based convolutional neural networks (CNN). In the encoder part, many successive convolution and pooling operations are applied to obtain features at a lower spatial resolution, and in the decoder part these features are up-sampled gradually and layer by layer, in order to make predictions in the original spatial resolution. However, the loss of spatial resolution caused by pooling affects the final classification performance negatively, which is compensated by skip-connections between corresponding features in the encoder and the decoder. The most popular ways to combine features are element-wise addition of feature maps and 1x1 convolution. In this work, we investigate skip-connections. We argue that not every skip-connections are equally important. Therefore, we conducted experiments designed to find out which skip-connections are important. Moreover, we propose a new cosine similarity loss function to utilize the relationship of the features of the pixels belonging to the same category inside one mini-batch, i.e. these features should be close in feature space. Our experiments show that the new cosine similarity loss does help the classification. We evaluated our methods using the Vaihingen and Potsdam dataset of the ISPRS 2D semantic labelling challenge and achieved an overall accuracy of 91.1% for both test sites.

Fully convolutional neural networks (FCN) are successfully used for the automated pixel-wise classification of aerial images and possibly additional data. However, they require many labelled training samples to perform well. One approach addressing this issue is semi-supervised domain adaptation (SSDA). Here, labelled training samples from a source domain and unlabelled samples from a target domain are used jointly to obtain a target domain classifier, without requiring any labelled samples from the target domain. In this paper, a two-step approach for SSDA is proposed. The first step corresponds to a supervised training on the source domain, making use of strong data augmentation to increase the initial performance on the target domain. Secondly, the model is adapted by entropy minimization using a novel weighting strategy. The approach is evaluated on the basis of five domains, corresponding to five cities. Several training variants and adaptation scenarios are tested, indicating that proper data augmentation can already improve the initial target domain performance significantly resulting in an average overall accuracy of 77.5%. The weighted entropy minimization improves the overall accuracy on the target domains in 19 out of 20 scenarios on average by 1.8%. In all experiments a novel FCN architecture is used that yields results comparable to those of the best-performing models on the ISPRS labelling challenge while having an order of magnitude fewer parameters than commonly used FCNs.

This paper proposes several methods for training a Convolutional Neural Network (CNN) for learning the similarity between images of silk fabrics based on multiple semantic properties of the fabrics. In the context of the EU H2020 project SILKNOW (http://silknow.eu/), two variants of training were developed, one based on a Siamese CNN and one based on a triplet architecture. We propose different definitions of similarity and different loss functions for both training strategies, some of them also allowing the use of incomplete information about the training data. We assess the quality of the trained model by using the learned image features in a k-NN classification. We achieve overall accuracies of 93–95% and average F1-scores of 87–92%.

State-of-the-art object detection approaches such as Fast/Faster R-CNN, SSD, or YOLO have difficulties detecting dense, small targets with arbitrary orientation in large aerial images. The main reason is that using interpolation to align RoI features can result in a lack of accuracy or even loss of location information. We present the Local-aware Region Convolutional Neural Network (LR-CNN), a novel two-stage approach for vehicle detection in aerial imagery. We enhance translation invariance to detect dense vehicles and address the boundary quantization issue amongst dense vehicles by aggregating the high-precision RoIs' features. Moreover, we resample high-level semantic pooled features, making them regain location information from the features of a shallower convolutional block. This strengthens the local feature invariance for the resampled features and enables detecting vehicles in an arbitrary orientation. The local feature invariance enhances the learning ability of the focal loss function, and the focal loss further helps to focus on the hard examples. Taken together, our method better addresses the challenges of aerial imagery. We evaluate our approach on several challenging datasets (VEDAI, DOTA), demonstrating a significant improvement over state-of-the-art methods. We demonstrate the good generalization ability of our approach on the DLR 3K dataset., Comment: 8 pages

Pixel-wise classification of remote sensing imagery is highly interesting for tasks like land cover classification or change detection. The acquisition of large training data sets for these tasks is challenging, but necessary to obtain good results with deep learning algorithms such as convolutional neural networks (CNN). In this paper we present a method for the automatic generation of a large amount of training data by combining satellite imagery with reference data from an available geospatial database. Due to this combination of different data sources the resulting training data contain a certain amount of incorrect labels. We evaluate the influence of this so called label noise regarding the time difference between acquisition of the two data sources, the amount of training data and the class structure. We combine Sentinel-2 images with reference data from a geospatial database provided by the German Land Survey Office of Lower Saxony (LGLN). With different training sets we train a fully convolutional neural network (FCN) and classify four land cover classes (code Building, Agriculture, Forest, Water/code). Our results show that the errors in the training samples do not have a large influence on the resulting classifiers. This is probably due to the fact that the noise is randomly distributed and thus, neighbours of incorrect samples are predominantly correct. As expected, a larger amount of training data improves the results, especially for the less well represented classes. Other influences are different illuminations conditions and seasonal effects during data acquisition. To better adapt the classifier to these different conditions they should also be included in the training data. © 2020 International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives.

This paper proposes a novel approach for linear feature detection. The contribution is twofold: a novel model for spatial point processes and a new method for linear feature detection. It describes a linear feature as a string of points, represents all features in an image as a configuration of a spatial point process, and formulates feature detection as finding the optimal configuration of a spatial point process. Further, a prior term is proposed to favor straight linear configurations, and a data term is constructed to superpose the points on linear features. The proposed approach extracts straight linear features in a global framework. The paper reports ongoing work. As demonstrated in preliminary experiments, globally optimal linear features can be detected. National Natural Science Foundation of China/41071263 National Natural Science Foundation of China/41571335 Zhejiang Provincial Natural Science Foundation of China/LY13D010003 Key Laboratory for National Geographic Census and Monitoring National Administration of Surveying, Mapping and Geoinformation/2014NGCM

We propose a novel hierarchical approach for the classification of airborne 3D lidar points. Spatial and semantic context is incorporated via a two-layer Conditional Random Field (CRF). The first layer operates on a point level and utilises higher order cliques. Segments are generated from the labelling obtained in this way. They are the entities of the second layer, which incorporates larger scale context. The classification result of the segments is introduced as an energy term for the next iteration of the point-based layer. This framework iterates and mutually propagates context to improve the classification results. Potentially wrong decisions can be revised at later stages. The output is a labelled point cloud as well as segments roughly corresponding to object instances. Moreover, we present two new contextual features for the segment classification: the distance and the orientation of a segment with respect to the closest road. It is shown that the classification benefits from these features. In our experiments the hierarchical framework improve the overall accuracies by 2.3% on a point-based level and by 3.0% on a segment-based level, respectively, compared to a purely point-based classification.

Online multi-person tracking in image sequences is commonly guided by recursive filters, whose predictive models define the expected positions of future states. When a predictive model deviates too much from the true motion of a pedestrian, which is often the case in crowded scenes due to unpredicted accelerations, the data association is prone to fail. In this paper we propose a novel predictive model on the basis of Gaussian Process Regression. The model takes into account the motion of every tracked pedestrian in the scene and the prediction is executed with respect to the velocities of all interrelated persons. As shown by the experiments, the model is capable of yielding more plausible predictions even in the presence of mutual occlusions or missing measurements. The approach is evaluated on a publicly available benchmark and outperforms other state-of-the-art trackers.

In this paper we describe learning of a descriptor based on the Siamese Convolutional Neural Network (CNN) architecture and evaluate our results on a standard patch comparison dataset. The descriptor learning architecture is composed of an input module, a Siamese CNN descriptor module and a cost computation module that is based on the L2 Norm. The cost function we use pulls the descriptors of matching patches close to each other in feature space while pushing the descriptors for non-matching pairs away from each other. Compared to related work, we optimize the training parameters by combining a moving average strategy for gradients and Nesterov's Accelerated Gradient. Experiments show that our learned descriptor reaches a good performance and achieves state-of-art results in terms of the false positive rate at a 95 % recall rate on standard benchmark datasets.

Tracking-by-detection is a widely used practice in recent tracking systems. These usually rely on independent single frame detections that are handled as observations in a recursive estimation framework. If these observations are imprecise the generated trajectory is prone to be updated towards a wrong position. In contrary to existing methods our novel approach uses a Dynamic Bayes Network in which the state vector of a recursive Bayes filter, as well as the location of the tracked object in the image are modelled as unknowns. These unknowns are estimated in a probabilistic framework taking into account a dynamic model, and a state-of-the-art pedestrian detector and classifier. The classifier is based on the Random Forest-algorithm and is capable of being trained incrementally so that new training samples can be incorporated at runtime. This allows the classifier to adapt to the changing appearance of a target and to unlearn outdated features. The approach is evaluated on a publicly available benchmark. The results confirm that our approach is well suited for tracking pedestrians over long distances while at the same time achieving comparatively good geometric accuracy.

In order to tackle the problem of consistently integrating 2D vector data and a DTM, we presented an approach for the adaptation of 2D GIS road objects to airborne laser scanning (ALS) data using active contours (snakes) in (Göpfert et al., 2011). In this paper the algorithm is modified for the integration of stereo images as an alternative data source for area-wide height information. For that reason, a new image energy is developed that exploits geometric and radiometric features derived from the image data. Afterwards, we compare the applicability of our method with respect to the ALS data and stereo images as input. In addition, a new approach is suggested that analyses the different energy terms of active contours after the optimisation process in order to automatically detect contour parts that did not reach a suitable position in the sensor data. This concept of an internal evaluation is able to guide the user during post processing. Experiments show that the snake approach with an image energy based on stereo images is generally able to adapt GIS road centrelines to the sensor data and thus to improve the quality of the 2D vector data. However, the comparison to the results for ALS data demonstrates that the algorithm perform slightly worse for image data in the high precision level.

In this paper, we propose a new framework for segmenting feature-based moving objects under affine subspace model. Since the feature trajectories in practice are high-dimensional and contain a lot of noise, we firstly apply the sparse PCA to represent the original trajectories with a low-dimensional global subspace, which consists of the orthogonal sparse principal vectors. Subsequently, the local subspace separation will be achieved via automatically searching the sparse representation of the nearest neighbors for each projected data. In order to refine the local subspace estimation result, we propose an error estimation to encourage the projected data that span a same local subspace to be clustered together. In the end, the segmentation of different motions is achieved through the spectral clustering on an affinity matrix, which is constructed with both the error estimation and sparse neighbors optimization. We test our method extensively and compare it with state-of-the-art methods on the Hopkins 155 dataset. The results show that our method is comparable with the other motion segmentation methods, and in many cases exceed them in terms of precision and computation time.

International audience; Quantum geographic information system (QGIS) is a free, open-source, cross-platform and scalable geographic information system (GIS) tool with plugin development in Python and C++ languages. Technically, QGIS integrates the geospatial data abstraction library (GDAL), which allows it to read and process a large number of geographic images. Distributed under the GNU/general public license (GPL) version 2, QGIS allows free access to a powerful, inexpensive GIS program that can be used on most platforms: GNU/Linux, Unix, Mac OS X and Windows. During the various updates of the QGIS version, the processing module was constantly improved by bug fixes and the addition of new functions. The processing module is proposed by default with the QGIS software and does not require the installation of an extension in order to be used. To use it, the QGIS graphical user interface (GUI) offers a "Processing" menu.

International audience; This chapter presents the various treatments offered by QGIS for network analysis and routing. After a quick definition of networks and graphs, it discusses the rules for construction and verification of networks using GIS. The chapter focuses on the presentation of examples of hydrographic network development and analysis. It consists of several practical works focused on a hydrographic network analysis and routing as part of a global project. The topology of GIS is expressed by a set of rules dealing with relations between spatial entities. In the case of networks, the notion of adjacency or contiguity implies that the spatial units have a common node. The use of a hydrographic networks allow to integrate spatially the physical parameters related to topography, geography, climate and hydrology and to accumulate them downstream and thus provide useful data in the fields of hydrology, geomorphology or ecology.

driving. While much progress has been made in recent years, imaging conditions in natural outdoor environments are still very challenging for current reconstruction and recognition methods. In this paper, we propose a novel unified approach which reasons jointly about 3D scene flow as well as the pose, shape and motion of vehicles in the scene. Towards this goal, we incorporate a deformable CAD model into a slanted-plane conditional random field for scene flow estimation and enforce shape consistency between the rendered 3D models and the parameters of all superpixels in the image. The association of superpixels to objects is established by an index variable which implicitly enables model selection. We evaluate our approach on the challenging KITTI scene flow dataset in terms of object and scene flow estimation. Our results provide a prove of concept and demonstrate the usefulness of our method.

Complex activity modeling and identification of anomaly is one of the most interesting and desired capabilities for automated video behavior analysis. A number of different approaches have been proposed in the past to tackle this problem. There are two main challenges for activity modeling and anomaly detection: 1) most existing approaches require sufficient data and supervision for learning; 2) the most interesting abnormal activities arise rarely and are ambiguous among typical activities, i.e. hard to be precisely defined. In this paper, we propose a novel approach to model complex activities and detect anomalies by using non-parametric Gaussian Process (GP) models in a crowded and complicated traffic scene. In comparison with parametric models such as HMM, GP models are nonparametric and have their advantages. Our GP models exploit implicit spatial-temporal dependence among local activity patterns. The learned GP regression models give a probabilistic prediction of regional activities at next time interval based on observations at present. An anomaly will be detected by comparing the actual observations with the prediction at real time. We verify the effectiveness and robustness of the proposed model on the QMUL Junction Dataset. Furthermore, we provide a publicly available manually labeled ground truth of this data set.

Land cover and land use exhibit strong contextual dependencies. We propose a novel approach for the simultaneous classification of land cover and land use, where semantic and spatial context is considered. The image sites for land cover and land use classification form a hierarchy consisting of two layers: a land cover layer and a land use layer. We apply Conditional Random Fields (CRF) at both layers. The layers differ with respect to the image entities corresponding to the nodes, the employed features and the classes to be distinguished. In the land cover layer, the nodes represent super-pixels; in the land use layer, the nodes correspond to objects from a geospatial database. Both CRFs model spatial dependencies between neighbouring image sites. The complex semantic relations between land cover and land use are integrated in the classification process by using contextual features. We propose a new iterative inference procedure for the simultaneous classification of land cover and land use, in which the two classification tasks mutually influence each other. This helps to improve the classification accuracy for certain classes. The main idea of this approach is that semantic context helps to refine the class predictions, which, in turn, leads to more expressive context information. Thus, potentially wrong decisions can be reversed at later stages. The approach is designed for input data based on aerial images. Experiments are carried out on a test site to evaluate the performance of the proposed method. We show the effectiveness of the iterative inference procedure and demonstrate that a smaller size of the super-pixels has a positive influence on the classification result.

In this paper we present a robust orientation approach for an imaging sensor flown on a micro-UAV based on image triplets. Our aim is to have the orientation available online, i.e. during image acquisition. The resulting point cloud and sensor orientations can then for instance be evaluated for navigation purposes of the UAV or to analyse the completeness of the point cloud. We use low quality imagery extracted from the downlink of an onboard PAL-camera. Trilinear constraints and cross-checked matches allow for a high robustness of the sensor orientation and the sparse 3D point cloud. In order to reach the goal of on-line processing given the large number of observations and unknowns, we make use of an incremental bundle adjustment. Estimated parameters are incrementally improved without explicitly considering previous observations. Our approach combines linear projective geometry for obtaining initial values using the trifocal tensor with non-linear perspective geometry for the estimation of the unknowns. This combination allows for a high precision of the estimation, while eliminating the need for initial values. We evaluate the performance of our approach by means of imagery we acquired of the facade of theWelfenschloss in Hannover, collected with a Microdrones md4-200 micro-UAV. The results are the orientation parameters of the images and a sparse 3D point cloud representing the object. They are compared to those from a commercial bundle adjustment software and analysed in terms of geometric precision.

© 2017 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY license Background Continuous electronic fetal heart-rate monitoring is widely used during labour, and computerised interpretation could increase its usefulness. We aimed to establish whether the addition of decision-support software to assist in the interpretation of cardiotocographs affected the number of poor neonatal outcomes. Methods In this unmasked randomised controlled trial, we recruited women in labour aged 16 years or older having continuous electronic fetal monitoring, with a singleton or twin pregnancy, and at 35 weeks' gestation or more at 24 maternity units in the UK and Ireland. They were randomly assigned (1:1) to decision support with the INFANT system or no decision support via a computer-generated stratified block randomisation schedule. The primary outcomes were poor neonatal outcome (intrapartum stillbirth or early neonatal death excluding lethal congenital anomalies, or neonatal encephalopathy, admission to the neonatal unit within 24 h for ≥48 h with evidence of feeding difficulties, respiratory illness, or encephalopathy with evidence of compromise at birth), and developmental assessment at age 2 years in a subset of surviving children. Analyses were done by intention to treat. This trial is completed and is registered with the ISRCTN Registry, number 98680152. Findings Between Jan 6, 2010, and Aug 31, 2013, 47 062 women were randomly assigned (23 515 in the decision-support group and 23 547 in the no-decision-support group) and 46 042 were analysed (22 987 in the decision-support group and 23 055 in the no-decision-support group). We noted no difference in the incidence of poor neonatal outcome between the groups—172 (0·7%) babies in the decision-support group compared with 171 (0·7%) babies in the no-decision-support group (adjusted risk ratio 1·01, 95% CI 0·82–1·25). At 2 years, no significant differences were noted in terms of developmental assessment. Interpretation Use of computerised interpretation of cardiotocographs in women who have continuous electronic fetal monitoring in labour does not improve clinical outcomes for mothers or babies. Funding National Institute for Health Research.

Domain adaptation techniques in transfer learning try to reduce the amount of training data required for classification by adapting a classifier trained on samples from a source domain to a new data set (target domain) where the features may have different distributions. In this paper, we propose a new technique for domain adaptation based on logistic regression. Starting with a classifier trained on training data from the source domain, we iteratively include target domain samples for which class labels have been obtained from the current state of the classifier, while at the same time removing source domain samples. In each iteration the classifier is re-trained, so that the decision boundaries are slowly transferred to the distribution of the target features. To make the transfer procedure more robust we introduce weights as a function of distance from the decision boundary and a new way of regularisation. Our methodology is evaluated using a benchmark data set consisting of aerial images and digital surface models. The experimental results show that in the majority of cases our domain adaptation approach can lead to an improvement of the classification accuracy without additional training data, but also indicate remaining problems if the difference in the feature distributions becomes too large.

Figure-ground image segmentation has been a challenging problem in computer vision. Apart from the difficulties in establishing an effective framework to divide the image pixels into meaningful groups, the notions of figure and ground often need to be properly defined by providing either user inputs or object models. In this paper, we propose a novel graph-based segmentation framework, called superpixel cut. The key idea is to formulate foreground segmentation as finding a subset of superpixels that partitions a graph over superpixels. The problem is formulated as Min-Cut. Therefore, we propose a novel cost function that simultaneously minimizes the inter-class similarity while maximizing the intra-class similarity. This cost function is optimized using parametric programming. After a small learning step, our approach is fully automatic and fully bottom-up, which requires no high-level knowledge such as shape priors and scene content. It recovers coherent components of images, providing a set of multiscale hypotheses for high-level reasoning. We evaluate our proposed framework by comparing it to other generic figure-ground segmentation approaches. Our method achieves improved performance on state-of-the-art benchmark databases.

The classification of building facade images is a challenging problem that receives a great deal of attention in the photogrammetry community. Image classification is critically dependent on the features. In this paper, we perform an empirical feature evaluation task for building facade images. Feature sets we choose are basic features, color features, histogram features, Peucker features, texture features, and SIFT features. We present an approach for region-wise labeling using an efficient randomized decision forest classifier and local features. We conduct our experiments with building facade image classification on the eTRIMS dataset, where our focus is the object classes building, car, door, pavement, road, sky, vegetation, and window.

The fully automated analysis of 3D point clouds is of great importance in photogrammetry, remote sensing and computer vision. For reliably extracting objects such as buildings, road inventory or vegetation, many approaches rely on the results of a point cloud classification, where each 3D point is assigned a respective semantic class label. Such an assignment, in turn, typically involves statistical methods for feature extraction and machine learning. Whereas the different components in the processing workflow have extensively, but separately been investigated in recent years, the respective connection by sharing the results of crucial tasks across all components has not yet been addressed. This connection not only encapsulates the interrelated issues of neighborhood selection and feature extraction, but also the issue of how to involve spatial context in the classification step. In this paper, we present a novel and generic approach for 3D scene analysis which relies on (i) individually optimized 3D neighborhoods for (ii) the extraction of distinctive geometric features and (iii) the contextual classification of point cloud data. For a labeled benchmark dataset, we demonstrate the beneficial impact of involving contextual information in the classification process and that using individual 3D neighborhoods of optimal size significantly increases the quality of the results for both pointwise and contextual classification.

In this paper we address the problem of classification of remote sensing images in the framework of transfer learning with a focus on domain adaptation. The main novel contribution is a method for transductive transfer learning in remote sensing on the basis of logistic regression. Logistic regression is a discriminative probabilistic classifier of low computational complexity, which can deal with multiclass problems. This research area deals with methods that solve problems in which labelled training data sets are assumed to be available only for a source domain, while classification is needed in the target domain with different, yet related characteristics. Classification takes place with a model of weight coefficients for hyperplanes which separate features in the transformed feature space. In term of logistic regression, our domain adaptation method adjusts the model parameters by iterative labelling of the target test data set. These labelled data features are iteratively added to the current training set which, at the beginning, only contains source features and, simultaneously, a number of source features are deleted from the current training set. Experimental results based on a test series with synthetic and real data constitutes a first proof-of-concept of the proposed method.

We present an interactive system to manipulate a virtual object by tracking multiple hands in 3D space using a Kinect device. The system segments hand shapes from a captured 3D scene by using depth information and active contours. A hand shape is recognized by a trained naive Bayes classifier, whether it belongs to a palm, a pointing hand form or both hands with simple occlusion. A plane is approximated by using RANSAC for a palm hand form, while a vector from the hand centroid to the fingertip is obtained for a pointing hand form by using ICP as tracking method. A shape of simple occluded two hands is split into a palm and an incomplete pointing hand form, whose missing data is estimated by using PCA. The system works in semi-real time.

This paper describes a method for automatic verification of 3D building models using airborne oblique images. The problem being tackled is identifying buildings that are demolished or changed since the models were constructed or identifying wrong models using the images. The models verified are of CityGML LOD2 or higher since their edges are expected to coincide with actual building edges. The verification approach is based on information theory. Corresponding variables between building models and oblique images are used for deriving mutual information for individual edges, faces or whole buildings, and combined for all perspective images available for the building. The wireframe model edges are projected to images and verified using low level image features – the image pixel gradient directions. A building part is only checked against images in which it may be visible. The method has been tested with models constructed using laser points against Pictometry images that are available for most cities of Europe and may be publically viewed in the so called Birds Eye view of the Microsoft Bing Maps. Results are that nearly all buildings are correctly categorised as existing or demolished. Because we now concentrate only on roofs we also used the method to test and compare results from nadir images. This comparison made clear that especially height errors in models can be more reliably detected in oblique images because of the tilted view. Besides overall building verification, results per individual edges can be used for improving the 3D building models.

Video surveillance systems are no longer a collection of independent cameras, manually controlled by human operators. Instead, smart sensor networks are developed, able to fulfil certain tasks on their own and thus supporting security personnel by automated analyses. One well-known task is the derivation of people's positions on a given ground plane from monocular video footage. An improved accuracy for the ground position as well as a more detailed representation of single salient people can be expected from a stereoscopic processing of overlapping views. Related work mostly relies on dedicated stereo devices or camera pairs with a small baseline. While this set-up is helpful for the essential step of image matching, the high accuracy potential of a wide baseline and the according good intersection geometry is not utilised. In this paper we present a stereoscopic approach, working on overlapping views of standard pan-tilt-zoom cameras which can easily be generated for arbitrary points of interest by an appropriate reconfiguration of parts of a sensor network. Experiments are conducted on realistic surveillance footage to show the potential of the suggested approach and to investigate the influence of different baselines on the quality of the derived surface model. Promising estimations of people's position and height are retrieved. Although standard matching approaches show helpful results, future work will incorporate temporal dependencies available from image sequences in order to reduce computational effort and improve the derived level of detail. BMBF/13N10809 BMBF/13N10810 BMBF/13N10811 BMBF/13N10812 BMBF/13N10813 BMBF/13N10814

Detection of buildings and vegetation, and even more reconstruction of urban terrain from sequences of aerial images and videos is known to be a challenging task. It has been established that those methods that have as input a high-quality Digital Surface Model (DSM), are more straight-forward and produce more robust and reliable results than those image-based methods that require matching line segments or even whole regions. This motivated us to develop a new dense matching technique for DSM generation that is capable of simultaneous integration of multiple images in the reconstruction process. The DSMs generated by this new multi-image matching technique can be used for urban object extraction. In the first contribution of this paper, two examples of external sources of information added to the reconstruction pipeline will be shown. The GIS layers are used for recognition of streets and suppressing false alarms in the depth maps that were caused by moving vehicles while the near infrared channel is applied for separating vegetation from buildings. Three examples of data sets including both UAV-borne video sequences with a relatively high number of frames and high-resolution (10 cm ground sample distance) data sets consisting of (few spatial-temporarily diverse) images from large-format aerial frame cameras, will be presented. By an extensive quantitative evaluation of the Vaihingen block from the ISPRS benchmark on urban object detection, it will become clear that our procedure allows a straight-forward, efficient, and reliable instantiation of 3D city models.

Our work addresses the problem of long-term visual people tracking in complex environments. Tracking a varying number of objects entails the problem of associating detected objects to tracked targets. To overcome the data association problem, we apply a Tracking-by-Detection strategy that uses Randomized Forests as a classifier together with a Kalman filter. Randomized Forests build a strong classifier for multi-class problems through aggregating simple decision trees. Due to their modular setup, Randomized Forests can be built incrementally, which makes them useful for unsupervised learning of object features in real-time. New training samples can be incorporated on the fly, while not drifting away from previously learnt features. To support further analysis of the automatically generated trajectories, we annotate them with quality metrics based on the association confidence. To build the metrics we analyse the confidence values that derive from the Randomized Forests and the similarity of detected and tracked objects. We evaluate the performance of the overall approach with respect to available reference data of people crossing the scene.

The precise classification and reconstruction of crossroads from multiple aerial images is a challenging problem in remote sensing. We apply the Markov Random Fields (MRF) approach to this problem, a probabilistic model that can be used to consider context in classification. A simple appearance-based model is combined with a probabilistic model of the co-occurrence of class label at neighbouring image sites to distinguish up to 14 different classes that are relevant for scenes containing crossroads. The parameters of these models are learnt from training data. We use multiple overlap aerial images to derive a digital surface model (DSM) and a true orthophoto without moving cars. From the DSM and the orthophoto we derive feature vectors that are used in the classification. One of the features is a car confidence value that is supposed to support the classification when the road surface is occluded by static cars. Our approach is evaluated on a dataset of airborne photos of an urban area by a comparison of the results to reference data. Whereas the method has problems in distinguishing classes having a similar appearance, it is shown to produce promising results if a reduced set of classes is considered, yielding an overall classification accuracy of 74.8%.

In this paper we propose a probabilistic supervised classification algorithm for LiDAR (Light Detection And Ranging) point clouds. Several object classes (i.e. ground, building and vegetation) can be separated reliably by considering each point's neighbourhood. Based on Conditional Random Fields (CRF) this contextual information can be incorporated into classification process in order to improve results. Since we want to perform a point-wise classification, no primarily segmentation is needed. Therefore, each 3D point is regarded as a graph's node, whereas edges represent links to the nearest neighbours. Both nodes and edges are associated with features and have effect on the classification. We use some features available from full waveform technology such as amplitude, echo width and number of echoes as well as some extracted geometrical features. The aim of the paper is to describe the CRF model set-up for irregular point clouds, present the features used for classification, and to discuss some results. The resulting overall accuracy is about 94 %.

Hyperspectral imaging, with its applications, offers promising tools for remote sensing and Earth observation. Recent development has increased the quality of the sensors. At the same time, the prices of the sensors are lowering. Anomaly detection is one of the popular remote sensing applications, which benefits from real-time solutions. A real-time solution has its limitations, for example, due to a large amount of hyperspectral data, platform’s (drones or a cube satellite) constraints on payload and processing capability. Other examples are the limitations of available energy and the complexity of the machine learning models. When anomalies are detected in real-time from the hyperspectral images, one crucial factor is to utilise a computationally efficient method. The Minimal Learning Machine is a distance-based classification algorithm, which can be modified for anomaly detection. Earlier studies confirms that the Minimal learning Machine (MLM) is capable of detecting efficiently global anomalies from the hyperspectral images with a false alarm rate of zero. In this study, we will show that by using a carefully selected lower threshold besides the higher threshold of the variance, it is possible to detect local and global anomalies with the MLM. The downside is that the improved method is highly sensitive with the respect to the noise. Thus, the second aim of this study is to improve the MLM’s robustness with respect to noise by introducing a novel approach, the piecewise MLM. With the new approach, the piecewise MLM can detect global and local anomalies, and the method is significantly more robust with respect to noise than the MLM. As a result, we have an interesting, easy to implement and computationally light method which is suitable for remote sensing applications.

Global warming is getting more and more attention, and one of the main reasons is the rapid growth of the surface temperature. Today more than fifty percent of the human population is living in the cities, and with the growing tendency by the year 2050 to seventy percent, urbanisation is becoming one of the main factors of global warming. Conducted research indicates that the heat island effect is increasing rapidly in a territory with fast urbanisation growth. The surface temperature and air temperature are increasing in surrounding rural areas. This research aims to analyze urbanisation's impact on creating heat islands in a specific time frame from 1990 until 2020 in five large cities with satellite imagery obtained from Landsat satellite missions. By conducting the land cover classification and interpretation of vegetation indexes with the help of GEE (Google Earth Engine) and GIS tools as well as their changes in time by using different channels of satellite imagery will try to show the correlation between urbanisation and heat islands. This research is important for the professional and academic society because the control of global warming and increasing the quality of life in cities is today's priority.

Classification and segmentation of buildings from airborne lidar point clouds commonly involve point features calculated within a local neighborhood. The relative change of the features in the immediate surrounding of each point as well as the spatial relationships between neighboring points also need to be examined to account for spatial coherence. In this study we formulate the point labeling problem under a global graph-cut optimization solution. We construct the energy function through a graph representing a Markov Random Field (MRF). The solution to the labeling problem is obtained by finding the minimum-cut on this graph. We have employed this framework for three different labeling tasks on airborne lidar point clouds. Ground filtering, building classification, and roof-plane segmentation. As a follow-up study on our previous ground filtering work, this paper examines our building extraction approach on two airborne lidar datasets with different point densities containing approximately 930K points in one dataset and 750K points in the other. Test results for building vs. non-building point labeling show a 97.9% overall accuracy with a kappa value of 0.91 for the dataset with 1.18 pts/m2 average point density and a 96.8% accuracy with a kappa value of 0.90 for the dataset with 8.83 pts/m2 average point density. We can achieve 91.2% overall average accuracy in roof plane segmentation with respect to the reference segmentation of 20 building roofs involving 74 individual roof planes. In summary, the presented framework can successfully label points in airborne lidar point clouds with different characteristics for all three labeling problems we have introduced. It is robust to noise in the calculated features due to the use of global optimization. Furthermore, the framework achieves these results with a small training sample size. © 2020 Copernicus GmbH. All rights reserved., International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, V, ISSN:1682-1750, ISSN:2194-9034, ISSN:1682-1777

Anomaly detection from hyperspectral data needs computationally efficient methods to process the data when the data gathering platform is a drone or a cube satellite. In this study, we introduce a minimal learning machine for hyperspectral anomaly detection. Minimal learning machine is a novel distance-based classification algorithm, which is now modified to detect anomalies. Besides being computationally efficient, minimal learning machine is also easy to implement. Based on the results, we show that minimal learning machine is efficient in detecting global anomalies from the hyperspectral data with low false alarm rate.

Hyperspectral image classification has drawn significant attention in the recent years driven by the increasing abundance of sensor-generated hyper- and multi-spectral data, combined with the rapid advancements in the field of machine learning. A vast range of techniques, especially involving deep learning models, have been proposed attaining high levels of classification accuracy. However, many of these approaches significantly deteriorate in performance in the presence of noise in the hyperspectral data. In this paper, we propose a new model that effectively addresses the challenge of noise residing in hyperspectral images. The proposed model, which will be called DCT-CNN, combines the representational power of Convolutional Neural Networks with the noise elimination capabilities introduced by frequency-domain filtering enabled through the Discrete Cosine Transform. In particular, the proposed method entails the transformation of pixel macroblocks to the frequency domain and the discarding of information that corresponds to the higher frequencies in every patch, in which pixel information of abrupt changes and noise often resides. Experiment results in Indian Pines, Salinas and Pavia University datasets indicate that the proposed DCT-CNN constitutes a promising new model for accurate hyperspectral image classification offering robustness to different types of noise, such as Gaussian and salt and pepper noise. © 2020 International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives., International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLIII, ISSN:1682-1750, ISSN:2194-9034, ISSN:1682-1777