Search

Wetlands are of great importance to the diversity of biota and ecology, thereby to humans. Monitoring such valuable areas is essential for sustainable development. When the sizes, geographic distribution, and total coverage of wetlands across the earth are taken into account, remote sensing shines out as the most economically and technically feasible method to realise the monitoring task. Concerning the utilisation of medium resolution satellite images as the input, the pixel-level approach falls short of understanding the wetland dynamics since vast amounts of pixels in such areas have mixed content. This study proposes a framework for determining the extent of wetlands and extracting their ground characteristics at the sub-pixel level. In the extent determination part, Tasselled Cap Water Index (TCWI) values are calculated on time series, and their variations throughout the year are modelled by fitting a double-sided sigmoid function. This information is coupled with Digital Terrain Model (DTM) thresholding to extract the final extent. A sub-pixel analysis is proposed for the latter part, which includes adopting a systematic approach using a three-element (soil, vegetation, water) scheme for establishing wetland ontology and implementing supervised spectral unmixing enhanced by band weight optimisation. Balı kdamı, one of the most impressive wetlands of Turkey, is chosen as the test area. Open-access optical satellite data acquired by the Sentinel-2 constellation are utilised as the primary data input. Since the abundance values of land cover classes in each Sentinel-2 pixel are estimated, reference abundance data with a 10 m ground sampling distance (GSD) are generated using four-band aerial images having a 30 cm GSD for the verification stage. A new method entitled ‘Abundance Confusion Matrix’ is introduced for comparison and detailed assessment of fractional land cover. Experimental results demonstrate that the extent determination is addressed with a precision of 99.21% and a miss rate of 5.75%. In addition, the abundance values of land cover classes are identified with an overall accuracy of 66.17% after the optimisation step. The proposed method proves to be a valuable tool for the detailed monitoring of wetlands. [ABSTRACT FROM AUTHOR]

This paper presents a novel approach for estimating wheat yields based on estimated abundances of endmembers attributed to photosynthetic pigments, using Landsat 8 images acquired during maximum greenness. Endmembers within the wheat pixels are found using an unmixing algorithm then they are further optimized to maximize the predictive power of the abundances for the yields. Similarity of the endmembers to photosynthetic pigment's spectral signatures and their predictive power suggests their relevance to the pigments. Although the initial unmixing of the intimate mixture of photosynthetic pigments is linear, interactions of abundances are used in the optimization to handle the non-linearity using bilinear model. Wheat yields are estimated with abundances, their relevant interactions, agrometeorological parameters and vegetation indices using three machine learning algorithms. Harvester records from 142 fields are used as ground truth for performance assessment. The yields are estimated with 82% accuracy (RMSE = 22.51) when Random Forest algorithm is used with important parameters. [ABSTRACT FROM AUTHOR]

Essential forest ecosystem services can be assessed by better understanding the diversity of vegetation, specifically those of Mediterranean region. A species level classification of maquis would be useful in understanding vegetation structure and dynamics, which would be an indicator of degradation or succession in the region. Although remote sensing was regularly used for classification in the region, maquis are simply represented as one to three categories based on density or height. To fill this gap, we test the capability of Sentinel-2 imagery, together with selected ancillary variables, for an accurate mapping of the dominant maquis formations. We applied Recursive Feature Selection procedure and used a Random Forest classifier. The algorithm is tested using ground truth collected from site and reached 78% and 93% overall accuracy at species level and physiognomic level, respectively. Our results suggest species level characterization of dominant maquis is possible with Sentinel-2 spatial resolution. [ABSTRACT FROM AUTHOR]

Remote sensing scene classification is gaining much more interest in the recent few years for many strategic fields such as security, land cover and land use monitoring. Several methods have been proposed in the literature and they can be divided into three main classes based on the features used: handcrafted features, features obtained by unsupervised learning and those obtained from deep learning. Handcrafted features are generally time consuming and suboptimal. Unsupervised learning based features which have been proposed later gave better results but their performances are still limited because they mainly rely on shallow networks and are not able to extract powerful features. Deep learning based features are recently investigated and gave interesting results. But, they cannot be usually used because of the scarcity of labelled remote sensing images and are also computationally expensive. Most importantly, whatever kind of feature is used, the neighbourhood information of them is ignored. In this paper, we propose a novel remote sensing scene representation and classification approach called Bag of Visual SubGraphs (BoVSG). First, each image is segmented into superpixels in order to summarize the image content while retaining relevant information. Then, the superpixels from all images are clustered according to their colour and texture features and a random label is assigned to each cluster that probably corresponds to some material or land cover type. Thus superpixels belonging to the same cluster have the same label. Afterwards, each image is modelled with a graph where nodes correspond to labelled superpixels and edges model spatial neighbourhoods. Finally, each image is represented by a histogram of the most frequent subgraphs corresponding to land cover adjacency patterns. This way, local spatial relations between the nodes are also taken into account. Resultant feature vectors are classified using standard classification algorithms. The proposed approach is tested on three popular datasets and its performance outperforms state-of-the-art methods, including deep learning methods. Besides its accuracy, the proposed approach is computationally much less expensive than deep learning methods. [ABSTRACT FROM AUTHOR]

Automated detection of buried anti-personnel landmines using remote sensing techniques is very important for clearing minefields without putting lives in danger. Although thermal infrared imaging is promising, it is far from applicable to the real world in its current state-of-the-art. The most serious problem is that experiments are generally held using sandboxes or levelled and cleared soil, but real fields are, at least partially, covered with plants. In this study, we present an algorithm for landmine detection that is robust enough to detect beyond the clutter caused by partial plant cover. The first part is a hypothesis generator based on circular Hough Transform applied to images that are filtered to enhance circular structures. The second part tests the candidate landmine coordinates using rotationally invariant features, including modified Histogram of Oriented Gaussians (HOG), over multiple images taken at different times after Wiener filtering to maximize signal-to-clutter ratio. The performances of various features and classifiers are compared. The overall performance of the algorithm is demonstrated on a dataset of real-world landmine images contaminated by simulated plants. Satisfactory results are obtained up to 40% equivalent plant coverage where more than 65% of the pixels are fully or partially covered by plants. [ABSTRACT FROM AUTHOR]

In order to apply hydrological models in the water resources investigation successfully, careful calibration and uncertainty analysis are required. Although many automatic calibration methods were developed, the time consumed for running the hydrologic model is still a problem for hydrologic modelers. To reduce the computational complexity and increase performance of the calibration procedure, a software package (Fast Automated Calibration Tool, FACT) that works on Soil and Water Assessment Tool (SWAT) was developed. Sequential Uncertainty Fitting Algorithm (SUFI‐2) was chosen to build the software package on, because SUFI‐2 combines optimization with uncertainty analysis and can handle a large number of parameters. SUFI‐2 implemented in SWAT‐CUP (a software program that was developed for SWAT) is very useful in an interactive manner, however; it has some drawbacks which are the time consumed, user interaction requirement and update problems of SWAT model files. In this study, the calibration procedure was implemented in a MATLAB script, which completes the full calibration in one single run. The developed tool was applied on Sarısu‐Eylikler Basin SWAT model that had r2 = 0.41, NSE = 0.11 between 1992 and 2010 without calibration. When the model was calibrated using SUFI‐2 in SWAT‐CUP, the model was improved to r2 = 0.57, NSE = 0.44, P‐factor = 0.69 and R‐factor = 1.00 at the end of 42 iterations using 100 simulation counts in each iteration. However; when the developed calibration tool was applied with 250 simulation counts without any user interaction for each iteration, the model was improved to r2 = 0.59, NSE = 0.57, P‐factor  = 0.72, and R‐factor = 1.32 at the end of three iterations. Thus, the developed calibration procedure took a shorter time (unlike SUFI2 in SWAT‐CUP, FACT takes several hours) compared to SWAT‐CUP with minimum user involvement. [ABSTRACT FROM AUTHOR]

Superpixel algorithms are aimed to partition an image into multiple similar sized segments based on similarity and proximity of pixels. In the heterogeneous regions, the boundaries of the objects should adhere well to the superpixels, and in the homogeneous parts, the pixels should be clustered so that compact superpixels are generated. Since speckle noise inherently exists in synthetic aperture radar (SAR) images their segmentation is considerably more difficult. In this article, the first contribution is the use of Mahalanobis distance instead of Euclidian, so that the superpixels can have elongated shapes to fit the complex structure of the real world better. Secondly, this geometric distance term is combined with similarity ratio term, which leads to even better performance on SAR images. Finally, the global constant that determines the relative importance of geometric proximity and pixel intensity similarity terms, whose best value should be chosen for each image, is considered. Instead of a global constant, its value is determined individually for each superpixel pair as a function of the average values of the superpixels. Experimental results with synthetic and real images demonstrate that the proposed approach has better segmentation performance than many of the existing state-of-the-art algorithms. [ABSTRACT FROM PUBLISHER]

Millions of buried and surface landmines throughout the world will continue to kill and injure people for many decades. Reliable landmine detection systems, especially those that do not require that an operator enters a minefield, are demanded. In such systems, data are captured by sensors and processed with signal-processing techniques. In recent studies, ground-penetrating radars, ultrasound transducers, metal detectors, and infrared sensors have been used to capture data as individual images or image time series. Bare soil and mines have different thermal characteristics, and this difference can be observed on the soil surface with thermal sensors. Since this is a dynamic behavior driven by radiation from the sun and changes with temperature during the day, it can be observed better in thermal image time series. In the proposed method, a two-step approach is adopted: pixelwise classification followed by mine detection. In the first step, a specific spatial filter is applied on images and the filtered image series is classified pixelwise using supervised classification. In the second step, the classified pixels are combined after a smoothing operation to detect mines. The types of the detected mines can also be determined. The method is tested on real data to quantify its classification and detection performance. The results are quite promising. Among the classifiers tested in this paper, the quadratic discriminate function classifier produces the best results. [ABSTRACT FROM PUBLISHER]