Search

This paper presents different approaches to map bark beetle infested forests in Croatia. Bark beetle infestation presents threat to forest ecosystems. Due to large unapproachable area, it also presents difficulties in mapping infested areas. This paper analyses available machine learning options in open-source software QGIS and SAGA GIS. All options are performed on Copernicus data, Sentinel 2 satellite imagery. Machine learning and classification options are maximum likelihood classifier, minimum distance, artificial neural network, decision tree, K Nearest Neighbor, random forest, support vector machine, spectral angle mapper and Normal Bayes. Kappa values respectively are: 0.71; 0.72; 0.81; 0.68; 0.69; 0.75; 0.26; 0.60; 0.41 which shows highest classification accuracy for artificial neural networks method and lowest for support vector machine accuracy. [ABSTRACT FROM AUTHOR]

Economic mineralization is often associated with alterations that are identifiable by remote sensing coupled geological analysis. The present paper aims to investigate the capabilities of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), Landsat-8 and Sentinel-2 data to map iron oxide and hydrothermally alteration zones in the Abhar area, NW Iran. To achieve this goal, the principal component analysis (PCA) and two machine learning methods including support vector machine (SVM) and artificial neural network (ANN) were employed. PCA method was carried out on four bands of all data and then the appropriate principal components were selected to map alterations. Due to the high precision of ASTER data within the short-wave infrared range, these data results are more satisfactory compared with Landsat-8 and Sentinel-2 sensors in detecting hydrothermally alterations through the PCA technique. Based on the obtained maps, the performance of all data types was approximately similar in the detection of iron oxide zones. Our desired data were classified by two methods of SVM and ANN. The results of these algorithms were presented as confusion matrix. According to these results, for hydrothermally alterations, ASTER data showed better performance in both SVM and ANN than other datasets by gaining values greater than 90%. These data did not perform well in the iron oxide zones detection, while Landsat-8 and Sentinel-2 have been more successful. For iron oxide, based on confusion matrix, Landsat-8 data have obtained the values of 78% and 79% through SVM and ANN algorithms, respectively, and also Sentinel-2 has acquired the values of 88.11% and 90.55% via SVM and ANN, respectively. Therefore, to map iron oxide zones, Sentinel-2 data are more favorable than Landsat-8 data. In addition, the ANN algorithm in ASTER data has represented the highest overall accuracy and Kappa coefficient with the values of 88.73% and 0.8453, respectively. [ABSTRACT FROM AUTHOR]

As a variant of Convolutional Neural Networks (CNNs) in Deep Learning, the Fully Convolutional Network (FCN) model achieved state-of-the-art performance for natural image semantic segmentation. In this paper, an accurate classification approach for high resolution remote sensing imagery based on the improved FCN model is proposed. Firstly, we improve the density of output class maps by introducing Atrous convolution, and secondly, we design a multi-scale network architecture by adding a skip-layer structure to make it capable for multi-resolution image classification. Finally, we further refine the output class map using Conditional Random Fields (CRFs) post-processing. Our classification model is trained on 70 GF-2 true color images, and tested on the other 4 GF-2 images and 3 IKONOS true color images. We also employ object-oriented classification, patch-based CNN classification, and the FCN-8s approach on the same images for comparison. The experiments show that compared with the existing approaches, our approach has an obvious improvement in accuracy. The average precision, recall, and Kappa coefficient of our approach are 0.81, 0.78, and 0.83, respectively. The experiments also prove that our approach has strong applicability for multi-resolution image classification. [ABSTRACT FROM AUTHOR]

Previous biophysical and empirical models of evapotranspiration retrieval are difficult to parameterize because of the effects of the nonlinear biophysics of plants, terrestrial and solar radiation and soils, despite attempts made using various satellite products. In this study, the multilayer feed-forward neural network approach with Levenberg–Marquardt back propagation (LM-BP) was used to successfully estimate evapotranspiration using the input of various satellite-based products. When applying neural network training, value-added satellite-based products such as normalized difference vegetation index (NDVI), normalized difference water index (NDWI), land surface temperature (LST), air temperature and insolation are used instead of only spectral information from satellite sensors to reflect the spatial representativeness of the neural network. The evapotranspiration estimated from the neural network with input parameters showed better statistical accuracy than the MODIS products (MOD16) and Priestley–Taylor methods when compared with ground station eddy flux measurements, which were considered as reference data. Additionally, the temporal variation in neural network evapotranspiration well reflected seasonal patterns of eddy flux evapotranspiration, especially for the high cloudiness in the summer season. [ABSTRACT FROM AUTHOR]

This paper describes an application of artificial neural networks for the recognition of volcanic lava flow hot spots using remote sensing data. Satellite remote sensing is a very effective and safe way to monitor volcanic eruptions in order to safeguard the environment and the people affected by such natural hazards. Neural networks are an effective and well-established technique for the classification of satellite images. In addition, once well trained, they prove to be very fast in the application stage.In our study a back propagation neural network was used for the recognition of thermal anomalies affecting hot lava pixels. The network was trained using the three thermal channels of the Advanced Very High Resolution Radiometer (AVHRR) sensor as inputs and the corresponding values of heat flux, estimated using a two thermal component model, as reference outputs.As a case study the volcano Etna (Eastern Sicily, Italy) was chosen, and in particular the effusive eruption which took place during the month of 2006 July. The neural network was trained with a time-series of 15 images (12 nighttime images and 3 daytime images) and validated on three independent data sets of AVHRR images of the same eruption and on two relative to an eruption occurred the following month.While for both nighttime and daytime validation images the neural network identified the image pixels affected by hot lava with a 100 per cent success rate, for the daytime images also adjacent pixels were included, apparently not interested by lava flow. Despite these performance differences under different illumination conditions, the proposed method can be considered effective both in terms of classification accuracy and generalization capability. In particular our approach proved to be robust in the rejection of false positives, often corresponding to noisy or cloudy pixels, whose presence in multispectral images can often undermine the performance of traditional classification algorithms. Future work shall address application of the proposed method to data acquired with a high temporal resolution, such as those provided by the spinning enhanced visible and infrared imager sensor on board the Meteosat second generation geostationary satellite. [ABSTRACT FROM AUTHOR]

Artificial neural networks (ANNs) have become a popular tool in the analysis of remotely sensed data. Although significant progress has been made in image classification based upon neural networks, a number of issues remain to be resolved. This paper reviews remotely sensed data analysis with neural networks. First, we present an overview of the main concepts underlying ANNs, including the main architectures and learning algorithms. Then, the main tasks that involve ANNs in remote sensing are described. The limitations and crucial issues relating to the application of the neural network approach are discussed. A brief review of the implementation of ANNs in some of the most popular image processing software packages is presented. Finally, we discuss the application perspectives of neural networks in remote sensing image analysis. [ABSTRACT FROM AUTHOR]

Upland vegetation represents an important resource that requires frequent monitoring. However, the heterogeneous nature of upland vegetation and lack of ground data require classification techniques that have a high degree of generalization ability. This study investigates the use of artificial neural networks as a means of mapping upland vegetation from remotely sensed data. First, the optimum size of support to map upland vegetation was estimated as being less than 4 m, which suggested that soft classification techniques and high spatial resolution IKONOS imagery were required. The use of high spatial resolution imagery for regional‐scale areas has introduced new challenges to the remote sensing community, such as using limited ground data and mapping land‐cover dynamics and variation over large areas. This work then investigated the utility of artificial neural networks (ANN) for regional‐scale upland vegetation from IKONOS imagery using limited ground data and to map unseen data from remote geographical locations. A Multiple Layer Perceptron was trained with pixels from an IKONOS image using early stopping; however, despite high classification accuracies when calculated for pixels from an area where training pixels were extracted, the networks did not produce high accuracies when applied to unseen data from a remote area. [ABSTRACT FROM AUTHOR]

Transportation engineers are commonly faced with the question of how to extract information from expensive and scarce field data. Modeling the distribution of trips between zones is complex and dependent on the quality and availability of field data. This research explores the performance of neural networks in trip distribution modeling and compares the results with commonly used doubly constrained gravity models. The approach differs from other research in several respects; the study is based on both synthetic data, varying in complexity, as well as real-world data. Furthermore, neural networks and gravity models are calibrated using different percentages of hold out data. Extensive statistical analyses are conducted to obtain necessary sample sizes for significant results. The results show that neural networks outperform gravity models when data are scarce in both synthesized as well as real-world cases. Sample size for statistically significant results is forty times lower for neural networks. [ABSTRACT FROM AUTHOR]

SPOT VEGETATION is a recent sensor at 1 km resolution for land surface studies. Cloud detection based on this sensor is complicated by the absence of a thermal band. An artificial neural network was thus trained for the cloud detection on atmospherically corrected S1 daily data and on top of the atmosphere reflectance P data, from the SPOT VEGETATION system. It consists of a multi-layer perceptron with one hidden sigmoid layer, trained with the Levenberg-Marquardt back-propagation algorithm and generalized by the Bayesian regularization. Two neural networks allowed optimal cloud detections to be obtained. The first used all four bands of S1 data with 13 hidden nodes, and the second employed all four bands of P data with 11 hidden nodes. The multiple-layer perceptrons lead to a cloud detection accuracy of 98.0% and 97.6% for S1 and P data, respectively, when trained to map three predefined values that classify cloud, water and land. The network was further evaluated using three SPOT VEGETATION images taken at different rates. The network detected not only bright thick clouds but also thin or less bright clouds. The analysis demonstrated the superior classification of the network over the standard cloud masks provided with the data. [ABSTRACT FROM AUTHOR]

Eucalypt tree dieback is a disease that threatens the survival of woodlands in Australian national parks. For mapping and monitoring the spatial distribution of dieback, airborne imaging technologies can be more effective than ground surveys. Amongst the numerous types of airborne sensors, the video camera provides images with very high spatial resolution. In order to detect individual defoliated Eucalyptus trees at Mt Eccles national park (south-western Victoria), aerial video data was acquired across the study site. Highlighting the health status of sparse and mainly unclustered defoliated eucalypts at Mt Eccles through video images was deemed to be achievable in several steps. This paper introduces a classification method based on a feedforward neural network, whose main goal is to perform a segmentation of the video frames into three classes, namely, bare branches or trunks, healthy canopy and understorey vegetation. The aim of the algorithm is to create a subset of the eucalypt tree group, including defoliated and dead trees, for further analysis. The results suggest that the recognition of trunks and systems of bare branches is feasible using the neural network architecture. This provides a means to pre-process the video data so as to analyse the health of trees and thus assist park managers with managing dieback. [ABSTRACT FROM AUTHOR]

Multilayer feed-forward (MLF) neural networks were employed to estimate air temperatures in Southern Québec (Canada) using Advanced Very High Resolution Radiometer (AVHRR) images. The input variables for the networks were the five bands of the AVHRR image, surface altitude, solar zenith angle, and Julian day. The estimation was carried out using a dataset collected during the growing season from June to September 2000. Levenberg--Marquardt back-propagation (LM-BP) was used to train the networks. The early stopping method was applied to improve the LM-BP and to generalize the networks. Bands 4 and 5, which are used for retrieval of surface temperature, were the most critical components for the estimation. The contribution of Julian day to the precision of estimated air temperature was much superior to that of altitude and solar zenith angle for the dataset of inter-seasonal air temperatures. The network using all five bands, Julian day, altitude, and solar zenith angle provided the best results, with 22 nodes in the hidden layer. In the time series of estimated and station air temperatures, the difference between the temperatures was generally maintained within 2°C on various canopies, even during steep variations in August and September. [ABSTRACT FROM AUTHOR]

More than most European cities, Istanbul is experiencing considerable pressure from urban development due to a rapidly increasing population. As a consequence the land use activities in urban and suburban areas are changing dramatically. To provide cost-effective information about the current state and how it is changing in order to develop integrated policies, multi-temporal remotely sensed data, with its synoptic and regular coverage, is being used. Nevertheless, the mapping and monitoring of urban change through remote sensing is difficult owing to the complex urban land use patterns. Although many image processing techniques have been developed for this purpose, they are complicated by differences amongst images caused by differences in the effects of the atmosphere, illumination, and surface moisture. One technique which is relatively unaffected by these problems is based on artificial neural network (ANN) classification algorithms. The main objective of this study was to examine the performance of two ANN classifiers for land use classification using Landsat TM data. Two different supervised ANN approaches were used: the multi layer perceptron (MLP) and the learning vector quantization (LVQ). The performance of these classifiers was compared to the more conventional maximum likelihood approach. [ABSTRACT FROM AUTHOR]

Recent progress in remote sensing technologies, coupled with ongoing and planned remote sensing missions, is expected to generate hydrologic data at spatial, temporal, and spectral resolutions never previously available. Artificial neural networks (ANNs), although at early stages of hydrologic applications, are rapidly becoming an attractive tool to characterize, model, and predict complex multisource remotely sensed hydrologic data. We review and examine the utility of ANNs for hydrologic applications, with particular emphasis on remote sensing of precipitation, soil moisture, and multisource land surface data. In addition to more popularly used multilayer feedforward networks, we also review recurrent neural networks for prediction and self-organization neural networks for spatial characterization of heterogeneous land surface processes. [ABSTRACT FROM AUTHOR]

Age is a powerful variable that can be of significant value when modelling the health of forest-dominated ecosystem. Traditional investigations have attempted to extract age information from remotely sensed data by regressing the spectral values with in situ derived age data. Traditional statistical approaches assume (a) normally distributed remote sensing and in situ data, (b) no collinearity among variables, and (c) linear data relationships. Artificial neural networks (ANNs) are not bound by such assumptions and may yield improved predictive modelling of forest stand biophysical parameters if properly utilized. This study investigated traditional statistical and ANN approaches to perform the predictive modelling of the age of loblolly pine (Pinus taeda) for large stands in southern Brazil using Thematic Mapper (TM) data. An extensive comparison of pattern associator and back-propagation ANNs versus both linear and nonlinear regression analysis was conducted. Various neural network architectures were investigated to determine the optimal configuration for this particular dataset. Certain back-propagation ANNs modelled stand age significantly better than traditional statistical approaches because of their ability to take into account nonlinear, nonnormally distributed data. The results suggest that ANN analysis may be of significant value when using remote sensing data to model certain forest variables. [ABSTRACT FROM AUTHOR]