Search

Deep learning classifiers exhibit remarkable performance for hyperspectral image classification given sufficient labeled samples but show deficiency in the situation of learning with limited labeled samples. Active learning endows deep learning classifiers with the ability to alleviate this deficiency. However, existing active deep learning methods tend to underestimate the feature variability of hyperspectral images when querying informative unlabeled samples subject to certain acquisition heuristics. A major reason for this bias is that the acquisition heuristics are normally derived based on the output of a deep learning classifier, in which representational power is bounded by the number of labeled training samples at hand. To address this limitation, we developed a feature-oriented adversarial active learning (FAAL) strategy, which exploits the high-level features from one intermediate layer of a deep learning classifier for establishing an acquisition heuristic based on a generative adversarial network (GAN). Specifically, we developed a feature generator for generating fake high-level features and a feature discriminator for discriminating between the real high-level features and the fake ones. Trained with both the real and the fake high-level features, the feature discriminator comprehensively captures the feature variability of hyperspectral images and yields a powerful and generalized discriminative capability. We leverage the well-trained feature discriminator as the acquisition heuristic to measure the informativeness of unlabeled samples. Experimental results validate the effectiveness of both (i) the full FAAL framework and (ii) the adversarially learned acquisition heuristic, for the task of classifying hyperspectral images with limited labeled samples. [ABSTRACT FROM AUTHOR]

Currently, effective crowd management based on the information provided by crowd monitoring systems is difficult as this information comes in at the moment adverse crowd movements are already occurring. Up to this moment, very little forecasting techniques have been developed that predict crowd flows a longer time period ahead. Moreover, most contemporary state estimation methods apply demanding pre-processing steps, such as map-matching. The objective of this paper is to design, train and benchmark a data-driven procedure to forecast crowd movements, which can in real-time predict crowd movement. This procedure entails two steps. The first step comprises of a cell sequence derivation method that allows the representation of spatially continuous GPS traces in terms of discrete cell sequences. The second step entails the training of a Recursive Neural Network (RNN) with a Gated Recurrent Unit (GRU) and six benchmark models to forecast the next location of pedestrians. The RNN-GRU is found to outperform the other tested models. Some additional tests of the ability of the RNN-GRU to forecast illustrate that the RNN-GRU preserves its predictive power when a limited amount of data is used from the first few hours of a multi-day event and temporal information is incorporated in the cell sequences. [ABSTRACT FROM AUTHOR]

Rice is one of the world's major staple foods, especially in China. Highly accurate monitoring on rice-producing land is, therefore, crucial for assessing food supplies and productivity. Recently, the deep-learning convolutional neural network (CNN) has achieved considerable success in remote-sensing data analysis. A CNN-based paddy-rice mapping method using the multitemporal Landsat 8, phenology data, and land-surface temperature (LST) was developed during this study. First, the spatial–temporal adaptive reflectance fusion model (STARFM) was used to blend the moderate-resolution imaging spectroradiometer (MODIS) and Landsat data for obtaining multitemporal Landsat-like data. Subsequently, the threshold method is applied to derive the phenological variables from the Landsat-like (Normalized difference vegetation index) NDVI time series. Then, a generalized single-channel algorithm was employed to derive LST from the Landsat 8. Finally, multitemporal Landsat 8 spectral images, combined with phenology and LST data, were employed to extract paddy-rice information using a patch-based deep-learning CNN algorithm. The results show that the proposed method achieved an overall accuracy of 97.06% and a Kappa coefficient of 0.91, which are 6.43% and 0.07 higher than that of the support vector machine method, and 7.68% and 0.09 higher than that of the random forest method, respectively. Moreover, the Landsat-derived rice area is strongly correlated (R2 = 0.9945) with government statistical data, demonstrating that the proposed method has potential in large-scale paddy-rice mapping using moderate spatial resolution images. [ABSTRACT FROM AUTHOR]

Unplanned urban settlements exist worldwide. The geospatial information of these areas is critical for urban management and reconstruction planning but usually unavailable. Automatically characterizing individual buildings in the unplanned urban village using remote sensing imagery is very challenging due to complex landscapes and high-density settlements. The newly emerging deep learning method provides the potential to characterize individual buildings in a complex urban village. This study proposed an urban village mapping paradigm based on U-net deep learning architecture. The study area is located in Guangzhou City, China. The Worldview satellite image with eight pan-sharpened bands at a 0.5-m spatial resolution and building boundary vector file were used as research purposes. There are ten sites of the urban villages included in this scene of the Worldview image. The deep neural network model was trained and tested based on the selected six and four sites of the urban village, respectively. Models for building segmentation and classification were both trained and tested. The results indicated that the U-net model reached overall accuracy over 86% for building segmentation and over 83% for the classification. The F1-score ranged from 0.9 to 0.98 for the segmentation, and from 0.63 to 0.88 for the classification. The Interaction over Union reached over 90% for the segmentation and 86% for the classification. The superiority of the deep learning method has been demonstrated through comparison with Random Forest and object-based image analysis. This study fully showed the feasibility, efficiency, and potential of the deep learning in delineating individual buildings in the high-density urban village. More importantly, this study implied that through deep learning methods, mapping unplanned urban settlements could further characterize individual buildings with considerable accuracy. [ABSTRACT FROM AUTHOR]