No abstract

Accurate prediction of future observations based on past data is the key to near real-time disturbance detection using satellite image time series (SITS). To overcome the limitations of existing methods, we present an attention-based long-short-term memory (LSTM) encoder-decoder model in which the historical time series of a pixel is encoded with a bidirectional LSTM encoder while the future time series is produced by another LSTM decoder. An attention mechanism is integrated into the encoder-decoder model to align the input time series with the output time series and to dynamically choose the most relevant contextual information while forecasting. Based on the proposed model, we develop a framework for near real-time disturbance detection and verify its effectiveness in the case of burned area mapping. The prediction accuracy of the proposed model is evaluated using moderate resolution imaging spectroradiometer (MODIS) time series and compared with state-of-the-art models. Experimental results show that our model achieves the best results in terms of lower prediction error and higher model fitness. We also evaluate the disturbance detection ability of the proposed framework. The proposed approach improves the detection rate of disturbances while suppressing false alarms, and increases the temporal accuracy. We suggest that the proposed methods provide new tools for enhancing current early warning systems in real time. Index Terms-Attention mechanism, encoder-decoder, longshort-term memory (LSTM), near real-time disturbance detection, satellite image time series (SITS). I. INTRODUCTIONT HE dramatic advances in data storage technology and high-performance computing over the past decade provide the opportunity to conduct time-series analyses with unprecedented volumes of data. Recently, near real-time disturbance Manuscript

This paper is dedicated to Professor Kunyang on the occasion of his 70th birthday.For MODIS NDVI time series with cloud noise and time distortion, we propose an effective time series clustering framework including similarity measure, prototype calculation, clustering algorithm and cloud noise handling. The core of this framework is dynamic time warping (DTW) distance and its corresponding averaging method, DTW barycenter averaging (DBA). We used 12 years of MODIS NDVI time series to perform annual land-cover clustering in Poyang Lake Wetlands. The experimental result shows that our method performs better than classic clustering based on ordinary Euclidean methods.

No abstract

No abstract

ABSTRACT:Anomaly regions in satellite images can reflect unexpected changes of land cover caused by flood, fire, landslide, etc. Detecting anomaly regions in satellite image time series is important for studying the dynamic processes of land cover changes as well as for disaster monitoring. Although several methods have been developed to detect land cover changes using satellite image time series, they are generally designed for detecting inter-annual or abrupt land cover changes, but are not focusing on detecting spatial-temporal changes in continuous images. In order to identify spatial-temporal dynamic processes of unexpected changes of land cover, this study proposes a method for detecting anomaly regions in each image of satellite image time series based on seasonal autocorrelation analysis. The method was validated with a case study to detect spatial-temporal processes of a severe flooding using Terra/MODIS image time series. Experiments demonstrated the advantages of the method that (1) it can effectively detect anomaly regions in each of satellite image time series, showing spatial-temporal varying process of anomaly regions, (2) it is flexible to meet some requirement (e.g., zvalue or significance level) of detection accuracies with overall accuracy being up to 89% and precision above than 90%, and (3) it does not need time series smoothing and can detect anomaly regions in noisy satellite images with a high reliability.

ObjectivesTranscutaneous auricular vagus nerve stimulation (taVNS) has been reported to be effective for chronic insomnia (CI). However, the appropriate population for taVNS to treat insomnia is unclear.MethodsTotal twenty-four patients with CI and eighteen health controls (HC) were recruited. Rest-state functional magnetic resonance imaging (Rs-fMRI) was performed before and after 30 min' taVNS at baseline. The activated and deactivated brain regions were revealed by different voxel-based analyses, then the seed-voxel functional connectivity analysis was calculated. In the CI group, 30 min of taVNS were applied twice daily for 4 weeks. Pittsburgh Sleep Quality Index (PSQI) and Flinders Fatigue Scale (FFS) were also assessed before and after 4 weeks of treatment in the CI group. The HC group did not receive any treatment. The correlations were estimated between the clinical scales' score and the brain changes.ResultsThe scores of PSQI (p < 0.01) and FFS (p < 0.05) decreased after 4 weeks in the CI group. Compared to the HC group, the first taVNS session up-regulated left dorsolateral prefrontal cortex (dlPFC) and decreased the functional connectivity (FCs) between dlPFC and bilateral medial prefrontal cortex in the CI group. The CI groups' baseline voxel wised fMRI value in the dlPFC were negatively correlated to the PSQI and the FFS score after 4 weeks treatment.ConclusionsIt manifests that taVNS has a modulatory effect on the prefrontal cortex in patients with CI. The initial state of dlPFC may predict the efficacy for taVNS on CI.

No abstract