No abstract

No abstract

No abstract

No abstract

Aircraft detection from very high resolution (VHR) remote sensing images has been drawing increasing interest in recent years due to the successful civil and military applications. However, several challenges still exist: 1) extracting the high-level features and the hierarchical feature representations of the objects is difficult; 2) manual annotation of the objects in large image sets is generally expensive and sometimes unreliable; and 3) locating objects within such a large image is difficult and time consuming. In this paper, we propose a weakly supervised learning framework based on coupled convolutional neural networks (CNNs) for aircraft detection, which can simultaneously solve these problems. We first develop a CNN-based method to extract the high-level features and the hierarchical feature representations of the objects. We then employ an iterative weakly supervised learning framework to automatically mine and augment the training data set from the original image. We propose a coupled CNN method, which combines a candidate region proposal network and a localization network to extract the proposals and simultaneously locate the aircraft, which is more efficient and accurate, even in largescale VHR images. In the experiments, the proposed method was applied to three challenging high-resolution data sets: the Sydney International Airport data set, the Tokyo Haneda Airport data set, and the Berlin Tegel Airport data set. The extensive experimental results confirm that the proposed method can achieve a higher detection accuracy than the other methods.Index Terms-Aircraft detection, convolutional neural networks (CNNs), weakly supervised learning.

The Insight-Hard X-ray Modulation Telescope (Insight-HXMT) is a broad band X-ray and gamma-ray (1-3000 keV) astronomy satellite. The High Energy X-ray telescope (HE) is one of its three main telescopes. The main detector plane of HE is composed of 18 NaI(Tl)/CsI(Na) phoswich detectors, where NaI(Tl) serves as primary detector to measure ~ 20-250 keV photons incident from the field of view (FOV) defined by the collimators, and CsI(Na) is used as an active shield detector to NaI(Tl) by pulse shape discrimination. CsI(Na) is also used as an omnidirectional gamma-ray monitor. The HE collimators have a diverse FOV: 1.1°x 5.7° (15 units), 5.7°x 5.7° (2 units) and blocked (1 unit), thus the combined FOV of HE is about 5.7°x 5.7°. Each HE detector has a diameter of 190 mm, resulting in the total geometrical area of about 5100 cm 2 . The energy resolution is ~15% at 60 keV. The timing accuracy is better than 10 μs and dead-time for each detector is less than 10 μs. HE is devoted to observe the spectra and temporal variability of X-ray sources in the 20-250 keV band either by pointing observations for known sources or scanning observations to unveil new sources, and to monitor the gamma-ray sky in 0.2-3 MeV. This paper presents the design and performance of the HE instruments. Results of the on-ground calibration experiments are also reported.

For autonomous driving, it is important to detect obstacles in all scales accurately for safety consideration. In this paper, we propose a new spatial attention fusion (SAF) method for obstacle detection using mmWave radar and vision sensor, where the sparsity of radar points are considered in the proposed SAF. The proposed fusion method can be embedded in the feature-extraction stage, which leverages the features of mmWave radar and vision sensor effectively. Based on the SAF, an attention weight matrix is generated to fuse the vision features, which is different from the concatenation fusion and element-wise add fusion. Moreover, the proposed SAF can be trained by an end-to-end manner incorporated with the recent deep learning object detection framework. In addition, we build a generation model, which converts radar points to radar images for neural network training. Numerical results suggest that the newly developed fusion method achieves superior performance in public benchmarking. In addition, the source code will be released in the GitHub.

Research has revealed systematic changes in warming rates with elevation (EDW) in mountain regions. However, weather stations on the Tibetan plateau are mostly located at lower elevations (3,000‐4,000 m) and are nonexistent above 5,000 m, leaving critical temperature changes unknown. Satellite LST (Land Surface Temperature) can fill this gap but needs calibrating against in situ air temperatures (Tair). We develop a novel statistical model to convert LST to Tair, developed at 87 high‐elevation Chinese Meteorological Administration stations. Tair (daily maximum/minimum temperatures) is compared with Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua LST (1330 and 0130 local time) for 8‐day composites during 2002‐2017. Typically, 80‐95% of the difference between LST and Tair (ΔT) is explained using predictors including LST diurnal range, morning heating/nighttime cooling rates, the number of cloud free days/nights, and season (solar angle). LST is corrected to more closely represent Tair by subtracting modeled ΔT. We validate the model using an AWS on Zhadang Glacier (5800 m). Trend analysis at the 87 stations (2002‐2017) shows corrected LST trends to be similar to original Tair trends. To examine regional contrasts in EDW patterns, elevation profiles of corrected LST trends are derived for three ranges (Qilian Mountains, NyenchenTanglha, and Himalaya). There is limited EDW in the Qilian mountains. Maximum warming is observed around 4,500‐5,500 m in NyenchenTanglha, consistent with snowline retreat. In common with other studies, there is stabilization of warming at very high elevations in the Himalaya, including absolute cooling above 6,000 m, but data there are compromised by frequent cloud.