The current rapid development of China's power transmission and transformation industry, inflatable switchgear has excellent environmental adaptability and small size, in the transmission and distribution network ...
详细信息
A multi-target synchronous measurement method is proposed for the vibration measurement problem of large flexible structures. Select the appropriate system hardware according to the actual measurement requirements, an...
详细信息
The working principles and methods included in the existing motion estimation techniques are classified, and the research on motion estimation in various fields at home and abroad is collected in the existing literatu...
详细信息
It is an important task to estimate a 3D bounding box from monocular images for autonomous driving. However, the monocular pictures do not have distance information, so it is difficult to acquire accurate results. For...
详细信息
The audio signal in the running process of the switchgear contains a lot of operating state information, which is of great significance to the monitoring of the internal insulation fault and abnormal state. In order t...
详细信息
This paper addresses collaborative transportation of autonomous surface vehicles (ASVs) operating at a low speed in an open water area to assist in the movement of a ship. A constrained collaborative control method is...
详细信息
The purpose of fusing infrared and visible photos is to provide high-quality combined images with clear objects and detailed textures, even in very dark environments. The visual quality and contrast of both infrared a...
详细信息
ISBN:
(数字)9798350390254
ISBN:
(纸本)9798350390261
The purpose of fusing infrared and visible photos is to provide high-quality combined images with clear objects and detailed textures, even in very dark environments. The visual quality and contrast of both infrared and visible images tend to suffer in low-light situations. Previous algorithms fused poor- quality source images directly, resulting in fused images that still had low contrast and lost detail information. Thus, in this paper, we first design a multi-level brightness adjustment unit to learn visible light illumination components layer by layer and perform luminance enhancement while fusing multi-source image information. Second, to enhance the network's capacity for feature extraction, we developed the parallel attention module. This module was inserted after each brightness adjustment module. Extensive trials show that our technique works better than nine common fusion methods when compared visually and using objective measures.
We propose three alternative measures for non-Gaussianity of quantum states: sine distance, Bures angle, and Bures distance, which are based on quantum fidelity introduced by Wang [Phys. Lett. A 373 58(2008)]. Using t...
详细信息
We propose three alternative measures for non-Gaussianity of quantum states: sine distance, Bures angle, and Bures distance, which are based on quantum fidelity introduced by Wang [Phys. Lett. A 373 58(2008)]. Using them, we evaluate the non-Gaussianity of some relevant single-mode and two-mode non-Gaussian states and find a good consistency of the three examined measures. In addition, we show that such metrics can exactly quantify the degree of Gaussianity of even Schrödinger-cat-like states of small amplitudes that can not be measured by other known non-Gaussianity measures such as the Hilbert–Schmidt metric and the relative entropy metric. We make a comparative study between all existing nonGaussianity measures according to the metric axioms and point out that the sine distance is the best candidate among them.
Unlike traditional path-following methods, this paper develops a parallel control framework to investigate the path-following behavior of autonomous surface ships (ASVs) in the presence of unknown disturbances and mod...
详细信息
Transfer learning(TL)utilizes data or knowledge from one or more source domains to facilitate learning in a target *** is particularly useful when the target domain has very few or no labeled data,due to annotation ex...
详细信息
Transfer learning(TL)utilizes data or knowledge from one or more source domains to facilitate learning in a target *** is particularly useful when the target domain has very few or no labeled data,due to annotation expense,privacy concerns,***,the effectiveness of TL is not always *** transfer(NT),i.e.,leveraging source domain data/knowledge undesirably reduces learning performance in the target domain,and has been a long-standing and challenging problem in *** approaches have been proposed in the literature to address this ***,there does not exist a systematic *** paper fills this gap,by first introducing the definition of NT and its causes,and reviewing over fifty representative approaches for overcoming NT,which fall into three categories:domain similarity estimation,safe transfer,and NT *** areas,including computer vision,bioinformatics,natural language processing,recommender systems,and robotics,that use NT mitigation strategies to facilitate positive transfers,are also ***,we give guidelines on NT task construction and baseline algorithms,benchmark existing TL and NT mitigation approaches on three NT-specific datasets,and point out challenges and future research *** ensure reproducibility,our code is publicized at https://***/chamwen/NT-Benchmark.
暂无评论