Abstract-Extracting informative image features and learning effective approximate hashing functions are two crucial steps in image retrieval . Conventional methods often study these two steps separately, e.g., learning hash functions from a predefined hand-crafted feature space. Meanwhile, the bit lengths of output hashing codes are preset in most previous methods, neglecting the significance level of different bits and restricting their practical flexibility. To address these issues, we propose a supervised learning framework to generate compact and bit-scalable hashing codes directly from raw images. We pose hashing learning as a problem of regularized similarity learning. Specifically, we organize the training images into a batch of triplet samples, each sample containing two images with the same label and one with a different label. With these triplet samples, we maximize the margin between matched pairs and mismatched pairs in the Hamming space. In addition, a regularization term is introduced to enforce the adjacency consistency, i.e., images of similar appearances should have similar codes. The deep convolutional neural network is utilized to train the model in an end-to-end fashion, where discriminative image features and hash functions are simultaneously optimized. Furthermore, each bit of our hashing codes is unequally weighted so that we can manipulate the code lengths by truncating the insignificant bits. Our framework outperforms state-of-the-arts on public benchmarks of similar image search and also achieves promising results in the application of person re-identification in surveillance. It is also shown that the generated bit-scalable hashing codes well preserve the discriminative powers with shorter code lengths.
A shared aperture metasurface (SA-MS) is proposed to achieve the wideband radar cross-section (RCS) reduction and gain enhancement of a circularly polarised (CP) microstrip antenna simultaneously. The SA-MS is composed of a wideband polarisation conversion surface and a partially reflecting surface (PRS). A 180°phase difference can be obtained by the polarisation conversion surface, which results in a wideband RCS reduction in normal incident direction. The Fabry-Perot resonator cavity is constructed by the PRS and the metallic ground of the CP microstrip antenna to enhance the gain. Experimental results show that the gain enhancement reaches 1.15 dB around the operating band and a wideband RCS reduction is obtained in the frequency range of 9-17 GHz, in which the average RCS reduction is 10.9 and 10.68 dB for dual polarisation in the normal incident direction.
We demonstrate the broadband cascaded four-wave mixing (FWM) and supercontinuum (SC) generation in a tellurite MOF which is made from 76.5TeO(2)-6ZnO-11.5Li(2)O-6Bi(2)O(3) (TZLB, mol%) glass. By using a 2-μm picosecond laser with the center wavelength of ~1958 nm as the pump source, the broadband FWM with the frequency separation of ~1.1 THz is obtained. The bandwidth of the frequency comb spans a range of ~630 nm from ~1620 to 2250 nm at the average pump power of ~125 mW. With the average pump power increasing to ~800 mW, the broadband mid-infrared SC generation with the spectrum from ~900 to 3900 nm is observed. Changing the pump source to a femtosecond laser (optical parametric oscillator, OPO) with the center wavelength of ~2000 nm, solitons and dispersive waves (DWs) are obtained.
A multidimension source model for generating broadband ground motions with deterministic 3D numerical simulations is proposed in this article. In this model, the source is composed of several superimposed layers, and the total seismic moment is assigned to these layers in different proportions. Each layer exactly fills up the seismic fault and is uniformly divided into subsources with size decreased progressively to reflect different levels of rupture details. Hence, the proposed multidimension source model may consider the realistic rupture process of an earthquake, that is, the spatial and temporal heterogeneity of source parameters, and generate broadband ground motions. To verify this source model, the 1994 Northridge earthquake is simulated with four multidimension source models, based on different source inversion results. The amplitudes, durations, and spectral characteristics of the observed ground motions of the 1994 Northridge earthquake are respectably reproduced in a range of frequencies up to 5 Hz. In addition, a scenario earthquake is also simulated with four multidimension source models, with different synthetic rupture process. The simulated ground motions of the scenario earthquake are generally in good agreement with the Next Generation Attenuation-West 2 ground-motion prediction equations. This demonstrates that it is promising to simulate realistic broadband ground motions of strong earthquakes with a proper source description and realistic Earth models.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.