Abstract The safety and reliability of mechanical equipment are ensured by rolling bearings, which play a crucial role as vital elements rotating machinery. However, the complex fault features bearings cannot be fully characterized single-channel data, feature distribution is significantly varied under varying operating conditions, leading to substantial decline in model diagnostic performance. To address these issues, multi-channel data-driven domain adaptation (DA) method based on central...

A new algorithm for lossless compression of hyperspectral imagery is proposed. First, the average value four neighbour pixels current pixel calculated as local mean, which subtracted by to eliminate correlation in band image. The residual produced this step called difference. differences co‐locate with previous bands form input vector recursive least square (RLS) filter, prediction difference produced. Then, sent adaptive arithmetic encoder. Experiment results show that proposed produces...

Abstract Currently, the diagnostic performance of many deep learning algorithms may drop dramatically when distribution training data is significantly different from that test data. Moreover, fault diagnosis approaches based on single-channel suffer problems such as large precision fluctuation, low reliability, and incomplete expression features. To overcome above deficiencies, a novel multi-channel data-driven recognition method fusion sparse filtering (SF) discriminative domain adaptation...

In this paper, a new method for generating different texture images is presented. This involves simple transform from certain one-dimensional (1-D) signal to an expected two-dimensional (2-D) image. Unlike traditional methods, the input generated by 1-D function in our work instead of sample texture. We first into frequency domain using fast Fourier transform. Based on sufficient analysis 2-D discrete cosine (DCT) domain, where each coefficients expresses feature direction, pseudo-DCT are...

Traditional image dehazing algorithms based on prior knowledge and deep learning rely the atmospheric scattering model are easy to cause color distortion incomplete dehazing. To solve these problems, an end-to-end algorithm residual attention mechanism is proposed in this paper. The network includes four modules: encoder, multi-scale feature extraction, fusion decoder. encoder module encodes input haze into map, which convenient for subsequent extraction reduces memory consumption; smoothed...

Abstract With the high yield of many wells represented by Well JT1 in Maokou Formation, has catalyzed a surge exploration activities along platform margin facies Formation central Sichuan and further showed significant potential northern slope. However, fracture cave body exhibits degree development, strong longitudinal horizontal heterogeneity, large formation pressure differences, drilling events such as gas kicks lost circulation occur frequently, which seriously affects efficient...

Abstract Bearings, as the core component for power transmission, are crucial in ensuring safe and reliable operation of equipment. However, fault information contained a single-channel vibration signal is inherently limited. Additionally, under time-varying speed conditions, features prone to drift, cross-domain diagnostic performance most traditional domain adaptation (DA) models may drop dramatically. To solve above problems enhance ability DA extracting invariant features, this paper...

The objective of this paper is to investigate the effectiveness transform domain adaptive algorithms in suppressing acoustic echo fullband and subband cancellers (AEC). AEC considered consists main auxiliary subbands designed provide a numerically efficient structure. Comparison results between using LMS algorithm are provided.

This work presents a novel rate-control model. Based on the famous heat conduction equation (HCE), rate transmission (RTE) is proposed to describe bit-rate behaviors during video coding. RTE inherently diffusion equation. A new model derived from special solution RTE. this model, scheme for MPEG-4 applications. Experiments show that achieves better buffer status and higher picture quality over Q2 scheme.

Recently, marvelous success has been obtained for machine learning approaches in solving the mechanical defect detection problems. However, traditional algorithms usually fail to generalize new input distributions, which may cause classification accuracy drops dramatically. To overcome this deficiency, domain adaptation techniques can be employed transferring and adapting source target domain. In paper, an effective unsupervised approach called subspace distribution alignment (SDA) is...

To extract buildings accurately, we propose a foreground-aware refinement network for building extraction. In particular, in order to reduce the false positive of buildings, design module using attention gate block, which effectively suppresses features nonbuilding and enhances sensitivity model buildings. addition, introduce reverse mechanism detail module. Specifically, this guides learn supplement missing details by erasing currently predicted regions achieves more accurate complete...