Abstract The rapid emergence of deep learning (DL) technology has resulted in its successful use various fields, including aquaculture. DL creates both new opportunities and a series challenges for information data processing smart fish farming. This paper focuses on applications aquaculture, live identification, species classification, behavioural analysis, feeding decisions, size or biomass estimation, water quality prediction. technical details methods applied to farming are also...

Fish species identification plays a vital role in marine fisheries resource exploration, yet datasets related to fish resources are scarce. In open-water environments, various often exhibit similar appearances and sizes. To solve these issues, we propose few-shot learning approach identifying species. Our involves two key components. Firstly, the embedding module was designed address challenges posed by large number of with phenotypes utilizing distribution relationships space. Secondly,...

Underwater images are widely used in ocean resource exploration and environment surveillance. However, due to the influence of light attenuation noise, underwater usually display degradation phenomena such as blurring color deviation; an enhancement method is required make more visible. Currently, there two major approaches for image enhancement: traditional methods based on physical or non-physical models, deep learning method. Inspired by fusion-based idea, this paper attempts combine with...

Deep Convolutional Neural Networks (DCNN) have the ability to learn complex features and are thus widely used in field of seismic signal denoising with low signal-to-noise ratio (SNR). However, current convolutional deep network for noise reduction does not make full use feature information extracted from all convolution layers network, cannot fit high SNR. To deal this problem, paper proposes DnRDB model, a time-frequency domain model combined residual dense blocks (RDB). The is mainly...

In recent years, residual learning has shown excellent performance on convolutional neural network (CNN)-based single-image super-resolution (SISR) tasks. However, CNN-based SISR approaches have focused mainly the design of deep architectures, and rectified linear units (ReLUs) used in these networks hinder shallow-to-deep information transfer. As a result, methods are unable to utilize some shallow information, improving model is difficult. To solve above issues, this paper proposes an...

A novel regularised image super‐resolution algorithm is proposed, building on the emerging cosparse or analysis sparse prior models, which are important complementary alternatives to widely used synthesis counterpart. Moreover, achieve adaptivity varying local structures of natural images, patch space partitioned into meaningful subspaces by clustering and learn sub‐dictionary for each cluster partitioned, performed online iteratively based solely current available information, maximum...

With the increasing demand of quality assurance and reliability additive manufacturing (AM), development advanced in-situ monitoring systems is increased to monitor process behavior. Optical-based camera are proved as effective ways observe part surface layer wise. For certain camera-based system, coverage build platform resolution images always a trade-off. In low-resolution images, detailed features (e.g. scan vector) often lost. Super (SR) algorithms discussed in literature, but there no...

Individual fish segmentation is a prerequisite for feature extraction and object identification in any machine vision system. In this paper, method of overlapping images aquaculture was proposed. First, the shape factor used to determine whether an overlap exists picture. Then, corner points were extracted using curvature scale space algorithm, skeleton obtained by improved Zhang-Suen thinning algorithm. Finally, intersecting obtained, overlapped region segmented. The results show that...

This paper presents a feature fusion and sparse transformer-based anomalous traffic detection system (FSTDS). FSTDS utilizes network to encode the data sequences extracting features, fusing them into coding vectors through shallow deep convolutional networks, followed by using transformer capture complex relationships between flows; finally, multilayer perceptron is used classify achieve anomaly detection. The of improves extraction from small sample data, encoder enhances understanding...

<div class="section abstract"><div class="htmlview paragraph">In the field of autonomous driving, in order to guarantee - robust perception performance at night and reduce cost data collection annotation, there are many day-to-night image translation methods based on Generative Adversarial Networks (GAN) generate realistic synthetic data. The vehicle light effect is great significance task (such as detection) scene. However, no research has ever focused problem translation....

In traditional fractal image coding, a candidate domain block D is optimally transformed to match the given range R. Besides spatial transformation S, baseline intensity mapping takes affine form of s-S(D) + o-I, where 5 and o are scaling factor offset coefficient respectively / refers constant vector with all its components equaling 1. Extension above surely possible has been studied in literature, this paper, we test simple extension which seems have not touched before, new coding scheme,...

For the epilepsy disease, many studies about thickness of cortex among various brain regions haven been implemented. However, there is few research cell distribution gray matter adjacent gyrus and sulcus. Our indicate it possible different that GFAP-IR interlaminar astrocytes somas between in gyri sulci epilepsy. Based on this motivation, we proposed a color slice image ROI (region interest) extracting method. A classifying method imported to compute comparability two pixels image. An...