Complex background in infrared images often challenges the accurate detection of small targets. To address this problem, we propose a weighted adaptive ring top-hat transformation (WARTH) for extracting infrared small targets in complex backgrounds. The method utilizes an adaptive ring-shaped structural element (SE) and a target awareness indicator to effectively measure local and global feature information to detect small targets while minimizing false alarms accurately. Firstly, the difference-of-structure tensors (DoST) is designed, and the positive smallest eigenvalue of DoST (PSEDoST) is computed to construct the adaptive ring-shaped SE that captures local feature information for background estimation. Secondly, the image is converted into the Fourier phase spectrum, and a two-stage sliding window filtering technique is designed to generate the target awareness indicator that perceives global feature information of small targets. Finally, the WARTH is defined by fusing the above two measurements, which can further eliminate false alarms and improve the robustness of target detection. The experimental results demonstrate that the WARTH is superior to several advanced methods in terms of false alarm reduction and small target detection in complex backgrounds.

Deep reinforcement learning (DRL) provides a new solution for autonomous robotic path planning in a known indoor environment. Previous studies mainly focused on robot path optimization but ignored blind areas in the indoor exploration, naturally result in low coverage rate and low exploration efficiency. The blind areas exploration is a crucial issue of the indoor environment. This work proposes an indoor blind area -oriented autonomous robotic path planning approach using DRL methods. First, the method optimization is based on a double deep Q -network (DDQN) with prioritized experience replay (PER). Then the Blocking and Blind Angle mechanism (BBA) is proposed to explore blind areas, assisted in selecting the optimal exploration points of next moment. Meanwhile it solves the common sparse reward problem in DRL. Finally, the presented method is successfully applied in simulation environment using a cleaning robot. Experiments show that the proposed BBAPER-DDQN not only explores the blind areas, but also accelerates the convergence speed. The results show that the training time is reduced from more than one hour to 36 minutes, and the coverage rate is increased by 11.37% higher than that of the baseline algorithms.

Factory recirculating aquaculture system (RAS) is facing in a stage of continuous research and technological innovation. Intelligent aquaculture is an important direction for the future development of aquaculture. However, the RAS nowdays still has poor self-learning and optimal decision-making capabilities, which leads to high aquaculture cost and low running efficiency. In this paper, a precise aeration strategy based on deep learning is designed for improving the healthy growth of breeding objects. Firstly, the situation perception driven by computer vision is used to detect the hypoxia behavior. Then combined with the biological energy model, it is constructed to calculate the breeding objects oxygen consumption. Finally, the optimal adaptive aeration strategy is generated according to hypoxia behavior judgement and biological energy model. Experimental results show that the energy consumption of proposed precise aeration strategy decreased by 26.3% compared with the manual control and 12.8% compared with the threshold control. Meanwhile, stable water quality conditions accelerated breeding objects growth, and the breeding cycle with the average weight of 400 g was shortened from 5 to 6 months to 3-4 months. (c) 2024 The Authors. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Virtual Reality (VR) is a key industry for the development of the digital economy in the future. Mobile VR has advantages in terms of mobility, lightweight and cost-effectiveness, which has gradually become the mainstream implementation of VR. In this paper, a mobile VR video adaptive transmission mechanism based on intelligent caching and hierarchical buffering strategy in Mobile Edge Computing (MEC)-equipped 5G networks is proposed, aiming at the low latency requirements of mobile VR services and flexible buffer management for VR video adaptive transmission. To support VR content proactive caching and intelligent buffer management, users' behavioral similarity and head movement trajectory are jointly used for viewpoint prediction. The tile-based content is proactively cached in the MEC nodes based on the popularity of the VR content. Second, a hierarchical buffer-based adaptive update algorithm is presented, which jointly considers bandwidth, buffer, and predicted viewpoint status to update the tile chunk in client buffer. Then, according to the decomposition of the problem, the buffer update problem is modeled as an optimization problem, and the corresponding solution algorithms are presented. Finally, the simulation results show that the adaptive caching algorithm based on 5G intelligent edge and hierarchical buffer strategy can improve the user experience in the case of bandwidth fluctuations, and the proposed viewpoint prediction method can significantly improve the accuracy of viewpoint prediction by 15%.

The finance-level Artificial Intelligence of Things (AIoT) is going to become a novel media in the 6G-driven digital society. Inside the financial AIoT environment, large-scale crowd credit assessment with the guarantee of low latency has been a general demand. Facing limited computational resources, there is still a lack of effective computation offloading methods for this purpose to ensure low latency. In order to deal with such an issue, this article introduces edge computing mode and proposes a low-latency edge computation offloading scheme for trust evaluation in financial AIoT. With different elements involved in the assessment process being denoted via mathematical description, a multiobjective optimization problem with constraints is formulated. Then, the aforementioned optimization problem is solved by a specific search algorithm, so that optimal task offloading schemes can be found. To assess the performance of the proposal, some simulation experiments are conducted to verify the proposed task offloading method. And it can be reflected from numerical results that latency can be well reduced compared with baseline methods.

This paper presents a robust scheme to extract weak small target under intricate backgrounds. Firstly, the maximally stable extremal regions (MSER) algorithm is employed to seek extremal regions whose size and shape are consistent with the definition of small target and whose gray level is relatively stable. Then, in view of the fact that small targets are relatively sparse defect areas in the whole image, the MSER-induced global saliency measure (MGSM) is developed to reduce regular backgrounds and enhance target signal. Meanwhile, based on the characteristics of small targets with compact gray levels and a certain contrast with its surrounding background, the MSER-induced local saliency measure (MLSM) is designed to reliably enlarge the target signal and remove strong clutter interferences. Finally, the reinforced MSER-induced saliency measure (RMSM) defined by fusing MGSM and MLSM can successfully eliminate complex backgrounds and highlight real targets. Results demonstrate that this method has superiority in enhancing dim target against various backgrounds and has strong robustness to different target shapes and sizes.

In the Recirculating Aquaculture Systems (RAS), the control of water quality indices remains essential to survival and growth of aquaculture objects. This requires effect prediction of future water status in advance, which can be adopted to help the generation of following control strategies. However, conventional methods of water quality prediction were mostly dependent on redundant parameters of model, which leads to inefficiency and low accuracy. In addition, the complexity of the RAS multi-units requires intelligent control of the water quality unit. Thus, a prediction and control framework for predicting water quality in RAS is proposed in this paper. Specifically, a hybrid deep learning structure which combines the Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU) and Attention mechanism is presented. To begin with, the CNN is utilized to extract local features for different timestamped water quality parameter. After the local features have been extracted, the proposed GRU model replicates the global sequential features of the parameters. The attention mechanism is then applied to focus on more critical features to promote the efficiency and accuracy of prediction. Finally, to demonstrate the efficiency and stability of the prediction and control framework with the mixture of CNN, GRU and Attention (PC-CGA), multiple groups of experiments and evaluations are carried out in a medium size RAS.

Real-time digital twin technology can enhance traffic safety of intelligent vehicular system and provide scientific strategies for intelligent traffic management. At the same time, real-time digital twin depends on strong computation from vehicle side to cloud side. Aiming at the problem of delay caused by the dual dependency of timing and data between computation tasks, and the problem of unbalanced load of mobile edge computing servers, a parallel intelligence-driven resource scheduling scheme for computation tasks with dual dependencies of timing and data in the intelligent vehicular systems (IVS) is proposed. First, the delay and energy consumption models of each computing platform are formulated by considering the dual dependence of sub-tasks. Then, based on the bidding idea of the auction algorithm, the allocation model of computing resources and communication resources is defined, and the load balance model of the mobile edge computing (MEC) server cluster is formulated according to the load status of each MEC server. Secondly, joint optimization problem for offloading, resource allocation, and load balance is formulated. Finally, an adaptive particle swarm with genetic algorithm is proposed to solve the optimization problem. The simulation results show that the proposed scheme can reduce the total cost of the system while satisfying the maximum tolerable delay, and effectively improve the load balance of the edge server cluster.

Device to device (D2D) is becoming one of the critic technologies of green smart cities. The uplink power allocation is investigated, when D2D users reuse subchannels of cellular users. The overall goal lies in minimization of energy consumption under guarantee of required data transmission rate, which belongs to a nondeterministic polynomial (NP) hard optimization problem. After analysis, we find that the constraint has implicit monotonicity, so we introduce a new vector to transform it into a monotonic optimization problem. Then, the equivalence of the problem before and after transformation is proved. To search global optimal energy conservation solution, we design a novel power allocation algorithm based on principle of reverse polyblock approximation and prove its convergence. It is demonstrated that our proposal is able to reduce power consumption in comparison to the most advanced alternatives. The proposed algorithm can be used as a benchmark to evaluate other algorithms.

With the development of intelligent aquaculture, the aquaculture industry is gradually switching from traditional crude farming to an intelligent industrial model. Current aquaculture man-agement mainly relies on manual observation, which cannot comprehensively perceive fish living con-ditions and water quality monitoring. Based on the current situation, this paper proposes a data-driven intelligent management scheme for digital industrial aquaculture based on multi-object deep neural network (Mo-DIA). Mo-IDA mainly includes two aspects of fish state management and environmental state management. In fish state management, the double hidden layer BP neural network is used to build a multi-objective prediction model, which can effectively predict the fish weight, oxygen con-sumption and feeding amount. In environmental state management, a multi-objective prediction model based on LSTM neural network was constructed using the temporal correlation of water quality data series collection to predict eight water quality attributes. Finally, extensive experiments were con-ducted on real datasets and the evaluation results well demonstrated the effectiveness and accuracy of the Mo-IDA proposed in this paper.

In a recirculating aquaculture system (RAS), feeding is an important factor affecting the growth of breeding objects. The traditional feeding methods relied on manual experience, which resulted in high labor costs and bait waste. To deal with these challenges, this paper proposes a dynamic scene images-assisted intelligent control method for industrial feeding through deep vision learning. First, a feeding video is processed according to the interframe difference method to obtain the image of the feeding state of the fish. Then, a modified VGG16 model is developed to determine the feeding state of the fish, transform it into a binary classification problem, and calculate the feeding frequency of the fish. After that, residual bait detection is deployed by adapting the YOLOv5 model. The results of the feeding frequency and the residual bait detection are then used to develop an intelligent feeding strategy to improve the growth rate of the fish and the conversion rate of the bait. Experimental tests on real-world scene images showed that the accuracy of identifying the feeding state by the modified VGG16 model reaches 92.4%. Through the verification of the medium-size RAS, compared with the traditional feeding method, the intelligent feeding method significantly saves manpower and reduce 15% of bait waste.

The whole process of tobacco production is composed of many components, in which their operation and administration are currently independent. It is required to deploy smart manufacturing workflow for the whole production process, in order to realize centralized effective global scheduling. This requires an advanced administration control platform that has strong abilities of multisource data integration and automatic decision support. To bridge such research gap, this paper designs an optimal scheduling control system for smart manufacturing processes of tobacco industry. First of all, this work discusses major characteristics of future-generation production control patterns in intelligent tobacco factories (ITF). Then, a five-layer architecture for optimal scheduling control of ITF is proposed, which contains Internet-of-Things layer, centralized control layer, model layer, platform layer and operation layer. In addition, a production scheduling optimization strategy is also developed for the proposed system to serve as the software algorithm that drives the running of whole smart manufacturing processes. Finally, this paper presents a comparative analysis of the proposed system’s transformation in a cigarette factory. Naturally, the effectiveness of the proposed production optimization scheduling strategy is verified through simulation.

With the development of emerging information technology, the traditional management methods of marine fishes are slowly replaced by new methods due to high cost, time-consumption and inaccurate management. The update of marine fishes management technology is also a great help for the creation of smart cities. However, some new methods have been studied that are too specific, which are not applicable for the other marine fishes, and the accuracy of identification is generally low. Therefore, this paper proposes an ecological Internet of Things (IoT) framework, in which a lightweight Deep Neural Networks model is implemented as a image recognition model for marine fishes, which is recorded as Fish-CNN. In this study, multi-training and evaluation of Fish-CNN is accomplished, and the accuracy of the final classification can be fixed to 89.89%–99.83%. Moreover, the final evaluation compared with Rem-CNN, Linear Regression and Multilayer Perceptron also verify the stability and advantage of our method.

Virtual reality (VR) is an important landing scenario for 5G and a key enabling technology for the digital economy, and the core challenge it faces is how to ensure real-time interaction and content distribution efficiency. Existing solutions rely on adaptive delivery, overall content edge caching, or asynchronous rendering in the cloud, with drawbacks such as high caching and computing costs and difficulty in ensuring real time. This article is oriented to content generation and synchronization of virtual reality, and aims at realizing intelligent transmission of VR content with high efficiency and low delay guarantee, and carries out research on cloud-edge-end collaborative mechanisms. First, for AI-aided VR content generation, a cloud-edge-end collaborative service architecture is constructed to support independent encoding and distribution of background and interactive content generation. Second, to address the issue that the MEC node with limited cache space cannot meet the dynamical demands of VR user, a cloud-edge collaborative caching strategy based on graph neural networks is proposed to achieve optimal caching and updating of background content, in which the background content is first cached MEC node according to the content request with node sharing scheme, then the content is updated according to a minimal cost update algorithm based on the graph neural networks (GNN)-aided request prediction. Finally, the proposed algorithm is simulated and tested, the simulation results show that the proposed caching and update algorithm achieve better quality of experiment (QoE) and higher cache hit ratio compared with comparison algorithms.

In industrial recirculating aquaculture systems (IRAS), the autonomous decision control of feeding strategies remains a practical concern. Conventionally, control schemes were established from data-driven view, which fails to comprehensively perceive activity status of fishes. To deal with this issue, a deep vision sensing-based fuzzy control scheme is proposed for smart feeding in IRAS. In the first stage, a deep learning-based object detection model is introduced to capture two aspects features as the decision factors: residual bait and eating frequency. In the second stage, a fuzzy neural network model is formulated to calculate control decision strategies via fuzzy inference. And experiments on real-world visual scenes are conducted to verify the proposal.

ABSTRACT Thermally-induced cracking has attracted extensive attention in improving reservoir permeability. In this paper, a thermo-elastic coupling model incorporating the strain-based elastic-brittle damage theory is used to analyze the cracking behaviors of the coal reservoir subjected to cryogenic liquid nitrogen shock. The evolution of temperature and thermally-induced stress with damage is analyzed. The effect of different factors including in-situ stress difference, elastic modulus, thermal expansion coefficient, thermal conductivity and quenching temperature on the induced crack morphology is investigated. It is found that the failure mechanism of the coal rock during cryogenic shock is mainly dominated by elastic brittle tensile damage. The induced fracture morphology is more sensitive to elastic modulus and thermal expansion coefficient relative to in-situ stress difference and thermal conductivity. The increases in elastic modulus and thermal expansion coefficient will bring about more fractures with greater complexity. The higher in-situ stress difference or lower thermal conductivity can generate more short thermal fractures. The critical quenching temperature for inducing thermal cracks around the wellbore is between −110 °C and −105 °C. The results of this study can provide some guidelines for cryogenic fracturing in coal reservoirs. INTRODUCTION Coalbed methane (CBM) reservoirs have the characteristics of low permeability and low porosity. At present, hydraulic fracturing is still the most widely used stimulation technology to improve the overall permeability of unconventional oil and gas reservoirs, including CBM, shale gas and tight gas (Montgomery et al., 2010; Kumari et al., 2018; Ma et al., 2021). However, hydraulic fracturing will bring about a series of thorny problems, such as high fracture initiation pressure in hard formation, pore plugging and water locking effect in water sensitive formation, serious waste of water resources, treatment of flowback fluid and environmental pollution. Compared with hydraulic fracturing, cryogenic liquid nitrogen (LN2) waterless fracturing can overcome these negative problems (Gregory et al., 2011; Rozell et al., 2012, Hung et al., 2020). The representative advantage of LN2 fracturing is that the low temperature LN2 with a boiling point of −195.8 °C produces a huge temperature difference in the rock (Jacobsen et al., 1986), resulting in tens of MPa of thermal stress. In addition, hundreds of times of the gas-liquid ratio of nitrogen also plays an important role in rock cracking. However, because of its high risk of vaporization, it is necessary to control the pressure in the wellbore through the safety valve, thus weakening the effect of nitrogen expansion-induced fracturing (Cha et al., 2018).

Vehicular named data networking (V-NDN) is promising to improve the content delivery efficiency in vehicular ad hoc networks (VANETs). However, the potential broadcast storm caused by Interest packet flooding and return path failures caused by vehicle mobility can significantly degrade the content delivery performance. Existing forwarding strategies based on outdated position information cannot address these issues well. In this paper, we propose a novel predictive forwarding strategy (PRFS) for V-NDN. In PRFS, long short-term memory (LSTM) is employed to amend the neighbor table (NBT) for preciser neighboring vehicles' positions. In addition, the next-hop forwarder is selected among the neighbors, taking into account the link reliability and the distance along the road (DR) in both directions. Furthermore, a new mechanism is designed to notify the selected next-hop forwarder by embedding the forwarder identity in the Interest packet header, so as to accelerate the forwarding process. Finally, extensive simulations are carried out, and experimental results demonstrate that PRFS can reduce the number of forwarded Interest packets and data packets by 21.29% and 25.75%, respectively, and improve the success ratio of satisfied Interest packets by 35.1% compared to the existing baseline algorithms.

In Recirculating Aquaculture System (RAS), feeding frequency is an important factor affecting fish growth, and precise feeding according to the state of fish is the key to improving the aquaculture efficiency. The current problems in detecting the feeding frequency of fish include low efficiency, high technical requirements, and the influenced by the aquaculture environment. In this paper, we adopt a computer vision method to calculate feeding frequency to achieve objective with high accuracy. Firstly, we process the video of the fish according to the inter-frame difference method to obtain the image of the feeding state of the fish. Then we propose a modified VGG16 model to determine the feeding state of the fish, transform it into a 0–1 classification problem and calculate the feeding frequency of the fish. The feeding frequency and the growth status of the fish are then used to develop an intelligent feeding strategy to improve the growth rate of the fish and the conversion rate of the bait. Tests have shown that the accuracy of identifying feeding state by the modified VGG16 model can reach 92.4%. The method has a positive effect on the development of recirculating aquaculture system.

In the process of Recirculating Aquaculture System (RAS), artificial feeding method has many problems, such as inappropriate feeding rhythm and waste of bait, which seriously affect the growth of fish. In order to tackle this problem, we designed an intelligent feeding algorithm applied to the RAS. First, the feeding frequency of fish in the culture pools and the amount of residual bait are acquired by processing the feeding image and video data captured by the camera on top of the culture pools by computer vision method. Then, the subsequent feeding amount are determined by the heuristic algorithm based on the amount of residual bait, the feeding frequency of fish and the other key parameters. The experimental results show that the intelligent feeding algorithm can save about 15% bait than the traditional manual feeding method without affecting the growth of fish.

Summary Cellular vehicular‐to‐everything (C‐V2X), as a key technology of Internet of Vehicles (IoV), is promised to be enhanced and strengthened to improve road traffic safety and achieve intelligent transportation in 6G era. However, computation‐intensive and latency‐sensitive computation tasks of autonomous driving create a great challenge for the computation and storage limited vehicles. Fortunately, mobile edge computing (MEC) architecture offers a possible solution for the challenge. In this paper, a joint computation offloading and resource allocation algorithm is proposed to solve the computation tasks offloading problem in the scenario of vehicular edge computing network. First, the computation offloading and resource allocation are jointly modeled as a mixed integer nonlinear optimization problem. Aiming at minimizing the total system cost (weighted sum of delay and energy consumption), a particle encoding method and a particle refinement algorithm are proposed based on the compression factor particle swarm optimization, and multilevel penalty function is adopted to deal with the constraints of the objective. Finally, experimental performances validate the effectiveness and feasibility of the proposed algorithm.