Search

Abstract The digital elevation model (DEM) provides important data support for 3D terrain modeling. However, due to the complex and changeable terrain in the real world and the high cost of field measurement, it is extremely difficult to obtain continuous and high-density elevation data directly. Therefore, it is necessary to rely on spatial interpolation technology to restore the DEM overall picture in the original sampling area. The traditional spatial interpolation method usually has the characteristics of low model complexity and high computational cost, which leads to low real-time performance and low precision of the interpolation process. The interpolation operation based on DEM data can be considered as a special image generation process where the input is a DEM image with missing values and the output is a complete DEM image. At present, a large number of studies have proved that deep learning methods are very effective in image generation tasks. However, the training of deep learning models requires the support of a large number of high-quality data sets. DEM data in various countries, especially in key regions, are usually restricted by privacy protection regulations and cannot be disclosed. The emergence of Federated Learning (FL) provides a new solution, which supports local training on multiple end nodes, without sending local data to a remote center server for centralized training, effectively protecting data privacy. In this study, we propose a DEM interpolation model based on FL and multiScale U-Net. The experimental results show that compared with the traditional method, this model has faster processing speed and lower interpolation precision. At the same time, this research result provides a new way for efficient and secure use of terrain information, especially in those application scenarios that have strict requirements for DEM data privacy and security.

The Autonomous Underwater Glider (AUG) is a kind of prevailing underwater intelligent internet vehicle and occupies a dominant position in industrial applications, in which path planning is an essential problem. Due to the complexity and variability of the ocean, accurate environment modeling and flexible path planning algorithms are pivotal challenges. The traditional models mainly utilize mathematical functions, which are not complete and reliable. Most existing path planning algorithms depend on the environment and lack flexibility. To overcome these challenges, we propose a path planning system for underwater intelligent internet vehicles. It applies digital twins and sensor data to map the real ocean environment to a virtual digital space, which provides a comprehensive and reliable environment for path simulation. We design a value-based reinforcement learning path planning algorithm and explore the optimal network structure parameters. The path simulation is controlled by a closed-loop model integrated into the terminal vehicle through edge computing. The integration of state input enriches the learning of neural networks and helps to improve generalization and flexibility. The task-related reward function promotes the rapid convergence of the training. The experimental results prove that our reinforcement learning based path planning algorithm has great flexibility and can effectively adapt to a variety of different ocean conditions.

The development of communication technology will promote the application of Internet of Things, and Beyond 5G will become a new technology promoter. At the same time, Beyond 5G will become one of the important supports for the development of edge computing technology. This paper proposes a communication task allocation algorithm based on deep reinforcement learning for vehicle-to-pedestrian communication scenarios in edge computing. Through trial and error learning of agent, the optimal spectrum and power can be determined for transmission without global information, so as to balance the communication between vehicle-to-pedestrian and vehicle-to-infrastructure. The results show that the agent can effectively improve vehicle-to-infrastructure communication rate as well as meeting the delay constraints on the vehicle-to-pedestrian link.

Remote sensing image analysis plays a vital role in achieving intelligent agricultural monitoring. However, the acquisition of high-resolution agricultural remote sensing data can be resource-intensive, resulting in an imbalance between training samples and artificial intelligence model parameters. In order to achieve accurate agricultural land recognition of limited-resolution remote sensing images, this article proposes a joint network of super-resolution and active learning (AL). The network introduces a pretrained image super-resolution model and optimizes this for remote sensing classification tasks. It effectively detects detailed features and completes the reconstruction. Based on the reconstructed data, an AL algorithm is proposed with a DBSS. It balances the contributions between interclass and boundary samples. Furthermore, we propose a semisupervisory assistance strategy based on consistency, it fully utilizes the predictive power of deep learning models aiming to reduce labeling costs. This framework is proved effective by experiments on an agricultural remote sensing image dataset, it reduces the cost of agricultural data annotation and improves the efficiency of model learning for low-resolution agricultural remote sensing.

As the foundation of human survival, the development of agriculture has always played an important role in social development. The quality of agricultural development determines the speed of social progress. With the development of computer science, using computer technology to solve problems related to agricultural development has become an important research direction in current computer development. In recent years, remote sensing detection has received widespread attention, and the application of remote sensing detection technology in agriculture can provide great convenience for the development of agriculture. Benefit from the development of deep learning, remote sensing detection has made gigantic achievements. However, we have to face some challenges. First, deep learning depends on large scale of data with annotations, which expends inestimable human resources. In addition, as the depth of detection network increases, the amount of parameters explosively extends. In this work, we carry out the research based on the two problems. At the beginning, an active learning method considering classification and localization task is proposed. Our method can choose some few but valuable images. It uses 34.48% amount of training set, up to 97.4% baseline performance, and realize the compression of the scale of dataset, which reduces the trouble of manual labeling. Due to imaging and other factors, there exists many small objects in remote sensing images. So we add the mixed convolution, dilated convolution, and mosaic data augmentation modules into HRDNet network. Experiments on the agriculture dataset indicate that the improved algorithm can obtain about 2% higher than HRDNet. To reduce the number of parameters, we adjust the weighted sum ratio of importance scores dynamically. With the pruning ratio of 80%, the model volume has only 184 MB, degrading 70%. Model compression accelerates the detection speed for seven times on NVIDIA AGX Xavier, with a speed of 6 FPS. Our work will lay a foundation for remote sensing detection.

The detection rate of mental health problems among undergraduates has recently risen significantly. However, undergraduates underutilize mental health services; approximately a third only of undergraduates in need of treatment use school counseling resources. Based on a social psychological theoretical framework, the health belief model, factors of undergraduates' willingness to seek help when dealing with psychological problems were investigated. A cross-sectional online questionnaire and a snowball sampling method with 446 undergraduates investigated perceived susceptibility, perceived severity, perceived behavioral benefits, perceived barriers, self-efficacy, and cues to action to understand how students' mental health-seeking behaviors are affected. We found that perceived susceptibility (p < 0.01), perceived severity (p < 0.01), perceived benefits (p < 0.01), perceived barriers (p < 0.01), self-efficacy (p < 0.01), and cues to action (p < 0.01) significantly correlated with behavioral intention. Encouragement or counseling from others would be more likely to motivate undergraduates to seek mental health help. In addition, we used a bias-corrected Bootstrap approach to test the significance of the mediating effect, the mediation effect of cues to action between undergraduates' perceived susceptibility and mental health help-seeking behavior was utterly significant [mediation effect value of 0.077, with an SE value of 0.027 and a 95% CI (0.028, 0.133)]. It demonstrated that those who perceived themselves to be at high risk of developing a mental illness and who had received encouragement or counseling to seek mental health help were more likely to be motivated to seek mental health help. Multiple regression analyses indicated that self-efficacy (Z = 5.425, p < 0.01) and cues to action (Z = 6.673, p < 0.01) independently influenced behavioral intentions. Encouragement or counseling from others would be more likely to motivate undergraduates to seek mental health help.

The ongoing data explosion introduced unprecedented challenges to the information security of communication networks. As images are one of the most commonly used information transmission carriers; therefore, their data redundancy analysis and screening are of great significance. However, most of the current research focus on the algorithm improvement of commonly used image datasets. Thus, we should consider an important question: Is there data redundancy in the open datasets? Considering the factors of model structures and data distribution to ensure the generalization, we conducted extensive experiments to compare the average accuracy based on few random data to the baseline accuracy based on all data. The results show serious data redundancy in the open datasets from different domains. For instance, with the aid of deep model, only 20% data can achieve more than 90% of the baseline accuracy. Further, we proposed a novel entropy-based information screening method, which outperforms the random sampling under many experimental conditions. In particular, considering 20% of data, for the shallow model, the improvement is approximately 10%, and for the deep model, the ratio to the baseline accuracy increases to greater than 95%. Moreover, this work can also serve as a new way of learning from a few valuable samples, compressing the size of existing datasets and guiding the construction of high-quality datasets in the future.

To efficiently complete a complex computation task, the complex task should be decomposed into sub-computation tasks that run parallel in edge computing. Wireless Sensor Network (WSN) is a typical application of parallel computation. To achieve highly reliable parallel computation for wireless sensor network, the network's lifetime needs to be extended. Therefore, a proper task allocation strategy is needed to reduce the energy consumption and balance the load of the network. This paper proposes a task model and a cluster-based WSN model in edge computing. In our model, different tasks require different types of resources and different sensors provide different types of resources, so our model is heterogeneous, which makes the model more practical. Then we propose a task allocation algorithm that combines the Genetic Algorithm (GA) and the Ant Colony Optimization (ACO) algorithm. The algorithm concentrates on energy conservation and load balancing so that the lifetime of the network can be extended. The experimental result shows the algorithm's effectiveness and advantages in energy conservation and load balancing.

Linear piezoelectric materials offer new opportunities for next-generation multifunctional micro/nano systems on chip aside from ferroelectrics. However, the advantage of quantitative pressure sensing of linear piezoelectric materials over ferroelectrics has not been demonstrated, while the thermal stability of linear piezoelectric zinc oxide (ZnO) is yet to be evaluated. In this work, pressure sensors based on linear piezoelectric undoped ZnO and ferroelectric vanadium doped ZnO (V-ZnO) films are studied side by side. Both films possess similar fiber textures oriented along the polar ZnO [002] direction, yet each of the V-ZnO grains consists of strip-like domains observed by SEM. Applying triangular-wave pressures, the linear piezoelectric ZnO sensor generates triangular-wave voltage signals that are one-to-one linearly correlated to the applied pressure, showing decisive advantages for quantitative pressure sensing compared to the ‘sail-like’ voltage signals of the ferroelectric V-ZnO sensor that is many-to-one correlated to the applied pressure. Superior thermal stability is demonstrated in the linear piezoelectric ZnO sensor showing stable pressure sensing properties up to 450 °C compared to the degraded sensing properties of the ferroelectric V-ZnO sensor at 150 °C. The thermally stable piezoelectric sensors for quantitative pressure sensing based on linear piezoelectric ZnO films serve as candidates for high-performance multifunctional micro/nano systems.

Electric power audit text classification is one of the important research problem in electric power systems. Recently, kinds of automatic classification methods for these texts based on machine learning or deep learning models have been applied. At present, the development of computing technology makes “pre-training and fine-tuning” the newest paradigm of text classification, which achieves better results than previous fully-supervised models. Based on pre-training theory, domain-related pre-training tasks can enhance the performance of downstream tasks in the specific domain. However, existing pre-training models usually use general corpus for pre-training, and do not use texts related to the field of electric power, especially electric power audit texts. This results in that the model does not learn too much electric-power-related morphology or semantics in the pre-training stage, so that less information can be used in the fine-tuning stage. Based on the research status, in this paper, we propose EPAT-BERT, a BERT-based model pre-trained by two-granularity pre-training tasks: word-level masked language model and entity-level masked language model. These two tasks predict word and entity in electric-power-related texts to learn abundant morphology and semantics about electric power. We then fine-tune EPAT-BERT for electric power audit text classification task. The experimental results show that, compared with fully supervised machine learning models, neural network models, and general pre-trained language models, EPAT-BERT can significantly outperform existing models in a variety of evaluation metrics. Therefore, EPAT-BERT can be further applied to electric power audit text classification. We also conduct ablation studies to prove the effectiveness of each component in EPAT-BERT to further illustrate our motivations.

With the advent of the 5G Internet of Things era, communication and social interaction in our daily life have changed a lot, and a large amount of social data is transmitted to the Internet. At the same time, with the rapid development of deep forgery technology, a new generation of social data trust crisis has also followed. Therefore, how to ensure the trust and credibility of social data in the 5G Internet of Things era is an urgent problem to be solved. This paper proposes a new method for forgery detection based on GANs. We first discover the hidden gradient information in the grayscale image of the forged image and use this gradient information to guide the generation of forged traces. In the classifier, we replace the traditional binary loss with the focal loss that can focus on difficult-to-classify samples, which can achieve accurate classification when the real and fake samples are unbalanced. Experimental results show that the proposed method can achieve high accuracy on the DeeperForensics dataset and with the highest accuracy is 98%.

In this study, we have fabricated a 368 nm LED with an epitaxial Indium Tin Oxide (ITO) contact layer. We analyze the thermal performance of the flip-chip LED with a symmetric electrode and metal reflective layer, applying ANSYS to build a coupled electro-thermal finite element model (FEM) of the temperature distribution in the multiple quantum wells (MQWs). We compare our system with the traditional Au-bump flip-chip LED and a flip-chip LED with a Distributed Bragg Reflector (DBR) layer. The simulation results have shown that the flip-chip LED with a metal reflective layer and symmetric electrode exhibits better heat dissipation performance, particularly at high input power. The influence of the insulating layer on the LED chip junction temperature is also examined. The simulation data establish an effect due to the thermal conductivity of the insulating layer in terms of heat dissipation, but this effect is negligible at an insulation layer thickness ≤1 µm.

Abstract With the rise of artificial intelligence, deep learning is gradually applied to the field of agriculture and plant science. However, the excellent performance of deep learning needs to be established on massive numbers of samples. In the field of plant science and biology, it is not easy to obtain a large amount of labeled data. The emergence of few-shot learning solves this problem. It imitates the ability of humans’ rapid learning and can learn a new task with only a small number of labeled samples, which greatly reduces the time cost and financial resources. At present, the advanced few-shot learning methods are mainly divided into four categories based on: data augmentation, metric learning, external memory, and parameter optimization, solving the over-fitting problem from different viewpoints. This review comprehensively expounds on few-shot learning in smart agriculture, introduces the definition of few-shot learning, four kinds of learning methods, the publicly available datasets for few-shot learning, various applications in smart agriculture, and the challenges in smart agriculture in future development.

Sea surface temperature (SST) is an important parameter to measure the energy and heat balance of sea surface. The change of sea surface temperature has an important impact on the marine ecosystem, marine climate and marine environment. Therefore, sea surface temperature prediction has become an significant research direction in the field of ocean. This article proposes a DBULSTM-Adaboost model based on ensemble learning. The model is composed of Deep Bidirectional and Unidirectional Long Short Term Memory (DBULSTM) and Adaboost strong learner. DBULSTM can capture the forward and backward dependence of time series, and the DBULSTM model is integrated with Adaboost strong learner to reduce the variance and bias of prediction and realize the short and medium term prediction of SST at a single point scale. Experimental results show that the model can improve the accuracy and stability of SST prediction. Experiments on the East China Sea and South China Sea with different prediction lengths show that the model is almost superior to other classical models in different sea areas and at different prediction levels. Compared with full-connected LSTM (FC-LSTM) model, the root-mean-square error is reduced by about 0.1.

In the field of ocean energy detection, Autonomous Underwater Vehicles (AUVs) offer significant advantages in terms of manpower, resource, and energy efficiency. However, the unpredictable nature of the ocean environment, particularly the real-time changes in ocean currents, poses navigational risks for AUVs. Therefore, effective path planning in dynamic environments is crucial for AUVs to perform specific tasks. This paper addresses the static path planning problem and proposes a model called the noise net double DQN network with prioritized experience replay (N-DDQNP). The N-DDQNP model combines a noise network and a prioritized experience replay mechanism to address the limited exploration and slow convergence speed issues of the DQN algorithm, which are caused by the greedy strategy and uniform sampling mechanism. The proposed approach involves constructing a double DQN network with a priority experience replay and an exploration mechanism using the noise network. Second, a compound reward function is formulated to take into account ocean current, distance, and safety factors, ensuring prompt feedback during the training process. Regarding the ocean current, the reward function is designed based on the angle between the current direction and the AUV’s heading direction, considering its impact on the AUV’s speed. As for the distance factor, the reward is determined by the Euclidean distance between the current position and the target point. Furthermore, the safety factor considers whether the AUV may collide with obstacles. By incorporating these three factors, the compound reward function is established. To evaluate the performance of the N-DDQNP model, experiments were conducted using real ocean data in various complex ocean environments. The results demonstrate that the path planning time of the N-DDQNP model outperforms other algorithms in different ocean current scenarios and obstacle environments. Furthermore, a user console-AUV connection has been established using spice cloud desktop technology. The cloud desktop architecture enables intuitive observation of the AUV’s navigation posture and the surrounding marine environment, facilitating safer and more efficient underwater exploration and marine resource detection tasks.

With the rapid development of information and communication technologies, artificial intelligence and IoTs, more and more advanced technologies, such as machine learning, reinforcement learning, neural networks and fuzzy systems, have been introduced into industrial practices. The application of advanced technologies has greatly promoted the process of industrial revolution. However, there is big gap between controlled simulation and real evolving environment, which results in the unsatisfactory performance of the typical algorithms in practical environments. For example, in Underwater IoTs, a dynamic and uncertain marine environment can cause equipment damage, resulting in huge financial losses. Therefore, improving the robustness and adaptability of algorithms and systems, and proposing new solutions in practical applications to meet the requirements of self-developing, self-organizing, and evolving systems is essential to promote intelligent industrial applications.

Abstract Illustrations frequently appear in advertisements, films, books, etc. Research into the popularity of illustrations shows practicality and necessity already, which plays an important role in the improvement of illustrations and advertising decisions. In this letter, an Illustration Popularity Assessment Network （IPANet） is designed, inspired by the illustration creation process. Atrous convolutional layers are densely concatenated to extract image composition information and Gram matrix is used to measure the image style. We also introduce a scoring model based on ‘views’ and ‘likes’ to score the popularity of collected illustrations. The IPANet is proved to be effective according to experimental results on the scored illustrations. As far as we know, we are the first to explore the popularity of illustrations and practice successfully.

Wearable sensing electronic systems (WSES) are becoming a fundamental platform to construct smart and intelligent networks for broad applications. Various physiological data are readily collected by the WSES, including biochemical, biopotential, and biophysical signals from human bodies. However, understanding these sensing data, such as feature extractions, recognitions, and classifications, is largely restrained because of the insufficient capacity when using conventional data processing techniques. Recent advances in sensing performance and system‐level operation quality of the WSES are expedited with the assistance of machine learning (ML) algorithms. Here, the state‐of‐the‐art of the ML‐assisted WSES is summarized with emphasis on how the accurate perceptions on physiological signals under different algorithms paradigm augment the performance of the WSES for diverse applications. Concretely, ML algorithms that are frequently implemented in the WSES studies are first synopsized. Then broad applications of ML‐assisted WSES with strengthened functions are discussed in the following sections, including intelligent physiological signals monitoring, disease diagnosis, on‐demand treatments, assistive devices, human–machine interface, and multiple sensations‐based virtual and augmented reality. Finally, challenges confronted for the ML‐assisted WSES are addressed.

Abstract Plastic waste caused by the extensive usage of face masks during COVID‐19 pandemic has become a severe threat to natural environment and ecosystem. Herein, an eco‐friendly approach to repurpose face mask waste for clean water production via solar thermal evaporation is proposed. By taking advantage of its interwind structure, face mask holds the promise to be an ideal candidate material for constructing photothermal evaporator. In‐situ surface modifications are performed successively with polyvinyl alcohol and polypyrrole to improve its wettability and solar absorption (97%). The obtained face mask‐based evaporator achieves significantly enhanced solar efficiency (91.5%) and long‐term salt‐rejection stability. The harvested clean water befits plant growing to enable farming on sea surface. A floating photothermal evaporation prototype is then developed to demonstrate autonomous solar ocean farming, with plants successfully cultivated over time. As such, the proposed strategy provides a promising solution towards ecological sustainability by tapping multiple benefits.

In this letter, we describe the design and fabrication of high-power AlGaN-based ultraviolet (UV) flip-chip high-voltage light-emitting diodes (LEDs) operating at 368 nm with an epitaxial indium tin oxide (ITO)/Al reflecting mirror and symmetry electrode layout. Metal-organic chemical vapor deposition (MOCVD) was used to grow an ITO thin film as a transparent electrode on the LED surface. At 365 nm, epitaxial ITO thin films exhibited a transmittance of up to 93.6%. Additionally, the epitaxial ITO/Al reflective mirror has a reflectance of 81.2% at 365-nm. To investigate the electrical characteristics, four types of HV-LED micro-cells were constructed with varying n-type mesa structures and p-type interconnect electrodes. We demonstrated a forward voltage (Vf) of 7.86 V at 350 mA with a 2 × 2 mico-cells high-voltage ultraviolet 368-nm flip-chip LED after optimising electrode structure and device process.

Scientific and moderate exercise is an effective remedy to boost the immune system and minimize the risk of infection in coronavirus disease 2019 (COVID-19) epidemic period. In this context, social media platforms play an essential role in communicating scientific fitness at home. Based on the integrated analysis of communication, physical geography, and health geography, this article takes short fitness-based videos on TikTok as the research object and comprehensively uses text analysis, mathematical statistics, and spatial analysis to reveal the characteristics of communication elements, fitness preferences, and the spatiotemporal evolution of these short videos, and to explore the communication mechanism and optimization path of short videos. The results show that: (1) the communication information focusing on “fighting against the epidemic and keeping fit at home” is made into short videos by female youth. Through the communication channels mainly composed of texts, pictures, and audio, the communication subjects carry out multilevel interactive communication with the communication objects; (2) fitness preferences are mainly composed of dances, unarmed fitness, and fitness with equipment; (3) from the perspective of time, taking February 20 as the node, the daily number of released short videos shows a trend of “increase before decrease”; from the perspective of space, the distribution is significantly unbalanced across provinces, and the direction is not apparent; and (4) this article explores the communication mechanism and targeted optimization countermeasures of short fitness-based videos on TikTok in China's COVID-19 epidemic period.

It is of great significance to accurately detect ships on the ocean. To obtain higher detection performance, many researchers use deep learning to identify ships from images instead of traditional detection methods. Nevertheless, the marine environment is relatively complex, making it quite difficult to determine features of ship targets. In addition, many detection models contain a large amount of parameters, which is not suitable to deploy in devices with limited computing resources. The two problems restrict the application of ship detection. In this paper, firstly, an SAR ship detection dataset is built based on several databases, solving the problem of a small number of ship samples. Then, we integrate the SPP, ASFF, and DIOU-NMS module into original YOLOv3 to improve the ship detection performance. SPP and ASFF help enrich semantic information of ship targets. DIOU-NMS can lower the false alarm. The improved YOLOv3 has 93.37% mAP, 4.11% higher than YOLOv3 on the self-built dataset. Then, we use the MCP method to compress the improved YOLOv3. Under the pruning ratio of 80%, the obtained compressed model has only 6.7 M parameters. Experiments show that MCP outperforms NS and ThiNet. With the size of 26.8 MB, the compact model can run at 15 FPS on an NVIDIA TX2 embedded development board, 4.3 times faster than the baseline model. Our work will contribute to the development and application of ship detection on the sea.

Combining remote sensing images with deep learning algorithms plays an important role in wide applications. However, it is difficult to have large-scale labeled datasets for remote sensing images because of acquisition conditions and costs. How to use the limited acquisition budget to obtaina better remote sensing image dataset is a problem worth studying. In response to this problem, this paper proposes a remote sensing image quality evaluation method based on node entropy, which can be combined with active learning to provide low-cost guidance for remote sensing image collection and labeling. The method includes a node selection module and a remote sensing image quality evaluation module. The function of the node selection module is to select representative images, and the remote sensing image quality evaluation module evaluates the remote sensing image information quality by calculating the node entropy of the images. The image at the decision boundary of the existing images has a higher information quality. To validate the method proposed in this paper, experiments are performed on two public datasets. The experimental results confirm the superiority of this method compared with other methods.

In the era of rapid development of the Internet of things, deep learning, and communication technologies, social media has become an indispensable element. However, while enjoying the convenience brought by technological innovation, people are also facing the negative impact brought by them. Taking the users’ portraits of multimedia systems as examples, with the maturity of deep facial forgery technologies, personal portraits are facing malicious tampering and forgery, which pose a potential threat to personal privacy security and social impact. At present, the deep forgery detection methods are learning-based methods, which depend on the data to a certain extent. Enriching facial anti-spoofing datasets is an effective method to solve the above problem. Therefore, we propose an effective face swapping framework based on StyleGAN. We utilize the feature pyramid network to extract facial features and map them to the latent space of StyleGAN. In order to realize the transformation of identity, we explore the representation of identity information and propose an adaptive identity editing module. We design a simple and effective post-processing process to improve the authenticity of the images. Experiments show that our proposed method can effectively complete face swapping and provide high-quality data for deep forgery detection to ensure the security of multimedia systems.

Agricultural robots are one of the important means to promote agricultural modernization and improve agricultural efficiency. With the development of artificial intelligence technology and the maturity of Internet of Things (IoT) technology, people put forward higher requirements for the intelligence of robots. Agricultural robots must have intelligent control functions in agricultural scenarios and be able to autonomously decide paths to complete agricultural tasks. In response to this requirement, this paper proposes a Residual-like Soft Actor Critic (R-SAC) algorithm for agricultural scenarios to realize safe obstacle avoidance and intelligent path planning of robots. In addition, in order to alleviate the time-consuming problem of exploration process of reinforcement learning, this paper proposes an offline expert experience pre-training method, which improves the training efficiency of reinforcement learning. Moreover, this paper optimizes the reward mechanism of the algorithm by using multi-step TD-error, which solves the probable dilemma during training. Experiments verify that our proposed method has stable performance in both static and dynamic obstacle environments, and is superior to other reinforcement learning algorithms. It is a stable and efficient path planning method and has visible application potential in agricultural robots.

Abstract High levels of humidity can induce thermal discomfort and consequent health disorders. However, proper utilization of this astounding resource as a freshwater source can aid in alleviating water scarcity. Herein, a low‐energy and highly efficient humidity control system is reported comprising of an in‐house developed desiccant dehumidifier and hygrometer (sensor), with an autonomous operation capability that can realize simultaneous dehumidification and freshwater production. The high efficiency and energy saving mainly come from the deployed super hygroscopic complex (SHC), which exhibits high water uptake (4.64 g g−1) and facile regeneration. Machine‐learning‐assisted in‐house developed low cost and high precision hygrometers enable the autonomous operation of the humidity management system. The dehumidifier can reduce the relative humidity (RH) of a confined room from 75% to 60% in 15 minutes with energy consumption of 0.05 kWh, saving more than 60% of energy compared with the commercial desiccant dehumidifiers, and harvest 10 L of atmospheric water in 24 h. Moreover, the reduction in RH from 80% to 60% at 32 °C results in the reduction of apparent temperature by about 7 °C, thus effectively improving the thermal comfort of the inhabitants.

Marine surveying is an important part of marine environment monitoring systems. In order to improve the accuracy of marine surveying and reduce investment in artificial stations, it is necessary to use high-precision GNSS for shipborne navigation measurements. The basic measurement is based on the survey lines that are already planned by surveyors. In response to the needs of survey vessels sailing to the survey line, a method framework for the shortest route planning is proposed. Then an intelligent navigation system for survey vessels is established, which can be applied to online navigation of survey vessels. The essence of the framework is that the vessel can travel along the shortest route to the designated survey line under the limitation of its own minimum turning radius. Comparison and analysis of experiments show that the framework achieves better optimization. The experimental results show that our proposed method can enable the vessel to sail along a shorter path and reach the starting point of the survey line at the specified angle.

Continuous respiratory monitoring is an important tool for clinical monitoring. The most widely used flow measure device is nasal cannulae connected to a pressure transducer. However, most of these devices are not easy to carry and continue working in uncontrolled environments which is also a problem. For portable breathing equipment, due to the volume limit, the pressure signals acquired by using the airway tube may be too weak and contain some noise, leading to huge errors in respiratory flow measures. In this paper, a cost-effective portable pressure sensor-based respiratory measure device is designed. This device has a new airway tube design, which enables the pressure drop efficiently after the air flowing through the airway tube. Also, a new back propagation (BP) neural network-based algorithm is proposed to stabilize the device calibration and remove pressure signal noise. For improving the reliability and accuracy of proposed respiratory device, a through experimental evaluation and a case study of the proposed BP neural network algorithm have been carried out. The results show that giving proper parameters setting, the proposed BP neural network algorithm is capable of efficiently improving the reliability of newly designed respiratory device.

With continuous progress of Internet of Things, multimedia analysis in it has attracted more and more attention. Specially, stereoscopic display technology plays an important role in the multimedia analysis processing. In the Internet of Things system, the quality of stereoscopic image will be reduced in the transmission process. In this mode, it will have a great impact on multimedia analysis to judge whether the quality of stereoscopic image meets the requirements. In this paper, a new no-reference stereoscopic image quality assessment model for multimedia analysis towards Internet of Things is built, which is based on a deep learning model to learn from the class labels and image representations. In our framework, images are represented by natural scene statistics features that are extracted from discrete cosine transform domain, and a regression model is employed to shine upon the quality from the feature vector. The training process of the proposed model contains an unsupervised pretraining phase and a supervised fine-tuning phase, enabling it to generalize over the whole distortion types and severity. The proposed model greatly shows the correlation with subjective assessment as demonstrated by experiments on the LIVE 3-D Image Quality Database and IVC 3-D Image Quality Database.

Virtual reality (VR), a new type of simulation and interaction technology, has aroused widespread attention and research interest. It is necessary to evaluate the VR quality and provide a standard for the rapidly developing technology. To the best of our knowledge, a few researchers have built benchmark databases and designed related algorithms, which has hindered the further development of the VR technology. In this paper, a free available data set (VRQ-TJU) for VR quality assessment is proposed with subjective scores for each sample data. The validity for the designed database has been proved based on the traditional multimedia quality assessment metrics. In addition, an end-to-end 3-D convolutional neural network is introduced to predict the VR video quality without a referenced VR video. This method can extract spatiotemporal features and does not require using hand-crafted features. At the same time, a new score fusion strategy is designed based on the characteristics of the VR video projection process. Taking the pre-processed VR video patches as input, the network captures local spatiotemporal features and gets the score of every patch. Then, the new quality score fusion strategy is applied to get the final score. Such approach shows advanced performance on this database.

In this paper, a novel image quality assessment (IQA) model using statistic features based on image information theory is proposed. Firstly, the image is decomposed into non-overlapped patches and transformed into the saliency information, specific information and entanglement information. Then considering the perception characteristics of human visual system, e.g. the center-surrounded of the receptive field and the contrast-gain masking process, have an important influence on image quality evaluation, the statistic features of the image information were employed to describe the image local contrast, structure, multi-scale and multi-direction properties, include the mean subtracted and contrast normalized (MSCN) features, the gradient magnitude (GM) features and the Laplacian of Gaussian (LOG) features. Finally, the mapping between the statistic features and the human subjective perception is established and used to measure the image quality. Experimental results on benchmark databases (LIVE, TID2013, CSIQ) indicate the rationality and validity of the proposed method.

Stereo matching is one of the most important and challenging subjects in the field of stereo vision. The disparity obtained in stereo matching can represent depth information in 3-D world to a great extent and shows great importance in stereo field. In general, stereo-matching methods primarily emphasize static image. However, the information provided by dynamic scene can be used fully and effectively to improve the results of stereo matching for dynamic scene, such as video sequences. In this paper, we propose a dynamic scene-based local stereo-matching algorithm which integrates a cost filter with motion flow of dynamic video sequences. In contrast to the existing local approaches, our algorithm puts forward a new computing model which fully considers motion information in dynamic video sequences and adds motion flow to calculate suitable support weight for accurately estimating disparity. Our algorithm can perform as an edge-preserving smoothing operator and shows improved behavior near the moving edges. The experimental results show that the proposed method achieves a better depth map and outperforms other local stereo-matching methods in disparity evaluation.

In the above article [1], there is minor mistake caused by numerical type position. The number 6.4912 in TABLE 4 should be typed as 0.7577. The number 0.5777 in TABLE 5 should be typed as 6.4912.

Carbon monoxide (CO) burns or explodes at over-standard concentration. Hence, in this paper, a Wifi-based, real-time monitoring of a CO system is proposed for application in the construction industry, in which a sensor measuring node is designed by low-frequency modulation method to acquire CO concentration reliably, and a digital filtering method is adopted for noise filtering. According to the triangulation, the Wifi network is constructed to transmit information and determine the position of nodes. The measured data are displayed on a computer or smart phone by a graphical interface. The experiment shows that the monitoring system obtains excellent accuracy and stability in long-term continuous monitoring.

Continuous respiratory monitoring is an important tool for clinical monitoring. Associated with the development of biomedical technology, it has become more and more important, especially in the measuring of gas flow and CO2 concentration, which can reflect the status of the patient. In this paper, a new type of biomedical device is presented, which uses low-power sensors with a piezoresistive silicon differential pressure sensor to measure gas flow and with a pyroelectric sensor to measure CO2 concentration simultaneously. For the portability of the biomedical device, the sensors and low-power measurement circuits are integrated together, and the airway tube also needs to be miniaturized. Circuits are designed to ensure the stability of the power source and to filter out the existing noise. Modulation technology is used to eliminate the fluctuations at the trough of the waveform of the CO2 concentration signal. Statistical analysis with the coefficient of variation was performed to find out the optimal driving voltage of the pressure transducer. Through targeted experiments, the biomedical device showed a high accuracy, with a measuring precision of 0.23 mmHg, and it worked continuously and stably, thus realizing the real-time monitoring of the status of patients.

The realization of the Internet of Things greatly depends on the information communication among physical terminal devices and informationalized platforms, such as smart sensors, embedded systems and intelligent networks. Playing an important role in information acquisition, sensors for stereo capture have gained extensive attention in various fields. In this paper, we concentrate on promoting such sensors in an intelligent system with self-assessment capability to deal with the distortion and impairment in long-distance shooting applications. The core design is the establishment of the objective evaluation criteria that can reliably predict shooting quality with different camera configurations. Two types of stereo capture systems—toed-in camera configuration and parallel camera configuration—are taken into consideration respectively. The experimental results show that the proposed evaluation criteria can effectively predict the visual perception of stereo capture quality for long-distance shooting.

Stereo cameras are the basic tools used to obtain stereoscopic image pairs, which can lead to truly great image quality. However, some inappropriate shooting conditions may cause discomfort while viewing stereo images. It is therefore considerably necessary to establish the perceptual criteria that can be used to evaluate the shooting quality of stereo cameras. This article proposes objective quality evaluation criteria based on the characteristics of parallel and toed-in camera configurations. Considering the different internal structures and basic shooting principles, this paper focuses on short-distance shooting conditions and establishes assessment criteria for both parallel and toed-in camera configurations. Experimental results show that the proposed evaluation criteria can predict the visual perception of stereoscopic images and effectively evaluate stereoscopic image quality.

With the explosive growth of ocean data, it is of great significance to use ocean observation data to analyze ocean pycnocline data in military field. However, due to natural factors, most of the time the ocean hydrological data is not complete. In this case, predicting the ocean hydrological data by partial data has become a hot spot in marine science. In this paper, based on the traditional statistical analysis literature, we propose a machine-learning ocean hydrological data processing process under big data. At the same time, based on the traditional pycnocline gradient determination method, the open Argo data set is analyzed, and the local characteristics of pycnocline are verified from several aspects combined with the current research about pycnocline. Most importantly, in this paper, the combination of kernel function and support vector machine(SVM) is extended to nonlinear learning by using the idea of machine learning and convex optimization technology. Based on this, the known pycnocline training set is trained, and an accurate model is obtained to predict the pycnocline in unknown domains. In the specific steps, this paper combines the classification problem with the regression problem, and determines the proportion of training set and test formula set by polynomial regression. Subsequently, the feature scaling of the input data accelerated the gradient convergence, and a grid search algorithm with variable step size was proposed to determine the super parameter c and gamma of the SVM model. The prediction results not only used the confusion matrix to analyze the accuracy of GridSearch-SVM with variable step size, but also compared the traditional SVM and the similar algorithm. At the end of the experiment, two features which have the greatest influence on the Marine density thermocline are found out by the feature ranking algorithm based on learning.

The human visual system (HVS) plays an important role in stereo image quality perception. Therefore, it has aroused many people's interest in how to take advantage of the knowledge of the visual perception in image quality assessment models. This paper proposes a full-reference metric for quality assessment of stereoscopic images based on the binocular difference channel and binocular summation channel. For a stereo pair, the binocular summation map and binocular difference map are computed first by adding and subtracting the left image and right image. Then the binocular summation is decoupled into two parts, namely additive impairments and detail losses. The quality of binocular summation is obtained as the adaptive combination of the quality of detail losses and additive impairments. The quality of binocular summation is computed by using the Contrast Sensitivity Function (CSF) and weighted multi-scale (MS-SSIM). Finally, the quality of binocular summation and binocular difference is integrated into an overall quality index. The experimental results indicate that compared with existing metrics, the proposed metric is highly consistent with the subjective quality assessment and is a robust measure. The result have also indirectly proved hypothesis of the existence of binocular summation and binocular difference channels.

Wireless local area network (WLAN) is a technology that combines computer network with wireless communication technology. The 2.4 GHz and 5 GHz frequency bands in the Industrial Scientific Medical (ISM) band can be used in the WLAN environment. Because of the development of wireless communication technology and the use of the frequency bands without the need for authorization, the application of WLAN is becoming more and more extensive. As the key part of the WLAN system, the antenna must also be adapted to the development of WLAN communication technology. This paper designs two new dual-frequency microstrip antennas with the use of electromagnetic simulation software—High Frequency Structure Simulator (HFSS). The two antennas adopt ordinary FR4 material as a dielectric substrate, with the advantages of low cost and small size. The first antenna adopts microstrip line feeding, and the antenna radiation patch is composed of a folded T-shaped radiating dipole which reduces the antenna size, and two symmetrical rectangular patches located on both sides of the T-shaped radiating patch. The second antenna is a microstrip patch antenna fed by coaxial line, and the size of the antenna is diminished by opening a stepped groove on the two edges of the patch and a folded slot inside the patch. Simulation experiments prove that the two designed antennas have a higher gain and a favourable transmission characteristic in the working frequency range, which is in accordance with the requirements of WLAN communication.

The Internet of Things is built based on various sensors and networks. Sensors for stereo capture are essential for acquiring information and have been applied in different fields. In this paper, we focus on the camera modeling and analysis, which is very important for stereo display and helps with viewing. We model two kinds of cameras, a parallel and a converged one, and analyze the difference between them in vertical and horizontal parallax. Even though different kinds of camera arrays are used in various applications and analyzed in the research work, there are few discussions on the comparison of them. Therefore, we make a detailed analysis about their performance over different shooting distances. From our analysis, we find that the threshold of shooting distance for converged cameras is 7 m. In addition, we design a camera array in our work that can be used as a parallel camera array, as well as a converged camera array and take some images and videos with it to identify the threshold.