Search

Antibiotics such as sulfonamides are widely used in agriculture as growth promoters and medicine in treatment of infectious diseases. However, the release of these antibiotics has caused serious environmental problems. In this paper, photocatalytic oxidation technology was used to degrade sulfadiazine (SDZ), one of the typical sulfonamides antibiotics, in UV illuminated TiO2 suspensions. It was found that TiO2 nanosheets (TiO2-NSs) with exposed (001) facets exhibit much higher photoreactivity towards SDZ degradation compared to TiO2 nanoparticles (TiO2-NPs) with a rate constant increases from 0.017 min−1 to 0.035 min−1, improving by a factor of 2.1. Under the attacking of reactive oxygen species (ROSs) such as superoxide radicals ( O2–) and hydroxyl radicals ( OH), SDZ was steady degraded on the surface of TiO2-NSs. Based on the identification of the produced intermediates by LC–MS/MS, possible degradation pathways of SDZ, which include desulfonation, oxidation and cleavage, were put forwards. After UV irradiation for 4 h, nearly 90% of the total organic carbon (TOC) can be removed in suspensions of TiO2-NSs, indicating the mineralization of SDZ. TiO2-NSs also exhibits excellent stability in photocatalytic degradation of SDZ in wide range of pH. Even after recycling used for 7 times, more than 91.3% of the SDZ can be efficiently removed, indicating that they are promising to be practically used in treatment of wastewater containing antibiotics.

At present, lane line detection is still one of the unsolved problems in the field of computer vision. The realization of automatic driving perception technology cannot lack lane line detection. However, most of the lane line detection algorithms proposed at present cannot meet the accuracy and real-time performance at the same time.[1] as a newly proposed polynomial regression lane line detection algorithm, has reached an excellent level in terms of accuracy and speed. This paper further explores on the basis of this paper and proposes a method that does not share Top-y. Experiments show that the newly proposed method is also very effective for lane line detection.

Currently, numerous studies have carried out to research the effect of biochars remediation soil heavy metals (HMs) contaminated, but there have been fewer explorations of the effect of biochars tubule on soil HMs remediation. This work aimed was to study the effect of passivation and remediation of lead (Pb) and chromium (Cr) contaminated soil after insert sewage sludge biochar (SSB) tubule. The results showed that the high risky fractions of Pb and Cr could be transformed into more stable fractions, also, Pb and Cr total contents are significantly decreased by SSB tubule. The mechanisms including adsorption, ion exchange, complexation and precipitation which are concluded from the characteristic analysis. Detailly, the passivation of Pb and Cr are better when the moisture at 25% and 35%, respectively, [Pb: exchangeable (F1), carbonate bound (F2) decreased by 25.1%, 16.8%, Fe-Mn oxides bound (F3) increased by 18.5%; Cr: F1 decreased by 73.0%, F2, F3, Organic matter bound (F4) increased by 13.2%, 23.9%, 30.8%), respectively]. The remediation of Pb and Cr is better when the moisture at 25% and 35%, respectively, (Pb: decreased by 23.3%; Cr: decreased by 38.4%, respectively). The findings showed that the SSB tubule is effective when used for soil HMs contaminated.

With the continuous advancement of science and technology, the artificial intelligence industry is developing more and more rapidly, and the research field focusing on autonomous driving continues to attract young engineers to join in. In the field of autonomous driving, the estimation of pedestrian intention at urban traffic intersections is a very important part of the entire field, and the technical difficulties and challenges faced are undoubtedly the biggest, because the life safety of pedestrians is involved here. Pedestrian trajectory prediction is also a very challenging topic in the field of autonomous driving or assisted driving. In this article, we design a splicing network that combines pedestrian intention and trajectory to build a set of network models that predict pedestrian intention and pedestrian trajectory. We call this model "IAT" (intention and trajectory). We train the collected data and test it on the validation set. The accuracy is 95%. The overlap of the trajectory and the walking trajectory of the human in reality almost overlap within 2s.

Recently, the rapid development of mobile multimedia intelligent voice systems has affected the rapid changes in user interaction. Meanwhile, the continuous optimization of recommended algorithms makes the change of interaction mode from the original user’s active search to the machine actively recommending continuously. It is an inevitable trend that multimedia intelligent voice dialogue system brings more efficient content services to users through active recommendation. However, the existing multimedia intelligent voice dialogue systems are mainly based on the user’s expression content to analyze the user’s subject willingness and then provide the reply contents to user. This kind of interaction only drives the topic through the user, which will cause some problems like single chat topic and missing points of interest. In order to address these problems, this paper proposes an interactive topic guiding model. This model uses the surrounding similar users and user context information to guide users’ topics, and taking into account the harmony of human-computer interaction, the game method is adopted to select the optimal content response strategy by considering the user’s and robot’s emotional friendliness and content friendliness. The experiments show that the model proposed in this paper can effectively guide topics. Moreover, it has significant improvement in content richness compared with other models in human-machine natural interaction, which can provide some research ideas for guiding the topic in the mobile multimedia intelligent voice system.

In order to improve the communication quality of relay-based D2D communication, this paper proposed a relay selection algorithm with the consideration of the information of the social domain and the physical domain. First, the transmission performance of D2D relay communication in the half-duplex relay mode and the full-duplex relay mode is analyzed separately in the physical domain. Second, we analyzed the user’s history check-in information to obtain the characteristics of user’s mobility and the social relationship strength between users. Finally, we proposed a relay communication gain metric to measure the extent of transmission performance improvement brought by D2D relay communication and developed a relay selection algorithm for maximizing the relay communication gain. The simulation results show that compared with the traditional D2D relay communication scheme, the transmission rate and stability of the transmission link of the D2D relay communication system using the proposed algorithm are greatly improved.

Along with the advances of virtual reality (VR) technologies, VR applications are found in various fields. For example, at many major exhibitions, VR is used to mimic the real world through utilizing a number of on-spot sensory equipment to provide immersive user experiences. Despite the advantages that VR brings, the enormous amount of distributed VR videos and the high density of VR users may impose a heavy burden on base stations. On the other hand, the quality of experience (QoE) may seriously suffer from the transmission delay and monetary costs of VR videos. To address the above issues, a D2D-assisted VR video distribution system is designed, and a pre-caching algorithm based on QoE gain is proposed. More specifically, by observing the user residence time at a given node, a user can pre-cache VR videos at the nodes to be visited for improved QoE. To achieve better veracity and efficiency of the pre-caching algorithm, the probability of a user entering each node is calculated based on that user’s interest and nodes’ popularity. Furthermore, the set of candidate schemes can be greatly narrowed down to a range of feasible schemes according to the real-time popularity threshold, the available storage space, and the energy of the user’s equipment. The scheme with maximized user’s QoE can finally be selected from the candidate sets according to user’s QoE gain. Numerical results show that our proposed pre-caching algorithm improves the level of user’s QoE while maintaining superior stability compared with the random-caching and greedy-caching algorithms.

The semantic analysis of computer vision is to let the machine automatically understand, analyze and generate reasonable semantic concepts for the content contained in visual signals such as images or videos. At present, many experts and scholars have carried out in-depth and extensive research on image recognition, and with the rise of neural networks, the use of deep learning should become the current research trend. In this paper, we use the industry-wide deep learning framework caffe training data set, and the results are deployed to Huawei's latest product, hilens, to obtain the final result through the startup file, and display the image and accuracy.

Traffic scene perception is key of autonomous driving. Computer vision and deep learning are popular basis of algorithms in this field. Current researches are mainly on single task, leading to high computing power demands and low operational efficiency. And algorithms may be trained on a small dataset, which means easier overfitting and more susceptible to noises. To overcome above shortcomings, we propose an end-to-end multi-task deep convolutional network model named ShuDA-RFBNet for multiple object detection and drivable area segmentation simultaneously. ShuDA-RFBNet adopts RFBNet for multiple object detection and DenseASPP for drivable area segmentation task. Both networks share a light-weight ShuffleNet V2 as the base net to extract features. ShuDA-RFBNet uses BDD100K to train, which is the current largest and diverse driving video dataset. Test results show that the mAP of multiple object detection task is 32.71% and the MOU of drivable area segmentation task is 82.67%. ShuDA-RFBNet can inference in real time.

The air pollution caused by diesel engine exhaust emission was studied. Firstly, the research direction of selective catalytic reduction technology (SCR) was determined by studying the current emission problems and the technological development status and regulations of various countries in the world. The exhaust emission datas of Wartsila 4L20 diesel engine were measured through the experiment, and the theoretical data of installing SCR system of this diesel engine was obtained through calculation. According to these data, the size of injection pump and catalyst were selected.

Lane detection has always been a critical task in the field of ADAS and autonomous driving, which determines the reliability of lane keeping, adaptive cruise control and path planning. However, the performance of most algorithms in the complex road traffic environment such as blurred lane markings, lane marking occlusion, shaded roads and laneless roads is not satisfactory. In this paper, we propose an end-to-end CNN network. The feature extraction part of the network adopts the most effective DenseNet. The dilated convolution and spatial pyramid pooling are used to expand the receptive field to maximize the context information of the lane marking. In addition, the combination of low-level features and advanced features is used to improve the smoothness of lane segmentation. Finally, a lightweight, fully connected network is used to suppress mis-segmentation in complex scenarios. According to its characteristics, we named the network DCLane. The proposed network is trained on the current largest lane detection dataset, called CULane. DCLane’s performance on the test sets exceeds the state-of-the-art model, while the model meets real-time requirements.

There are some shortcomings in both safety and convenience for existing vehicle unlocking methods, mostly due to the separation between the vehicle and its key. As an improvement, we propose a vehicle unlocking system based on face recognition. The system includes hardware and software. The hardware scheme adopts a modular design, and try to make full use of existing devices of the autonomous driving system and ADAS. The software uses two different deep learning algorithms: FaceNet to verify facial identity, and modified ResNet to distinguish between real faces and secondary faces. A prototype is implemented using the embedded platform Raspberry Pi, and a series of tests are carried out on it. The prototype verifies the feasibility of the hardware scheme and the effectiveness of the algorithms. Test results show that the system is available day and night, quick in unlocking, safe and reliable.

Traffic scenes recognition is the cornerstone of autonomous driving. However, most of the current algorithms are individually trained for tasks such as object detection and road segmentation. In addition, the training data used are mainly concentrated in small datasets such as KITTI, and the trained models are highly susceptible to weather, lighting and other factors. In order to solve the above problems, we propose an end-to-end CNN model for drivable area segmentation and multiple object detection. The feature extraction part of the network is powerful DenseNet. The atrous convolution and spatial pyramid pooling are used for road segmentation, and single shot detection is used for multiple object detection. According to its characteristics, we named the network Dense-ACSSD. Dense-ACSSD is trained on the current largest autonomous driving dataset, called BDD100K. The final training results show that the mIOU of the drivable area segmentation part is as high as 84.15%, and the mAP of the multiple object detection part reaches 30.82%. In addition, the inference time of Dense-ACSSD can meet real-time requirements.

This paper describes how to use online learning platform for traffic light identification, as well as including lane-line identification. First, the traditional traffic light and lane-line identification method were explained; then explaining the concept of neural network and its application in driverless car field; finally, the paper explains how to use the learning platform on the line to train, which can get outputs the model. Using the model, we will get the corresponding results. Based on the continuous optimization of previous studies, this paper makes full use of the advantages of online learning platforms to improve learning methods, to some extent, which enables students to broaden their minds and understand the important position of deep learning in the field of unmanned driving.

Kuang, C.; Liang, H.; Mao, X.; Karney, B.; Gu, J.; Huang, H.; Chen, W., and Song, H., 2017. Influence of potential future sea-level rise on tides in the China Sea. This study investigates the diurnal and semidiurnal tidal responses of the entire China Sea to a potential rise in sea level of 0.5–2 m. A modified two-dimensional tidal model based on MIKE21 is primarily configured and validated for the present situation; then, three (0.5, 1, 2 m) sea-level rise (SLR) scenarios are simulated with this model. The predicted results show that the principal lunar semidiurnal (M2) and diurnal (K1) tidal constituents respond to SLR in a spatially nonuniform manner. Generally, changes of M2 and K1 amplitudes in shallow waters are larger than those in the deep sea, and significant tidal alterations mainly occur in the Bohai and Yellow seas, Jianghua Bay, Hangzhou Bay, Taiwan Strait, Yangtze River estuary, Pearl River estuary, and Beibu Bay. Possible mechanisms further discussed for these changes mainly relate...

As a kind of social attribute of bounded rational users, attitude continuously selects and updates in the process of interaction between nodes, which makes the behavioral game between communication subjects and affects the dissemination of information. To analyze the impact of attitude changes on information, this paper proposes an information propagation model based on evolutionary game. Firstly, from the individual point of view, the update rules of the node's attitude are defined according to non-Bayesian social learning rules. Secondly, the game matrix between nodes based on attitude value is established, and the information dissemination model based on node attitude is established according to the evolutionary analysis paradigm. By replicating the equilibrium solution of dynamic equations with both positive and negative attitudes, and analyzing the stability of the corresponding equilibrium points, the evolution mechanism of dynamic interaction between attitude nodes is obtained. Finally, the impact of different attitudes on information dissemination was analyzed by combining SIS model. The validity of the proposed model is verified by numerical analysis and simulation experiments, and the simulation results show that the change of different attitude of nodes plays an important role in information dissemination.

There are a large number of redundant transmissions in current D2D multicast content delivery systems, which seriously reduces the utilization efficiency of resources. This paper designs a novel user-information-aware D2D video distribution mechanism. More specifically, by predicting users’ video requests, the video can be pushed to potential service requesters while distributing video for service requesters. Firstly, the willingness of potential requesters to accept the pushed video is estimated based on the users’ interests, the popularity of the videos and the residual-energy of the users’ devices, and the user-demand-aware clustering algorithm is proposed. Secondly, considering social and interference information, the utility metric of D2D multicast is proposed to measure the value of content distribution service. Finally, this paper proposes a D2D video distribution mechanism to optimize the utility value. Simulation results show that the proposed mechanism significantly improves throughput, energy and spectrum efficiency compared to the traditional distribution mechanism.

The connection of atomic clusters has been investigated using molecular-dynamics simulation to explain the splitting of the second peak of the pair-distribution function in amorphous alloys. It is found that the unevenness of the connecting style of atomic clusters results in the splitting of the second peak and the two subpeaks is caused by a three-atom-shared connection and a one-atom-shared connection between atomic clusters. The underlying reason is that metallic glasses have higher density than liquid alloys and a different connecting style of atomic clusters from crystals.

This paper studies the melting of Cu-Pd bimetallic clusters with different Pd positions by using molecular dynamics simulation with a general embedded-atom method. The melting of clusters with 55 Pd atomic distributing different positions where the core-layer, second-layer, third-layer, fourth-layer and mixed. It is found that the changing of melting point is strongly related to the position of Pd atomic. The results indicate that the Pd atoms doped in the core layer and surface layer, below the melting point of the second layer and third layer. Meanwhile, this indicate that the Pd atomic doped in the second layer and third layer, the structure of cluster is relatively stable. The irregular phenomena of the melting were induced by the Pd position. This gives a new method to tune the melting point in bimetallic clusters.

The disturbance observer is a specific method of designing a two degree of freedom control structure to achieve insensitivity to modeling error and disturbance rejection. It has been successfully applied in a variety of motion control applications. In motion control, the major sources of uncertainties are friction, inertia, and external disturbances. These uncertainties should be taken into account by any robust motion controller. In this paper, this element is a PD (proportional-derivative) controller. The disturbance observer proves its advantages through the simulation and experiments. With disturbance observer, better tracking performance can be achieved with less control energy.

In bimetallic cluster, research on the frozen structure with the changing concentration plays an important role in exploring new structural materials. This paper studies the freezing processes of (AgCu)309clusters with different Ag concentrations. The results indicated that the structural transformation was strongly related to concentration. It was found that the frozen structures were changed form icosahedron, hcp and fcc-hcp with the change of Ag concentration. The frozen structures were formed icosahedral for the clusters with Ag concentration at 10%, 20%, 30%, and the pure Ag309. For the clusters with Ag content at 40%, 50%, 60%, 70%, and 80%, the frozen structures were formed defect icosahedral. It was also found that the frozen structure have hcp character for the pure Cu309cluster. Meanwhile, the frozen structure of (AgCu)309with 90% Ag concentration was formed fcc-hcp structure. The segregation effects of the Ag-Cu are the key reason for the structural transformation.

In this paper, the structural transitions of the rapidly cooled Ni nanomaterial under different pressures were studied by using molecular dynamics simulation. The work gives the structural properties, including the potential energy, pair-correlation function and Honeycutt-Andersen (HA) and Voronoi indices. This pair-correlation function of liquid Ni agrees well with the experimental results. The local structures are characterized by Honeycutt-Andersen (HA) indices and Voronoi tessellation. Our results indicate that with the increasing pressure the Ni nanomaterial transformed from the disordered structure to the ordered structure. The icosahedral and defect icosahedral decrease, the hcp-like structure increased at 300 K. These phenomena were shown that the crystalline state is hcp-like local structure.

This Delay/Disruption Tolerant Network(DTN) is a New Network Architecture for Communication in Restricted Condition with High Latency and Intermittent Connectivity.In Order to Solve the Problem that Messages can Not be Delivered Reliably because of Mobility and Limited Buffer Size of Nodes in DTN,a Probabilistic Routing Algorithm Based on Node Free Buffer-Utilization (NFBU)is Proposed.The Algorithm Sets Node Free Buffer Warning/excitation Threshold,through Warning/incentive Factor to Increase/decrease the Node Forwarding Probability,and Set up the Buffer Management Strategy,message Time to Live(TTL)value is Smaller,the Higher its Priority,at the same Time Remove the Oldest Message to get more Buffer Space.Simulations Indicate that the Algorithm Improves Delivery Ratio and Reduces Network Latencies in Different Node Free Buffer-Utilization State Compared with some other Routing Algorithms,and it is Highly Adaptable to Network.