Accurate prediction of pedestrian motion is crucial for autonomous driving, particularly in path planning and collision avoidance applications. Most current methods concentrate on spatiotemporal feature parameters (e.g., velocity continuity, social force parameters, and poses) extracted from historical trajectories to model pedestrian movement. However, these methodologies fail to adequately capture pedestrian intent and do not dynamically account for behavioral heterogeneity, leading to significant discrepancies with real-world observations. To address this issue, a dynamic pedestrian intention understanding (DPIU) framework is proposed, which links future intentions to historical experiences. Grounded in cognitive decision-making mechanisms derived from human physiology, the DPIU framework is designed to predict pedestrian motion by inferring inherent movement intentions. To establish a comprehensive historical perspective, a multiscale detail feature module is employed, incorporating a time-scale-based trajectory segmentation strategy to enhance the representation of pedestrian states. Subsequently, a goal intent prediction module is introduced, employing a probabilistic model to estimate pedestrians’ inclination toward the spatial scope of their intended goals. This module assesses the similarity between the current scene and historical experiences, thereby optimizing the utilization of time-fragmented information. Finally, a dynamic optimization module is developed, which superimposes intent point probabilities and applies a Bayesian-based density estimation method to ensure that the predicted outcomes closely align with real-world behaviors. Experimental evaluations on the Stanford drone drones (SDD), ETH-UCY, and ApolloScape datasets demonstrate that the proposed DPIU framework outperforms existing methods in predicting future trajectories and optimizing multimodal forecasting outcomes. The method substantially improves predictive performance in dynamic scenarios, providing a valuable tool for autonomous driving applications.

Age-related macular degeneration (AMD) is a chronic, progressively degenerative disorder of the macular. Wet AMD is responsible for 80 to 90 percent of all AMD-related blindness. Nowadays, ophthalmologists can diagnose wet AMD based on subretinal fluid and hemorrhage in ocular fundus images. However, this diagnosis process is time-consuming and influenced by subjective factors. Deep learning methods have achieved significant improvements for segmentation tasks in the retinal image, but only a few studies have focused on the subretinal fluid and hemorrhage lesion segmentation of wet AMD. This paper proposed AMD-Net, a novel U-Net architecture, which can segment the subretinal fluid and hemorrhage lesions of the wet AMD in ocular fundus images. The proposed AMD-Net consists of three components: encoder feature fusion unit (EFFU), skip connection block (SKB), and decoder attention block (DAB). The AMD-Net adopts the EFFU to extract and fuse the multi-scale features. And with the attention module, the EFFU can assign a higher weight for discriminative features. To reduce the semantic gap between the encoder features and decoder features, the AMD-Net designs an SKB. Furthermore, the AMD-Net introduces a DAB, a new module based on the UNet 3+ decoder, designed to leverage the local context information and enhance the contribution of high-level semantic features in tiny regions segmentation. We evaluate the proposed method on a private dataset collected by ourselves. Our model achieves 67.18% subretinal fluid Dice, 66.51% hemorrhage Dice, and 77.07% average Dice on this dataset. The experimental results indicate that our proposed AMD-Net is superior to the state-of-the-art deep learning methods.

To improve the efficiency of detecting abnormal traffic incidents on the road network and reduce the false alarm rate, a real-time traffic anomaly detection framework based on a graph spatiotemporal pattern learning (GSTPL) network is proposed. In this framework, a traffic pattern search algorithm based on a fluctuation similarity measure is designed to screen traffic flow data with the same traffic pattern, and a traffic pattern graph tuple is constructed as the input of the network model to avoid the sample imbalance problem and the effect of single-sample randomness for traffic pattern learning. Then the GSTPL network is designed to extract, unsupervised, the traffic spatiotemporal pattern features and make a reasonable prediction of future traffic parameters as the basis for anomaly evaluation. To further restrain the effect of random fluctuations in traffic flow parameters, an abnormal state evaluation method is designed to calculate the anomaly state likelihood by prediction error distribution learning. The overall detection framework realizes stable prediction of network key node traffic parameters by using spatiotemporal pattern features to construct the traffic pattern graph tuple, and gives incident evaluation results in real time by combination with the detection data. The experiment uses I90 and I405 highway traffic data in Seattle, WA, from 2015. Through comparative analysis, the proposed incident detection method based on GSTPL has a higher detection rate and lower false alarm rate, can adaptively learn dynamic changes of the traffic pattern, and has strong adaptability and stability to different traffic environments.

The unreasonable layout of taxi stands (TS) in urban areas not only fails to provide bidirectional guidance for drivers and passengers but also wastes spatial resources and aggravates the surrounding traffic. This paper compares the performance of three classical location models in optimizing TS spatial layout, and develops an extended model integrating the p-median and distance factor to support TS site selection in urban planning from multiple perspectives. To this end, taxi demand with spatial–temporal dynamics is extracted from taxi global positioning system (GPS) data to uncover the restrictive distribution characteristics of the setting areas and specific locations of TS with GIS platform. Taxi demand is then subdivided, and potential service points are set up on the road network. With the constraints of the supply and demand environment, we design the TS location models (TSLM) based on the set covering problem (SCP), the maximal covering location problem (MCLP), and the p-median problem (PMP), respectively. Furthermore, the TSLM based on PMP is extended to consider the maximum acceptable distance for passengers. A genetic algorithm-based procedure is introduced for solving the extended TSLM. An experiment conducted in China compares the facility coverage capacity, taxi demand allocation, and passenger access willingness of the optimal layout schemes obtained from four TSLMs. The number of parking spaces at TS is also evaluated. The result demonstrates that extended TSLM outperforms the other three models in the validity of locating TS.

目的 ：探讨实施全身振动训练对脑卒中偏瘫下肢肌张力、步行功能的影响。方法 ：纳入 80 例脑卒中偏瘫患者随机分成两组，其中 40 例对照组患者给予常规康复训练干预，40 例观察组患者则实施康复训练 + 全身振动训练，对两组经 2 周干预的下肢肌张力及步行功能变化进行评价。结果 ：干预后两组伸髋肌群肌张力与伸膝肌群肌张力评分均降低，但观察组显著低于对照组（P ＜ 0.05）；干预后两组 10MWT 时间均缩短，观察组短于对照组（P ＜ 0.05）。结论 ：对脑卒中偏瘫患者，全身振动训练可明显改善患者下肢肌张力，提高步行功能。

目的 探究优质护理服务模式对子宫脱垂患者护理的影响效果.方法 选择2020年10月到2021年3月我院收治的100例子宫脱垂患者,随机分类为观察组和对照组,每组50例.对照组患者采用常规护理方式,观察组采用优质护理模式,比较两组患者的生活质量以及护理满意度情况.结果 两组生活质量评分比较,观察组患者的物质生活、身体功能、社会功能、心理功能评分均高于对照组,差异具有统计学意义(P<0.05);两组患者护理满意度情况比较观察组患者在满意程度上高于对照组.结论 运用优质护理服务模式在子宫脱垂患者治疗中的应用效果更为显著,在提升患者生活质量和满意度方面效果甚佳,具有极大的临床意义.

After the release of the NVIDIA RTX platform, ray tracing methods are available in real-time rendering and improves the render quality. Therefore, realistic results can be produced in real time by combining ray tracing with rasterization techniques. However, rendering transparent objects with roughness and caustics in real time is still a challenge. We focus on these points and propose a model to render transparent objects based on real-time ray tracing. The model consists of two parts: Iterative Ray Tracing Transparency (IRTT) and Light Half Path Caustics (LHPC). IRTT approximates accurate reflections and refractions by tracing rays, computes volumetric absorption using the length that a ray propagates inside a transparent object, and simulates the rough transparency by bending the normal vector. Meanwhile, LHPC approximates caustics with a method named Local Light Tracing (LLT), where the main task is finding the point of incidence by tracing rays in the direction of the light source. After, a denoiser assisted by a distribution method is applied to suppress the noise. In the experiments, we design a combined pipeline and compare it with our model in terms of performance and render quality. In addition, our results are also compared with rasterization techniques. The experimental results reveal that our model improves the quality of the rendered image and offers a playable framerate.& nbsp; (c) 2021 Elsevier Ltd. All rights reserved.

目的 分析研究预见性护理在难产护理中的临床效果.方法 选择2020年6月到2021年3月我院收治的80例难产患者,随机分类为观察组和对照组,每组40例.对照组患者采取常规护理方法,观察组患者采取预见性护理方法,比较两组患者新生儿并发症发病情况.结果 两组并发症发病情况比较,观察组患者窒息、臂丛神经损伤、锁骨骨折、吸入性肺炎的发生率均低于对照组,有差异(P<0.05).结论 运用预见性护理手法对难产患者进行护理,可以有效降低患者以及新生儿并发症情况,具有十分显著的临床价值.

In the field of traffic safety, the occurrence of accidents remains a cause of concern for road regulators as well as users. Exploring risk factors inducing the accidents and quantifying the accident risk will not only benefit the prevention and control of traffic accidents but also assist in developing effective risk propagation model for road accidents. This study uses detailed accident record data to mine the risk factors affecting the occurrence of accidents, and quantify the accident risk under the combination of risk factors. First, by reviewing relevant literature and analyzing historical accident, we construct a multi-dimension characterization framework of risk factors with bi-level structure. The Human Factors Analysis and Classification System (HFACS) is applied to supplement and improve the framework. Next, under this framework, we identify the risk factors in traffic accident record, and analyze the statistical characteristics from the level of risk sources and risk characteristics. Then, the concept of accident liability weight is proposed to measure the impact of risk factors on accident occurrence. Through the liability affirmation of risk factors, the accident probability are updated. Last, we establish an accident risk quantify model (ARQM) based on the mean mutual information to compare the likelihood of accidents in different scenarios. In addition, we compare the accident probability and risk under equivalent liability and liability affirmation, as well as give some fundamental ideas regarding how to effectively prevent accidents.

Diabetic retinopathy, a complication of diabetes, is a significant cause of vision loss and blindness. Early detection of diabetic retinopathy can help reduce the risk of blinding. However, automatic diabetic retinopathy identification is a challenging task due to their different morphology during a different stage. Aiming at the problem of low efficiency for most of the existing methods, we developed a diabetic retinopathy recognition system based on transfer learning. The system utilizes transfer learning, which trains a neural network based on the DenseNet201 network model and Messidor Data Set, and can not only train an active network quickly, but also has a reasonable effect on the classification of diabetic retinopathy.

A large number of photos are taken for each athlete during a marathon competition, therefore, how to classify photos of specific athletes accurately and effectively has become the focus of attention. In this paper, we propose a compound deep neural network for marathon athletes number recognition to make classification more efficient and accurate. The proposed model is divided into three modules: image preprocessing module, text detection module, and text recognition module. Firstly, in the preprocessing module, we make use of the You Only Look Once version 3, and set the detection threshold and similarity threshold to reduce unnecessary detection. Secondly, we combine the efficient text detector Connectionist Text Proposal Network and the excellent text recognition general framework Convolutional Recurrent Neural Network (CRNN) to recognize the athletes number plates. Besides, to improve the accuracy of detection, we use transfer learning to fine-tune the CRNN. Finally, we design an effective tree filtering algorithm to avoid the interference caused by the text detection module. It can filter out invalid results, thereby improving the accuracy of the model. Our model is capable of performing classification on photos of marathon athletes with high precision. The model is feasible and effective, as indicated by the experiment results.

Wernicke脑病（Wernicke’s encephalopathy,WE）,是由硫胺素（维生素B1）缺乏所致的一种神经系统代谢性脑病。我们报道2例临床表现典型、经影像学检查确认的WE病例,以增加对本病的认识。1临床资料1.1患者1,男性,60岁,因胃术后1个月,

This paper presents an optophone which aims to make the sight-disabled people have the same reading view as normal people. The complete set of equipment includes a Raspberry Pi, a camera and an Android App. We connected the CSI camera to the raspberry PI interface to complete the image acquisition of physical paper reading materials and connected the playback device to the Raspberry Pi audio interface to provide users with reading services. The Android app communicates with the Raspberry Pi to control the Raspberry Pi's actions, including recognizing images to generate text, selecting the appropriate reading file, and choosing to play or pause of file reading. Unlike such devices developed in the past, the service is made easier by the fact that visually impaired people can operate this optophone directly from their phones. The technical overview of each subsystem of the optophone is presented and discussed.

The Dynamic Traffic Assignment (DTA) is one of the important measures to alleviate urban network traffic congestion. The congestions are usually caused by stochastic traffic demands, which are generally unassignable from time dimension in the real-world but are assumed to be assignable in existing DTA methods (i.e. real-time travel demands). In this paper, a distributed DTA method for preventing urban network traffic congestion caused by stochastic real-time travel demands by improving Multi-Agent Reinforcement Learning (MARL). A team structure, which consists of decision-makers and advisers, is designed to learn parallelly in realistic DTA tasks. To reduce the size of the solution space adaptively, the dynamic critical values advised by adviser agents are adopted as constraints for the strategy space of decision-makers (i.e. main agents). A collaborative heterogeneous-adviser mechanism is designed to avoid deviation of guidance. To enhance the adaptability of DTA to the changeable external environment, the mixed strategy concept is introduced to improve the decision-making process of main agents. The respective mapping mechanisms are designed to define adaptive learning rates to improve the sensitivity of MARL. The Sioux Falls (SF) network is established as a test platform via a Dynamic Network Loading (DNL). The effectiveness of the suggested DTA method is assessed through numerical simulations SF network. Under the influence of the scenario with stochastic real-time travel demands, the results show that the proposed method outperforms in terms of the throughput of the network and the individual average travel time among the overall network. Additionally, the ability of the proposed method in response to the external environment rapidly has also been demonstrated. Adopting the suggested method can improve the state of the art to assign stochastic real-time travel demands dynamically and to avoid potential traffic congestion fundamentally.

Urban network traffic congestion can be caused by disturbances, such as fluctuation and disequilibrium of traffic demand. This paper designs a distributed control method for preventing disturbance-based urban network traffic congestion by integrating Multi-Agent Reinforcement Learning (MARL) and regional Mixed Strategy Nash-Equilibrium (MSNE). To enhance the disturbance-rejection performance of Urban Network Traffic Control (UNTC), a regional MSNE concept is integrated, which models the competitive relationship between each agent and its neighboring agents in order to improve the decision-making process of MARL. The learning rate is enhanced with a self-adaptive ability to avoid a local optimal dilemma; Jensen-Shannon (JS) divergence is utilized to define the learning rate of the modified MARL. A two-way rectangular grid network with nine intersections is modeled via a Cell Transmission Model (CTM). A probability distribution mechanism, which can update the turn ratio of each approach dynamically and discretely, is established to represent the segmented route-decision process of the vehicles. The effectiveness of the proposed control method is evaluated through simulations in the grid network. The results show the influence of major disturbances, such as fluctuation of vehicle arrival rate, fluctuation of traffic demand (e.g. a rapidly rising flow and extreme changes in origin-destination distribution), and disequilibrium of traffic demand (e.g. different arrival flows at each boundary of the urban network), on the performance of the suggested control method. The results can be used to improve the state of the art in order to reduce urban network traffic congestion due to these disturbances.

Automatic retinal vessel segmentation has drawn significant attention in early diagnosis and treatment of many diseases, such as diabetes, retinal diseases, and coronary heart disease. However, due to vessels exhibit variations in morphology and low contrast, it is still challenging to obtain accurate segmentation results. In this paper, aiming at upgrading the accuracy and sensitivity of existing vessel segmentation methods, we propose a Multi-Scale Convolutional Neural Network with Attention Mechanisms (MSCNN-AM). For extraction of blood vessels at different scales, we introduce atrous separable convolutions with varying dilation rates, which could capture global and multi-scale vessel information better. Meanwhile, in order to reduce false-positive predictions for tiny vessel pixels, we also adopt attention mechanisms so that the proposed MSCNN-AM can pay more attention to retinal vessel pixels instead of background pixels. Because the green channel shows better vessel contrast and less noise than other channels in the RGB image, our proposed MSCNN-AM is trained and tested with green channel images only, excluding extra pre-processing and post-processing steps. The proposed method is evaluated on three public datasets, including DRIVE, STARE, and CHASE_DB1. In addition, we adopt six objective metrics to verify the performance of the MSCNN-AM, including sensitivity (Se), specificity (Sp), accuracy (Acc), F1-score, an area under a receiver operating characteristic curve (AUC-ROC), and an area under precision/recall curve (AUC-PR). Experimental results indicate that our proposed method outperforms most of the existing methods with a sensitivity of 0.8342/0.8412/0.8132 and an accuracy of 0.9555/0.9658/0.9644 on DRIVE, STARE, and CHASE_DB1 separately.

A taxi stand can effectively regulate the behavior of taxi picking up passengers, reduce empty-run rate, and provide a convenient and orderly waiting environment for the public. However, the unreasonable setting of the existing taxi stands in most cities leads to an extremely low utilization rate and a waste of public space resources. This paper presents a novel three-stage strategy to address the taxi stands location problem (TSLP) incrementally. First, taxi demands hotspots are mined from a massive taxi Global Positioning System (GPS) data with GIS platform, and the optimal area for taxi stands siting in the following stages is determined. Then, the spatial interaction between taxi demands and taxi stands is explored to generate demand subsections and stand candidates along both the sides of the road. At last, a taxi stand location model (TSLM) is developed to minimize the total cost, which contains the access cost of passengers and the construction cost of taxi stands. The genetic algorithm-based procedure is adopted for TSLM optimization. A case study conducted in China verifies the effectiveness of the location strategy and investigate the impact of the maximum acceptable distance for passengers on TSLP. The experimental results describe the number and layout of taxi stand under a different demand coverage, which indicates that the proposed approach is beneficial to provide scientific reference for the municipal department in taxi stand site decisions and make a tradeoff between the interests of planners and users.

In this paper, an Augmented Reality Navigation Application based on iOS platform is designed. This application is composed of five modules: the Location Module, the Digital Map Module, the Approximate Module, the Estimate Module and the AR Module. The Location Module requests current location via GPS in real-time and obtains the details of destination from map server. The Digital Map Module provides a Satellite Imagery View which superimposes navigation routes with marks such as current location, starting and destination. The Approximate module produced an approximation to fix the bias when converting the geographical coordinates between WGS-84 and GCJ-02 system. The Estimate Module produced a 3D position estimation according to a geographical coordinate provided by GPS or map server. This estimation will be used for placing the marks and routes in AR view. The AR Module provides an Augmented Reality scene superimposed destination and the route on a real-world background with a real-world scale.

In augmented reality, smart devices need to sense their own position in real physical space and complex scene structure to achieve good virtual reality interaction and three-dimensional registration. This paper proposes a technical framework of visual SLAM, which uses VO and BA methods based on feature points to obtain pose, and applies it to the augmented reality system to realize pose estimation, registration, tracking, collision and other effects of intelligent equipment independent of artificial or natural landmarks.

In our daily life, many face applications need to complete three tasks: face detection, facial landmark localization and head pose estimation. Currently, most methods accomplish these three tasks separately. Multi-task cascade convolution neural network(MTCNN) adpots the idea of casecading that combines face detection and face alignment. Inspired by MTCNN, we combine the three tasks of face detection, head pose estimation and key points detection under a cascade framework. Simultaneously, we increased the number of key points detected by MTCNN from 5 to 21. By training and testing our model on the WIDER and Umdfaces datasets, we explored the inherent correlation between these three facial tasks and demonstrated the excellent results of the model tested in an unconstrained environment.