A roadside sensing unit can provide over-the-horizon perception information for autonomous vehicles due to its high perspective. However, numerous challenges need be overcome such as the missing detection of small objects and occluded objects. To this end, study proposed a fixed perspective (FPP) module, which considered background subtraction camera object detection. The FPP module was divided into two parts: grayscale (GBS) submodule background–current image fusion (BCF) submodule....

Abstract To improve the safety and efficiency of train operation, autonomous driving have developed rapidly in recent years. Among them, signal detection is one most basic functions. However, due to small size light complicated railway environment, still a huge problem. The existing methods, such as approach based on Hough circle transformation, are hard meet practical application requirements. In this paper, real time lights Yolov5 introduced. And lot experiments were conducted prove...

Abstract Virtual reality (VR) displays aim to create highly immersive virtual environments based on the principle of binocular disparity, which reproduces spatial information scenes through fusion processing disparity by visual system. However, due differences between VR and real‐world scenes, challenge rendering in a manner that aligns with users' natural depth perception principles has not been fully addressed. In this paper, image distances (VIDs) RGB channels head‐mounted display (HMD)...

The rapid adoption of immersive technologies in educational contexts has heightened research interest analyzing the specific behavioral patterns learners within learning environments. However, existing on technical affordances and pedagogical potential analysis remains fragmented. This study contributes by developing a sustainable conceptual framework that amalgamates requirements, specification, evaluation, iteration into an integrated model to identify benefits hurdles A systematic review...

Rail-track detection is a core function for automated rail transit perception. However, existing methods cannot effectively detect the rail-tracks in complex environment, especially turnout scenarios. In this study, we propose topology guided method to which includes following four parts: Firstly, neural network used obtain pixels of rail-lanes, and geometric relationship between rail-lanes mined by inverse perspective transformation. Secondly, rail-lanes' are converted key points...

At present, drones are rapidly developing in the aviation industry and applied to all aspects of life. However, letting autonomously avoid obstacles is still focus research by scholars at this stage. current automation mostly based on human experience determine obstacle avoidance strategy UAV. And method only rely machine very few. In paper, UAV collect visual distance sensor information make autonomous decision through deep reinforcement learning algorithm, algorithm tested v-rep simulation...

Rail track segmentation is key to environmental perception of autonomous train. However, due the complexity railway environment, critical issues such as detection rail tracks with different curvatures remain be overcome. In this study, a novel architecture called FarNet proposed for long-range point cloud segmentation. The mainly divided into three parts, i.e., spherical projection, attention-aggregation network and results refinement. Specifically, projection converts LiDAR pseudo range...

Roadside light detection and ranging (LiDAR) is commonly used to record the traffic data of whole intersection scene or road segment in intelligent transportation systems (ITSs). However, general deep-learning object methods do not adequately consider static background captured by roadside LiDAR. Moreover, critical issues remain be overcome using LiDAR: false alarms caused complex interference multiscale objects with limited characteristics. To this end, a feature enhancement cascade network...

Rail-track detection is a crucial function for an active obstacle avoidance system in trains. However, existing methods face challenges effectively detecting rail-tracks, particularly turnout scenarios. This study introduces novel rail-track approach using key-point estimate network. The network treats the as pair and constructs dedicated model detection. Additionally, pseudo-attention mechanism leverages output from previous stages, enabling to focus on region. Also, dislocation assignment...

The development of 6G has led to the need for communication systems realize low latency, high throughput and stable connectivity. To achieve these goals, Reconfigurable Intelligent Surface emerged. However, RIS with massive elements, overhead caused by channel estimation feedback is not negligible. In this article, we address problem using irregular RIS, which essentially involves irregularly rationing a specified number reflective elements on an surface, providing additional spatial degrees...

Depth completion of forward objects is critical to safe and effective autonomous driving as it can estimate dense-depth from sparse LiDAR (i.e., Light Detection Ranging) data RGB images. The study proposes a self-supervised depth method address the deficiency in sensing images caused by uneven distribution point-clouds generated long-range LiDAR. proposed uses stacked-hourglass-network structure backbone, which allows perform multi-level deep fusion extract features point-clouds. In...

With the extending use of LiDAR SLAM in various areas, interference external disturbances on is becoming more and obvious. Huge efforts have been made to reduce drift error using graph-based methods. However, mapping results can be severely affected by under extreme conditions, which will limit performance This study proposes a new strategy static local scale identifying compensating influence based localization SLAM. Contrast experiments were first designed performed analyze potential...

In the intelligent unmanned systems, aerial vehicle (UAV) obstacle avoidance technology is core and primary condition. Traditional algorithms are not suitable for in complex changeable environments based on limited sensors UAVs. this article, we use an end-to-end deep reinforcement learning (DRL) algorithm to achieve UAV autonomously avoid obstacles. For problem of slow convergence DRL, a Multi-Branch (MB) network structure proposed ensure that can get good performance early stage;...

Object detection plays an important role in underground intelligent vehicles and transportation systems. Due to the uneven light mining scenarios, infrared cameras are one of typical onboard sensors for environmental perception. Although object has been studied decades, it still confronts challenge detecting objects mines. The contributing factors include weak small images similar environments scenarios. In this paper, a Feature Enhancement Guided Network (FEGNet) is proposed address these...

The integration of cameras and LiDAR sensors has emerged as a promising approach to enhance environmental perception 3D reconstruction capabilities in autonomous vehicles robotic systems. Precise extrinsic calibration is paramount importance achieve effective multi-sensor fusion applications. Traditional methods often rely on manual procedures specific targets, which can be time-consuming prone errors. In contrast, Convolutional Neural Networks (CNNs) have shown potential devising end-to-end...

Virtual Reality (VR) offers a valuable platform for real-life skills training. However, previous research has indicated that human's perception of depth in VR differs from the real world. Such perceptual conflicts can impact immersion and learning skills, thus attracting widespread attention. Various methods have been proposed to enhance users' perception, yet underlying mechanisms still require further research. In this paper, we used Error-Related Potentials (ErrPs) electroencephalography...