A5062124960- Robotics and Sensor-Based Localization
- Robotic Path Planning Algorithms
- Autonomous Vehicle Technology and Safety
- Advanced Vision and Imaging
- Video Surveillance and Tracking Methods
- Advanced Image and Video Retrieval Techniques
- Control and Dynamics of Mobile Robots
- Traffic control and management
- Remote Sensing and LiDAR Applications
- Robotic Locomotion and Control
- Indoor and Outdoor Localization Technologies
- Advanced Neural Network Applications
- Vehicle Dynamics and Control Systems
- 3D Surveying and Cultural Heritage
- Image and Video Stabilization
- Inertial Sensor and Navigation
- Generative Adversarial Networks and Image Synthesis
- Underwater Vehicles and Communication Systems
- Infrared Target Detection Methodologies
- Advanced Image Processing Techniques
- Optical measurement and interference techniques
- Advanced Measurement and Detection Methods
- Smart Parking Systems Research
- Biomimetic flight and propulsion mechanisms
- Anomaly Detection Techniques and Applications
Beijing Institute of Technology
2016-2025
Wuhan University of Technology
2023-2025
Chongqing University
2024
The University of Melbourne
2024
Chongqing University of Posts and Telecommunications
2023
Guangzhou University
2023
Beijing Satellite Navigation Center
2023
Chongqing Science and Technology Commission
2023
Nanjing Forestry University
2023
Nanjing University of Science and Technology
2016-2021
This paper presents a lightweight stereo vision-based driving lane detection and classification system to achieve the ego-car's lateral positioning forward collision warning aid advanced driver assistance systems (ADAS). For detection, we design self-adaptive traffic lanes model in Hough Space with maximum likelihood angle dynamic pole region of interests (ROIs), which is robust road bumpiness, structure changing while interferential markings on ground. What's more, this can be improved...
This paper presents real-time obstacles detection and their status classification method for collision warning in the vehicle active safety system. Specifically, stereo cameras millimeter wave (mmw)-radar are fused to help driving ego-vehicle find "Danger" or "Potential Danger" a timely way through combining with kinematic model. The proposed makes full use of unique advantages mmw-radar sense environment several modules. Cameras mainly used detect near lateral dynamic objects obtain region...
Comparing to image inpainting, outpainting receives less attention due two challenges in it. The first challenge is how keep the spatial and content consistency between generated images original input. second maintain high quality results, especially for multi-step generations which regions are spatially far away from initial To solve problems, we devise some innovative modules, named Skip Horizontal Connection Recurrent Content Transfer, integrate them into our designed encoder-decoder...
Nowadays, deep learning techniques are widely used for lane detection, but application in low-light conditions remains a challenge until this day. Although multi-task and contextual-information-based methods have been proposed to solve the problem, they either require additional manual annotations or introduce extra inference overhead respectively. In paper, we propose style-transfer-based data enhancement method, which uses Generative Adversarial Networks (GANs) generate images conditions,...
Intersections are quite important and complex traffic scenarios, where the future motion of surrounding vehicles is an indispensable reference factor for decision-making or path planning autonomous vehicles. Considering that trajectory a vehicle at intersection partly obeys statistical law historical data once its driving intention determined, this paper proposes long short-term memory based (LSTM-based) framework combines prediction together. First, we build prior trajectories model (IPTM)...
Integrated navigation of unmanned ground vehicles (UGV) is significant for many advanced intelligent transportation system applications. Adaptive information fusion technique based on observability analysis has a great potential to enhance UGV integrated systems the capability high-precision positioning and navigation. In an system, tolerance against unknown time-varying observation conditions key factor satisfy specific requirements high-precision, self-adaption, high reliability. Thus,...
Deep reinforcement learning (DRL) has been successfully applied to end-to-end autonomous driving, especially in simulation environments. However, common DRL approaches used complex driving scenarios sometimes are unstable or difficult converge. This paper proposes two improve the stability of policy model training with as few manual data possible. For first approach, is combined imitation train a feature network small amount for parameters initialization. second an auxiliary added framework,...
Object Transfiguration replaces an object in image with another from a second image. For example it can perform tasks like putting exactly those eyeglasses A on the nose of person B. Usage exemplar images allows more precise specification desired modifications and improves diversity conditional generation. However, previous methods that rely feature space operations, require paired data and/or appearance models for training or disentangling objects background. In this work, we propose model...
Object Transfiguration replaces an object in image with another from a second image. For example it can perform tasks like "putting exactly those eyeglasses A on the nose of person B". Usage exemplar images allows more precise specification desired modifications and improves diversity conditional generation. However, previous methods that rely feature space operations, require paired data and/or appearance models for training or disentangling objects background. In this work, we propose...
With the rapid development of computer vision, vision-based simultaneous localization and mapping (vSLAM) plays an increasingly important role in field unmanned driving. However, traditional SLAM methods based on a monocular camera only perform well simple indoor environments or urban with obvious structural features. In off-road environments, situation that encounters could be complicated by problems such as direct sunlight, leaf occlusion, rough roads, sensor failure, sparsity stably...
Autonomous driving, including intelligent decision-making and path planning, in dynamic environments (like highway) is significantly more difficult than the navigation static scenarios because of additional time dimension. Therefore, correlating dimension space through prediction to create a spatio-temporal map can make planning such kinds environment much easier. In this article, NGSIM data analysed processed from perspective ego-vehicle (using as an ego-vehicle's perception results). Based...
Intelligent decision making and efficient trajectory planning are closely related in autonomous driving technology, especially highway environment full of dynamic interactive traffic participants. This work integrates them into a unified hierarchical framework with long-term behavior (LTBP) short-term (STDP) running two parallel threads different horizon, consequently forming closed-loop maneuver system that can react to the effectively efficiently. In LTBP, novel voxel structure 'voxel...
The efficiency of the rapidly exploring random tree (RRT) falls short when efficiently guiding targets through constricted-passage environments, presenting issues such as sluggish convergence speed and elevated path costs. To overcome these algorithmic limitations, we propose a narrow-channel path-finding algorithm (named NCB-RRT) based on Bi-RRT with addition our proposed research failure rate threshold (RFRT) concept. Firstly, three-stage search strategy is employed to generate sampling...
Point cloud registration is a fundamental task in the fields of computer vision and robotics. Recent advancements transformer-based methods have demonstrated enhanced performance this domain. However, standard attention mechanisms employed these approaches tend to incorporate numerous points low relevance, therefore struggle focus their weights on sparse yet meaningful points. This inefficiency leads limited local structure modeling capabilities quadratic computational complexity. To...
Neural Radiance Fields (NeRF) use multi-view images for 3D scene representation, demonstrating remarkable performance. As one of the primary sources images, multi-camera systems encounter challenges such as varying intrinsic parameters and frequent pose changes. Most previous NeRF-based methods assume a unique camera rarely consider scenarios. Besides, some NeRF that can optimize extrinsic still remain susceptible to suboptimal solutions when these are poor initialized. In this paper, we...
Point cloud data are often accompanied by noise and irregularities, which bring great challenges to the extraction of point surface traces discontinuous rock masses. Most existing feature line methods rely on traditional geometric or statistical techniques, less resistant noise. To address this issue, paper proposes a novel method for trajectory recognition surfaces mass clouds. The first detects extracts points using normal tensor voting theory based symmetry at different periods. Then,...