A dual motion-model tracker that explicitly accounts for non-linear perspective transformations during vehicle approach is introduced, substantially improving temporal consistency over linear motion assumptions, and a semantic attribute classification pipeline that estimates occlusion level, readability, sign embeddedness, and road relevance is developed, providing actionable context to downstream planning.
Abstract
Reliable traffic sign detection is a prerequisite for the global deployment of autonomous driving systems, where regulatory compliance and road safety depend on perceiving signs correctly across regions, ranges, and weather conditions. Despite recent progress, vision-based methods continue to face three fundamental limitations: poor cross-regional generalization due to high diversity across countries, degraded performance on small-object detection at long ranges (traffic signs occupy as little as $10{\times}10$ pixels at 200m), and fragile temporal tracking under the strongly non-linear perspective distortion that occurs as a vehicle approaches a sign. In this paper, we address the problem of robust, long-range, region-agnostic traffic sign perception by combining camera and Light Detection and Ranging (LiDAR) sensing. We present a multi-modal detection framework whose Intensity-Aware Deformable Fusion module aligns retro-reflective LiDAR cues with camera features, anchoring detection on geometric invariants rather than region-specific visual appearance. We further introduce a dual motion-model tracker that explicitly accounts for non-linear perspective transformations during vehicle approach, substantially improving temporal consistency over linear motion assumptions. Additionally, we develop a semantic attribute classification pipeline that estimates occlusion level, readability, sign embeddedness, and road relevance, providing actionable context to downstream planning. Extensive evaluation on our dataset, spanning 60+ countries and 2,500+ hours of driving data, shows that the proposed pipeline achieves an Object Miss Ratio (OMR) of 0.49% across 221,068 evaluation sequences, demonstrating globally generalizable traffic sign perception in commercial-grade autonomous driving systems.
Object detection and tracking are fundamental components of perception systems for autonomous driving. Achieving robust performance under adverse conditions such as limited visibility, sensor noise, and failures remains an open challenge, particularly in autonomous racing, where vehicles operate at very high speeds, experience strong vibrations, and interact under small safety margins. This paper presents a multi-modal late-fusion perception pipeline for object detection and tracking in the autonomous racing domain. The proposed system extends previous work by exploiting all onboard sensors through a late-fusion approach and a dedicated multi-object tracking framework. Independent detections from cameras, LiDARs, and RADARs are combined to provide timely and robust state estimates of surrounding vehicles. The tracking method explicitly compensates for detection delays and embeds in its model prior knowledge of vehicle dynamics and track layout. Experimental evaluation on real-world data across diverse critical scenarios, representative of challenging edge cases also in urban driving, confirms the effectiveness of the proposed pipeline and its suitability to support safe and adaptive planning decisions.
Davide Malvezzi, Michele Pestarino, Vittoria Cavicchioli et al.· 0 citations
Urban traffic areas, such as intersections, significantly escalate the risk of accidents due to the high density and diverse behaviour of road users. Achieving higher automation levels necessitates robust object recognition and tracking capabilities to mitigate accident risk. Modern vehicle architectures utilize multiple sensor systems to ensure robustness and high-quality perception. For camera systems in particular, multi-camera multi-target tracking (MCMTT) with visual re-identification (ReID) is fundamental for understanding object behaviour across different fields of view. This work directly addresses this need by presenting and evaluating a deep learning-based concept for near-field MCMTT using Waymo’s real-world perception dataset [1]. This work enables comprehensive and reliable tracking, focusing on the investigation of concept parameters and highlighting the essential factors for the automated driving task.
Marius Westendorf, Jonas Brinkmann, Marcel Kascha et al.· Automotive and Engine Techno...· 0 citations
Accurate traffic scene recognition serves as a critical foundation for decision-making and safe driving in autonomous driving and intelligent transportation systems. Existing methods mostly rely on single visual data vulnerable to environmental variations, or vision-LiDAR fusion schemes with insufficient capacity to represent road topology and traffic semantic information, limiting recognition accuracy and robustness. To address these limitations, this paper proposes an end-to-end traffic scene recognition network based on the fusion of monocular camera images and corresponding road map top-down view data. We design a learnable cross-view spatial alignment module to eliminate perspective discrepancy, and a bidirectional cross-attention fusion module to enable deep bidirectional interaction between visual semantic and map topology features. Experiments on a self-built dataset covering five typical traffic scenes show that the proposed method achieves an overall recognition accuracy of 92.6%, outperforming the best single-input baseline Swin-Tiny by 3.3%. Ablation studies further validate the effectiveness of each core module.
ZhenYu Cheng, Haoyu Kon· International Conference on...· 0 citations
Traffic sign detection is an integral part of Advanced Driver Assistance Systems (ADAS) and self-driving cars wherein correct and timely detection helps ensure a safer driving experience. However, currently available models are trained on benchmark datasets that pertain to European or Chinese traffic scenarios. Such models lack the efficiency required for traffic scenarios in India due to multiple factors such as the variety of sign shapes and languages present, occlusions, varied light intensities and prevalence of smaller objects. In this paper, a region-based traffic sign detection system has been proposed using the enhanced YOLOv8 framework for Indian traffic scenarios. A subset of the Mapillary Traffic Sign Dataset (MTSD) has been curated, annotated and data-augmented with mosaic augmentation, change in brightness, adaptive scaling, and high-resolution training to overcome challenges posed by occlusions, varying light intensity and distant traffic signs.
L. M, S. N., Sohan Js et al.· 2026 7th International Confe...· 0 citations
Traffic sign detection is a critical perception component in autonomous driving, yet it remains highly challenging due to occlusions caused by leaves, vehicles, and buildings. These visual obstructions can lead to catastrophic decision-making errors in autonomous vehicles, directly threatening passenger safety. To support reliable autonomous navigation under occlusion, this paper proposes ODE-Net. It is an advanced YOLOv11-based detection framework with three key contributions. First, the hierarchical multi-scale feature extraction module designed within the backbone network enhances the representation of partially visible features under various obstructions. Second, the high-low frequency fusion neck module introduced at the neck is incorporated to strengthen semantic information, enabling the capture of edge details and overall contours at occlusion boundaries. Finally, an occlusion ratio prediction branch integrated into the detection head, along with the improved area-aware regression loss, is proposed to optimize bounding box regression. We also construct the Occluded Traffic Sign Dataset, which contains 8699 images with precise occlusion-percentage annotations. Compared with recent detectors including YOLOv11, YOLOv8, HYPER-YOLO, and RT-DETR-l, experiments on CCTSDB and OTSD show that ODE-Net achieves an mAP50 of 80.6% on CCTSDB and an mAP50-95 of 53.5%, improving over YOLOv11 by 3.3 and 3.4 percentage points, respectively. On OTSD, ODE-Net reaches the best mAP50-95 of 64.5%, surpassing YOLOv11, RT-DETR-l, and HYPER-YOLO by 2.2, 2.5, and 1.6 percentage points, respectively. The model achieves 94.4% occlusion-prediction accuracy while maintaining a compact size of 2.58M parameters and a low computational cost of 6.7 GFLOPs, thereby achieving a superior accuracy-efficiency balance.
Jiacai Liao, Le Luo, Lin Hu et al.· Proceedings of the Instituti...· 0 citations
Bus-mounted vision sensing provides a practical and complementary perspective for intelligent transportation systems, but reliable traffic object detection from bus front-view cameras remains challenging because elevated viewpoints induce severe scale skewness, dense interactions around bus stops and intersections, and frequent heterogeneous occlusion. To support this sensing scenario while avoiding ambiguity with previously used dataset acronyms, we construct the Bus Front-view Traffic Dataset (BFTD), a high-resolution benchmark collected from forward-facing cameras mounted on multiple buses operating on urban routes during real-world service. The BFTD contains 8131 images and 56,137 annotated instances across five traffic-participant categories, covering dense pedestrians, mixed-traffic flow, illumination variation, rain, fog, and occlusion-prone scenes. Based on the visual characteristics of bus-mounted cameras, we propose YOLO-M2LA, a local–global attention detection framework in which CBS-SPD preserves fine-grained information during early downsampling and M2LA couples multi-scale local context modeling with efficient global dependency aggregation. Extensive experiments on BFTD and public benchmarks show that the proposed framework improves detection accuracy, particularly for small and visually crowded traffic participants, while maintaining a practical accuracy–efficiency trade-off. Dataset statistics, condition-specific evaluation, ablation analysis, and qualitative visualization further support the effectiveness of BFTD and YOLO-M2LA for vision-based traffic sensing. The dataset and implementation are publicly available online.
Wenjing Gao, N. Zou· Italian National Conference...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
