Aug 2026· International Conference on Automation, Control and Robotics Engineering· pp. 39-44· 0 citations· 11 references
Abstract
In intelligent transportation, roadside perception systems act as city’s eyes. However, monocular camera-based 3D object detection remains challenging, as inferring 3D geometry from 2D images is ill-posed, especially under occlusion, scale variation, and adverse weather. To address this, we propose MonoYOLO, featuring a dual-branch architecture for deep fusion of semantic and geometric information. The Cross-Modal Fusion (CMF) module integrates RGB semantic features with depth maps spatial cues, enhancing 3D attribute perception. The Multi-Scale Attention (MSA) module adaptively focuses on key features, improving robustness to scale variation. The Detect3D head decouples the 3D detection task into sub-tasks, ensuring accurate and stable bounding box prediction. Experiments on the challenging Rope3D dataset show that MonoYOLO achieves superior comprehensive detection accuracy while maintaining high computational efficiency, offering an effective solution for low-cost, high-performance roadside 3D perception.
This work introduces SAM-AD, a domain-specific pretraining strategy that fine-tunes SAM on autonomous-driving imagery to extract feature representations with rich semantic information, and develops the Depth-Guided Wavelet Attention (DGWA) module, which suppresses high-frequency sensor noise while preserving critical c...
Zi-Ying Song, Lin Liu, Hong-Yu Pan et al.· 0 citations
Multimodal perception integrating light detection and ranging (LiDAR) and cameras has become a key paradigm for 3-D object detection, as it leverages both geometric structure and semantic information. However, in real-world autonomous driving scenarios, calibration errors, adverse weather, and sensor degradation can in...
Hui-Lin Huang, Yan Bai, Peng-Yuan Wang et al.· IEEE Sensors Journal· 0 citations
Results confirm the effectiveness of combining transformer-based global encoding with lightweight convolutional decoding for high-quality, real-time monocular depth estimation.
A aggregated Euclidean distance weighted box fusion method, which aggregates complementary information from multiple candidate boxes during post-processing to improve bounding-box selection and localization accuracy, and a hybrid deformable half-conv (HDHC) module that jointly enhances global and local feature represen...
Di Tian, Jia-Wei Wang, Jia-Bo Li et al.· Measurement science and tech...· 0 citations
Three-dimensional object detection plays a critical role in intelligent robotics and autonomous driving, where accurate and robust perception remains challenging under multimodal fusion settings. Existing bird’s-eye-view (BEV)-based multimodal methods still suffer from unreliable camera depth estimation, insufficient a...
Jian-Qiang Su, Ji-Yuan Wang, Yong-Sheng Qi et al.· IEEE Transactions on Instrum...· 0 citations
