• Laser & Optoelectronics Progress
  • Vol. 62, Issue 2, 0237005 (2025)
Ruoying Liu1,2,3,*, Miaohua Huang1,2,3, Liangzi Wang1,2,3, Yongkang Hu1,2,3, and Ye Tao1,2,3
Author Affiliations
  • 1Hubei Key Laboratory of Advanced Technology for Automotive Components, Wuhan University of Technology, Wuhan 430070, Hubei , China
  • 2Hubei Collaborative Innovation Center for Automotive Components Technology, Wuhan University of Technology, Wuhan 430070, Hubei , China
  • 3Hubei Research Center for New Energy & Intelligent Connected Vehicle, Wuhan University of Technology, Wuhan 430070, Hubei , China
  • show less
    DOI: 10.3788/LOP241187 Cite this Article Set citation alerts
    Ruoying Liu, Miaohua Huang, Liangzi Wang, Yongkang Hu, Ye Tao. Roadside Object Detection Algorithm Based on Multiscale Sequence Fusion[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0237005 Copy Citation Text show less
    Architecture diagram of MQ-YOLO algorithm
    Fig. 1. Architecture diagram of MQ-YOLO algorithm
    Architecture diagram of D_C2f module
    Fig. 2. Architecture diagram of D_C2f module
    Architecture diagram of triple feature encoder module and scale sequence feature fusion module
    Fig. 3. Architecture diagram of triple feature encoder module and scale sequence feature fusion module
    Architecture diagram of DyDetect module
    Fig. 4. Architecture diagram of DyDetect module
    Visualization results of DAIR-V2X-I dataset. (a) Category and quantity of labels; (b) distribution of center position of detection target; (c) distribution of width and height of label
    Fig. 5. Visualization results of DAIR-V2X-I dataset. (a) Category and quantity of labels; (b) distribution of center position of detection target; (c) distribution of width and height of label
    Object detection results of DAIR-V2X-I dataset. (a) Dense occlusion scene; (b) small target scene; (c) night scene
    Fig. 6. Object detection results of DAIR-V2X-I dataset. (a) Dense occlusion scene; (b) small target scene; (c) night scene
    Object detection results of the DAIR-V2X-SPD-I dataset. (a) Dense occlusion scene; (b) small target scene; (c) rainy scene
    Fig. 7. Object detection results of the DAIR-V2X-SPD-I dataset. (a) Dense occlusion scene; (b) small target scene; (c) rainy scene
    Roadside object detection system. (a) System deployment location; (b) system architecture; (c) visual interface
    Fig. 8. Roadside object detection system. (a) System deployment location; (b) system architecture; (c) visual interface
    Experimental results of roadside object detection system. (a) Sunny scene; (b) rainy scene; (c) night scene
    Fig. 9. Experimental results of roadside object detection system. (a) Sunny scene; (b) rainy scene; (c) night scene
    MethodTypeParameter /MGFLOPsmAP@50 /%mAP@(50‒95) /%FPS /(frame/s)
    1YOLOv83.008.1086.764.6150.8
    2YOLOv8+D-C2f3.008.1086.864.8157.7
    3YOLOv8+PSF2.8617.0088.967.1121.3
    4YOLOv8+DyDetect3.6210.4088.066.994.7
    5YOLOv8+NWD3.018.2087.064.7142.2
    6YOLOv8+D-C2f+PSF3.4815.3088.867.3110.2
    7YOLOv8+D-C2f+PSF+DyDetect3.1719.4090.570.369.6
    8

    YOLOv8+

    D-C2f+PSF+DyDetect+NWD

    3.9622.2090.670.669.2
    Table 1. Ablation experiments on DAIR-V2X-I dataset
    MethodParameter /MGFLOPsmAP@50 /%mAP@(50‒95) /%FPS/(frame/s)
    SSD26.3062.7068.839.634.4
    YOLOXs54.20156.0082.349.8101.0
    YOLOv5n2.517.1086.264.0139.2
    YOLOv5s9.1123.8089.468.7133.3
    YOLOv7-Tiny6.0213.2087.760.5117.6
    YOLOv8n3.008.1086.764.6150.8
    YOLOv8s11.1328.4089.369.2146.2
    YOLOv9-Tiny4.2518.6088.267.341.1
    MQ-YOLO3.9622.2090.670.669.2
    Table 2. Comparison of detection results of different algorithms on DAIR-V2X-I dataset
    MethodmAP@50 /%Mean value /%
    CarPedestrianCyclistBusVanTruck
    YOLOv8n97.477.891.799.498.298.593.8
    MQ-YOLO98.792.795.899.499.199.497.5
    Table 3. Comparison of detection results of different algorithms on DAIR-V2X-SPD-I dataset
    Ruoying Liu, Miaohua Huang, Liangzi Wang, Yongkang Hu, Ye Tao. Roadside Object Detection Algorithm Based on Multiscale Sequence Fusion[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0237005
    Download Citation