Currently, the main method for detecting lane lines is semantic segmentation. However, there are still some problems with this approach. Firstly, as semantic segmentation requires the extraction of feature information and the classification of all pixels in the vehicle image, lane lines usually represent only a small fraction of the pixels, resulting in a huge computational waste and making it difficult to guarantee real-time lane detection. Furthermore, with semantic segmentation, lane lines can only be distinguished from background information, and it is not possible to form different The lane line instances cannot be formed. Secondly, existing lane detection methods generally have a high false detection rate, making it difficult to ensure the safety of unmanned driving. The safety of unmanned driving is not guaranteed. Therefore, this paper proposes the detection of lane lines using an improved hourglass model to generate key points and incorporates CBAM (Convolutional Attention Mechanism) and channel shuffle modules to further improve detection accuracy and reduce the false detection rate. Firstly, the CBAM attention mechanism is embedded into the encoding-decoding structure to enhance useful features and suppress useless ones. Secondly, the ShuffleNet module is introduced to strengthen cross-channel information exchange between multi-level features at different scales, thus improving the accuracy of recognition for obstructed key points. The model is trained and verified using the TuSimple Lane Line open dataset. Experimental results show that when the model is composed of different numbers of hourglass modules, the proposed lane detection algorithm achieves accuracies of 96.34%, 96.66%, 96.71%, and 96.73%, respectively, verifying the effectiveness of the method proposed in this paper for lane detection, namely, detecting the feasible driving area of lanes.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Lane Detection with the Fusion of CBAM Attention Mechanism and Channel Shuffle Module


    Contributors:
    Cheng, Xin (author) / Sun, Zhenzhong (author) / Zhou, Jingmei (author)


    Publication date :

    2023-08-04


    Size :

    1441421 byte





    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English



    Multi-channel Avionics Module Fault Classification Based on Inception-CBAM Network

    Yu, Lubin / Peng, YiXi / Han, Xingjian et al. | IEEE | 2024


    EAFMNet: An Efficient Attention Fusion Mechanism for Lane Detection

    Ran, Hao / Yin, Yunfei / Huang, Faliang et al. | IEEE | 2022


    Improved Asphalt Pavement Crack Detection Model Based on Shuffle Attention and Feature Fusion

    Tursun Mamat / Abdukeram Dolkun / Runchang He et al. | DOAJ | 2025

    Free access

    Ship target detection based on CBAM-YOLOv8

    Zhang, Jiandong | SPIE | 2024