【摘要】 针对车道线检测在阴影、光照不足等复杂环境中精度受限的问题，设计了基于编码器-解码器架构的车道线检测网络。该网络的核心在于将优化后的CBAM注意力机制整合至MobileNetV2网络的倒置残差结构中，大幅提升了车道线特征的提取能力。在解码器模块，采用了高效的上采样技术，精准重构高分辨率分割图像。经过模型对比与实验验证，所提出的车道线检测算法，在应对复杂的交通场景时，展现出了尤为出色的检测性能。更多还原

【Abstract】 Addressing the problem of limited accuracy in lane detection under complex environments, such as shadows and inadequate lighting, this paper proposes a lane detection network based on an encoder-decoder architecture.The core of this network lies in the integration of an optimized Convolutional Block Attention Module(CBAM)attention mechanism into the inverted residual structure of the MobileNetV2 network, significantly enhancing the lane feature extraction capability.In the decoder module, efficient upsampling techniques are employed to accurately reconstruct high-resolution segmentation images.Results from model comparisons and experimental validations demonstrate that the proposed lane detection algorithm exhibits particularly outstanding performance when dealing with complex traffic scenarios.更多还原