LaneSegNet:一种高效的车道线检测方法

车道线检测在智能交通领域占有重要地位,其检测的准确度和速度对于辅助驾驶以及自动驾驶有重要影响.针对目前深度学习方法识别车道线精度差、速度慢的问题,提出了一种高效的车道线分割方法LaneSegNet.首先基于编码和解码网络原理构建主干网络Lane-Net,用于提取车道线特征信息并分割出车道线;然后使用多尺度空洞卷积特征融合网络,可以极大地扩充模型的感受野,提取全局特征信息;最后使用混合注意力网络获取丰富的车道线特征,并增强与当前任务相关的信息.实验结果表明:在TuSimple数据集上,该方法检测车道线的准确率为97.6％;在CULane数据集上,该方法在标准路面的检测准确率达到92.5％,多种路面综合检测准确率为75.2％.本文提出的LaneSegNet车道线检测方法分割精确度和推理速度优于其他对比模型,且具有更强的适应性和鲁棒性.

LaneSegNet: an efficient lane line detection method

Lane detection plays an important role in intelligent transportation.The accurate and fast lane detection is important for assisted driving and automatic driving.In view of the poor accuracy and slow speed of deep learning methods for lane line recognition,a method abbreviated as LaneSegNet is proposed for efficient lane line segmentation.First,based on the principle of encoding and decoding network,a backbone network Lane-Net is constructed to extract the lane line features and segment the lane lines.Then,the multi-scale dilated convolution feature fusion network is used to greatly expand the receptive field of the model and extract the global features.Finally,the hybrid attention network is used to obtain rich lane line features and enhance the information related to the current task.The experimental results show that the accuracy of this method is 97.6% on TuSimple dataset,while on the CULane dataset,the detection accuracies are 92.5% and 75.2% for standard pavement and multiple pavements,respectively.Compared with other models,the proposed LaneSegNet has better segmentation accuracy and reasoning speed,and has stronger adaptability and robustness.