基于 MultiRes+UNet 网络的车道线检测算法

无人驾驶技术改变人类生活方式,带车道线属性的高精地图,是无人驾驶领域的重要一环。 针对现有算法在车道线检测时存在准确率低、效率低等问题提出基于 MultiRes+UNet 检测方法。 该方法通过空洞卷积扩大卷积感受野,从而对全局信息统筹,运用 MultiRes block 和 Res path 结构减轻编码器-解码器特征之间的差异,大大降低了内存的需求。 实验结果表明,此算法在保证检测准确率的同时, 提高了算法的稳定性和运行速率,在纯车道、复合车道、阴影污损车道等多情况下,调和平均值分数分别为 0. 959、0. 942、0. 891,该算法存在高效性、高鲁棒性。

Lane line detection algorithm based on MultiRes+UNet network

Unmanned technology changes human lifestyle. The high-precision map with the lane line attribute plays a crucial role in theunmanned field. Proposing a detection method based on MultiRes + UNet network, aiming at the problem that the low accuracy and lowefficiency of existing algorithms in the detection of composite lane lines. This method expands the convolution receptive field by dilatedconvolution to co-ordinate global information. The MultiRes block and Res path structure are used to reduce the difference between theencoder and decoder features, which greatly reduces the memory requirement. The experimental results show that the proposed algorithmimproves the detection speed of the algorithm while ensuring the detection accuracy. When the pure lane, compound lane and shadowfouling lanes are obtained, the harmonic mean F1 scores are 0. 959, 0. 942, and 0. 891, the algorithm is high efficiency and highrobustness.