基于DD-CycleGAN的道路检测模型研究

Research on road detection model based on DD-CycleGAN

道路检测对于辅助驾驶而言仍具有挑战性。为了获得更准确的道路检测结果,提出一种结合深度学习与自适应检测的道路检测模型,该模型可以有效地提取道路特征并完成道路检测任务。首先,采用双判别器周期一致的生成对抗网络(DD-CycleGAN)作为全文的基础框架网络。其次,在生成器中添加空间卷积神经网络(CNN)以及残差密集块,进一步提升生成器的性能。最后,提出一种自适应的优化模型来提高道路检测的准确度。实验结果表明:提出的模型在KITTI道路基准数据集上精度达到了92.15%,明显优于传统的道路检测算法。

Road detection is still challenging for assisted driving.In order to obtain more accurate road detection results,a road detection model combining deep learning and adaptive detection,which can effectively extract road features and complete road detection tasks is proposed.Firstly,double discriminant(DD)-Cycle generative adversarial network(GAN)is used as the basic framework network of the full text.Secondly,add a spatial convolutional neural network(CNN)and a residual dense block to the generator to further improve the performance of the generator.Finally,an adaptive optimization model is proposed to improve the accuracy of road detection.The experimental results show that the precision of the proposed model on the KITTI road benchmark dataset reaches 92.15%,which is significantly prior to the traditional road detection algorithm.