A method of lane alignment restoration based on resnet network with fan-grid data characteristics

The development of lane detection technology has a crucial impact on the field of driverless environment detection. In order to solve the common non-visual detection problem in lane detection, this paper proposes a lane detection algorithm based on Res-Net network and sector grid road dataset annotation model on the basis of deep learning-based lane detection method. Firstly, a road sector grid is built to label the data set. Then Res-Net18 lightweight deep learning network is used to train and learn the model. Finally, the trained model is compared with Ultra-Fast-Lane-Detection network algorithm and lanenet method. The experimental results show that the detection accuracy of the proposed method is 2.4% and 1% higher than that of Ultra-Fast-Lane-Detection and lanenet methods, respectively, and the detection robustness in the absence of vision in lane detection problems is improved. In this paper, the algorithm can effectively restore and predict the blocked lane line under the condition of no visual road, and improve the detection accuracy.