Research on Lane Line Detection Based on Around View System

Traditional lane detection methods are limited by factors such as camera position and perspective, and often encounter issues such as false detection and missed detection. This article conducts research on lane detection methods from the perspective of multi camera BEV, and proposes a circular lane detection method based on convolutional neural networks (CNN). In order to solve the occlusion problem perceived from traditional forward looking perspectives, an around system was constructed using multiple cameras, and a multi classification semantic segmentation network was innovatively designed to predict obstructions, greatly reducing the false detection rate of obstructed lane lines. After verification, the algorithm proposed in this article can achieve good lane line detection results in different environments.