A Neural Network Model for Autonomous Vehicles Safety Check Using Camera-only/Camera-Lidar Images

Engineering notes

Experimental results demonstrate that the proposed methods significantly improve the safety and performance of autonomous driving by ensuring reliable sensor data fusion, enabling safe navigation even in challenging environments.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Camera and LiDAR sensors play a crucial role in vehicle perception systems, enabling accurate detection of obstacles and other vehicles in autonomous driving technologies. Despite their complementary advantages, ensuring the reliability of camera data under varying environmental conditions poses a significant challenge. This paper introduces a novel system for camera data consistency checking through two methods: (1) a camera-only consistency checking mechanism utilizing an effective point-to-point feature extraction neural network inspired by the D2-Net method, and (2) a camera-LiDAR fusion data consistency checking approach employing advanced sensor fusion techniques. By integrating a critical reliability check for camera data and fusing it with LiDAR information, our system leverages the strengths of both sensor types to enhance the robustness and efficiency of autonomous vehicle navigation. Experimental results demonstrate that the proposed methods significantly improve the safety and performance of autonomous driving by ensuring reliable sensor data fusion, enabling safe navigation even in challenging environments.

Links and sources

• Official / arXiv page

Comments