现代电子技术2025，Vol.48Issue(20)：10-17,8.DOI:10.16652/j.issn.1004-373x.2025.20.002
基于Point-Attention点云分类的激光雷达故障诊断方法研究
Method of LiDAR fault diagnosis based on Point-Attention point cloud classification
摘要
Abstract
In the field of intelligent vehicles and autonomous robots,LiDAR sensors are widely used for environmental perception and object detection due to their high precision and reliability.Therefore,fault diagnosis of LiDAR is particularly important.Internal faults of LiDAR often come with firmware alerts,but fault detection caused by external environmental factors presents greater challenges.For example,occlusion faults in LiDAR point clouds caused by vehicle deformation or dirt are difficult to reflect directly at the firmware level and can only be diagnosed by means of external detection.On this basis,a method of Point-Attention-based LiDAR occlusion fault diagnosis is proposed.The model's ability to extract key features from point cloud data is enhanced by combining with multi-head geometric attention mechanism module,CBAM module and residual connection mechanism,thereby improving classification accuracy and robustness.On the real ScanObjectNN dataset and the ModelNet40 benchmark dataset,the experimental verification of the Point-Attention model is conducted.It can realize classification accuracy of 93.7%and 82.5%,respectively.The time feature capture mechanism is fused to make model better adapt to the temporal correlation in realscenarios,thereby handling LiDAR occlusion faults more accurately.The experimental results demonstrate that the proposed method can effectively diagnose LiDAR occlusion faults,and best overall accuracy can reach over 99%,providing an efficient and accurate solution for LiDAR fault diagnosis.关键词
激光雷达/故障诊断/点云分类/残差连接/遮挡检测/时间特征捕捉Key words
LiDAR/fault diagnosis/point cloud classification/residual connection/occlusion detection/time feature capture分类
信息技术与安全科学引用本文复制引用
谭光兴,程星,陈海峰..基于Point-Attention点云分类的激光雷达故障诊断方法研究[J].现代电子技术,2025,48(20):10-17,8.基金项目
国家自然科学基金项目(61563005) （61563005）
