The invention relates to a train track detection method based on a laser radar, and the method comprises the following steps: collecting the point cloud data of a track in front of a train through the laser radar, and creating a point cloud aerial view; sampling the point cloud aerial view to form a data set, and performing track labeling on each picture in the data set; constructing a trunk convolutional neural network model, extracting image features, extracting local features based on the image features, and adding an attention mechanism to the local features to form global features; constructing a track detection model based on the local features and the global features; and inputting the to-be-detected point cloud aerial view into the trained track detection model to obtain predicted track line information, and marking the train track in the original point cloud data. According to the method, the three-dimensional point cloud data collected by the laser radar is used for identifying and positioning the track, the track is marked in the point cloud, and the method is not easily influenced by environmental factors such as illumination, is wider in applicability, and has higher detection precision and detection rate at the same time.
本发明涉及一种基于激光雷达的列车轨道检测方法,包括如下步骤:通过激光雷达采集列车前方轨道的点云数据,创建点云鸟瞰图;对点云鸟瞰图进行采样形成数据集,对数据集中的每张图片进行轨道标注;构建主干卷积神经网络模型,提取图像特征,基于图像特征提取局部特征,在局部特征上加入注意力机制,形成全局特征;基于局部特征和全局特征,构建轨道检测模型;将待检测的点云鸟瞰图输入到训练好的轨道检测模型中,获取预测的轨道线信息,并在原始点云数据中标记列车轨道。本发明利用激光雷达收集的三维点云数据对轨道进行识别和定位,在点云中标记出来,该方法不易受光照等环境因素的影响,适用性更广,同时具有更高的检测精度和检测速率。
Train track detection method based on laser radar
一种基于激光雷达的列车轨道检测方法
2023-07-28
Patent
Electronic Resource
Chinese
City domain train track obstacle detection system based on machine vision and radar
European Patent Office | 2024
|European Patent Office | 2023
|