基于激光雷达与机载光电成像融合的高速轨道形变检测系统设计

Design of high-speed orbit deformation detection system based on the fusion of LiDAR and airborne optoelectronic imaging

摘要:

高速轨道形变是指高速铁路轨道在运营过程中由于列车荷载、温度变化、地基沉降等因素而发生的变形现象,极大影响了高速铁路车辆的行驶安全性,为此,设计了基于激光雷达与机载光电成像融合的高速轨道形变检测系统。通过激光发射器、激光信号接收器的结构设计与组装,实现激光雷达成像模块的设计,从无人机、光电摄像机、云台等方面,完成对机载光电成像模块的设计,实现图像数据的融合处理。在硬件优化系统的支持下,生成高速轨道的激光雷达与机载光电图像,经过去噪、畸变校正、去雾等操作,完成对初始图像的预处理。采用加权融合的方式,得到激光雷达与机载光电融合图像,并提取融合图像的边缘特征,计算高速轨道位置坐标,通过坐标计算结果与轨道初始结构数据的比对,输出高速轨道形变的检测结果。实验结果表明:所设计系统的平均状态误检率为2.3%,形变量检测误差平均值为0.15mm,具有良好的检测能力。

Abstract:

High speed track deformation refers to the deformation phenomenon caused by train load, temperature change, foundation settlement and other factors during the operation of high-speed Railway track, which greatly affects the driving safety of high-speed railway vehicles. Therefore, a high-speed track deformation detection system based on the fusion of laser radar and airborne photoelectric imaging is designed. By designing and assembling the structure of the laser transmitter and laser signal receiver, the design of the LiDAR imaging module is achieved. The design of the airborne photoelectric imaging module is completed from aspects such as drones, photoelectric cameras, and PTZ, achieving the fusion processing of image data. With the support of the hardware optimization system, generate high-speed orbit LiDAR and airborne optoelectronic images, and complete the preprocessing of the initial image through operations such as denoising, distortion correction, and defogging. By using a weighted fusion method, the fusion image of LiDAR and airborne optoelectronics is obtained, and the edge features of the fusion image are extracted. The high-speed track position coordinates are calculated, and the detection results of high-speed track deformation are output by comparing the coordinate calculation results with the initial structure data of the track. The experimental results show that the average state error detection rate of the designed system is 2.3%, and the average error value of shape variable detection is 0.15mm, indicating good detection ability.