Fault Recognition Based On Three Dimensional Imaging Of OCS

Electric railway power supply system is constituted by overhead contact system(OCS)and traction substation.OCS is the main structure of electric power supply system for high speed railway,which is comprised of contact suspension,support device,positioning device and other parts.OCS’s performance directly affects the speed and safety of high-speed train.OCS works in the harsh and complicated environments along the railway,which leads to frequent faults.Therefore,OCS’ real-time detection is crucial and essential.The detection method of OCS mainly includes contact and non-contact ways.Contact detection technology obtains dynamic parameters of pantograph-OCS system by adding sensors,which non-contact technology obtains the parameters by means of laser,ultrasonic,radar and other devices.The non-contact detection uses the obtained parameters data analysis to evaluate the operating status of pantograph.Traditional contact detection is time-consuming and inefficient,which doesn’t meet the needs of the development of modern railway.In this paper,the 3D model of the OCS is reconstructed by the way of three-dimensional computer modeling,then extracted the defective insulator from the 3D model and detected it in non-contact way.With reference to the difficulty of identifying the parameters of the catenary resulting from the analysis on two-dimensional plane image of the OCS 6C system being highly affected by light and surface reflection of the parts,in this paper,three-dimensional model is created by using continuous acquisition of multi frame catenary point cloud data and analysis on physique characteristic of the three-dimensional model through the parts of catenary is conducted so that the parameters of the catenary can be determined.Keypoints are acquired from the depth images of surface point cloud of the catenary through the algorithm of NARF(Normal Aligned Radial Feature,NARF)and algorithm of FPFH(Fast Point Feature Histograms,FPFH)is used to describe the feature of the key points and determine the matching relationship.After the mismatching points are eliminated through the use of SAC-IA(Sample Consensus Initial Alignment,SAC-IA),ICP(Iterative Closest Point,ICP)algorithm is used for precise registration.Ultimately,the three-dimensional model of the catenary is obtained via curved surface reconstruction.On the basis of this,both measuring the key structural parameters of the reconstructed three-dimensional model of the catenary and verifying the accuracy of the reconstructed OCS model are conducted.After extracting the defect insulator in reconstructed OCS model by point cloud segmentation,detect the insulator based on its normal surface information and quantitatively display the fault.The normal three-dimensional model of the corresponding parts is identified through CAD modeling platform.After that,use the Geomaigc Qualify software to analyze the differences between normal model and fault model.Finally,by comparing the efficiency and accuracy of the two methods,the advantages of the proposed non-contact method are verified.