Research On Online Monitoring System Of Catenary Compensator Based On Combined PCA Dimensionality Reduction And HOG Feature Extraction

With the continuous increase of the speed and density of railway locomotive in China,the bow net flow of electric locomotive is increasing,and the stable operation of the catenary tension compensator is particularly important.In order to ensure the power supply safety of the catenary of the railway system,the compensator monitoring device Came into being.This device monitors the distance from the fall to the ground(b value of the compensator),thereby judging the operating status of the compensator,and eliminating potential safety hazards in a timely manner.It is an important part of the ground monitoring device(6C system)of the contact network and power supply equipment.There are two prominent problems with the current compensator monitoring system.The first is that due to the harsh working environment and the system is exposed to natural environments for a long time,the ranging sensor probe may be blocked by debris,affecting the true value of the ranging and causing the processor to misjudge.Increase the workload of personnel maintenance.The second is that the main processor of the on-site monitoring sub-station cannot implement complex algorithms.The data must be uploaded to the monitoring center for processing.The real-time performance is not high and data loss is prone to occur.Aiming at the above problems,this paper designed a catenary compensator online monitoring system combining PCA dimensionality reduction and HOG feature extraction combined with digital image processing technology.The main tasks include:(1)The hardware circuits and software functions of the data acquisition module,network communication module and power module of the monitoring system are designed respectively.A set of on-site monitoring system with Raspberry Pi as the core controller and a combination of various hardware acquisition equipment is set up for real-time monitoring.On-site environment temperature and humidity and b value of the compensator and data over-judgment to realize timely processing of data and avoid information loss.The hardware circuits and software functions of the data acquisition module,network communication module and power module of the monitoring system are designed respectively.A set of on-site monitoring system with Raspberry Pi as the core controller and a combination of various hardware acquisition equipment is set up for real-time monitoring.On-site environment temperature and humidity and b value of the compensator and data over-judgment to realize timely processing of data and avoid information loss.(2)The on-site monitoring sub-station collects on-site images when the data exceeds the limit,and uses digital image processing technology to realize the secondary judgment.To prevent the loss of image detail information,first use side window(SMF)-bilateral filtering to globally denoise the image to reduce the effects of noise such as electromagnetic interference;through the setting of the time threshold,use the CLAHE algorithm to judge the system as low-light at night The image is enhanced.This algorithm effectively prevents noise amplification and significantly improves edge details while enhancing.Due to the special operating environment,in order to reduce redundant calculations,the ranging probe part in the image is extracted as the target area.The station collects environmental images only when the data exceeds the limit.The image interval between adjacent frames may be longer and the background changes.The traditional frame-to-frame difference method cannot be used to accurately identify the image.Therefore,a combination of PCA dimensionality reduction and HOG feature extraction is adopted.Perform verification of the target area.Experiments show that the method has low computational overhead and high accuracy,and can complete the automatic diagnosis of foreign body intrusion limits in the target area and reduce false positives.(3)Based on the Alibaba Cloud platform development system,the cloud server and the My SQL database are used to upload and store data;using the C # programming language and relying on the Visual Studio integrated design environment to develop client software for the monitoring center,staff users can view alarm information and historical data At the same time,it provides the parameter threshold setting function to achieve differentiated management.After the test,the system runs stably and the detection accuracy is high,real-time monitoring of the compensation device is realized,and it provides technical support for improving the lean level of safety management of catenary power supply.