Claims:I/We claim:
A vision based system for inspection of air brake in railway coaches comprising:
one or more data acquisition sub system, for sensing, aggregating and transmitting images;
at least one data processing sub system, for processing of images using image processing algorithm; and
a result visualization sub system, as displaying device for rendering of results.
The system as claimed in claim 1, wherein said data acquisition subsystem further comprises one or more devices for image sensing; and a network switch for aggregation and transmission of image.
The system as claimed in claim 2, wherein said device for image sensing is IP camera operate in night or infrared mode.
The system as claimed in claim 2, wherein said network switch for aggregation and transmission of images is power over ethernet (POE) switch used to power cameras.
The system as claimed in claim 4, wherein said switch is 10 port IEEE 802.3at/af standard.
The data processing subsystem as claimed in claim 1, wherein the data processing unit uses image processing algorithm configured to: cropping of acquired images, determining the edge of images, detection of straight lines, determination of the length of lines above threshold for rending of results.
The subsystem as claimed in claim 8, wherein the data processing unit consists of an Intel Celeron 2Ghz industrial PC.
A method of inspection of air brake in railway coaches using computer vision, the method comprising:
releasing of control instructions from the command computer;
providing acknowledgement of receive request from processing unit and instructing image sensing device;
starting capture of images using image sensing device;
aggregating and transmitting the images using ethernet switch;
processing and analysing of images using image processing algorithm at computing device; and
rendering of result at command computer.
The method as claimed in claim 8, wherein the step of processing of images comprises cropping of acquired images,
determining the edge of images,
detection of straight lines,
determining the length of lines above threshold for rending of results.
The method as claimed in claim 9, wherein canny edge algorithm is applied to determine the edge of images.
The method as claimed in claim 9, wherein Hough line transform is applied for detection of straight lines.
The method as claimed in claim 9, wherein Euclidean distance is used to determine the length of lines. , Description:Field of the invention
The present invention generally relates to field of image processing. In particular, the invention relates to computer vision-based inspection system that includes capturing and processing of input Infra-red (IR) images and display the status of disc brake in real time. The computer vision is evolving as an advanced branch of image processing which overcomes the limitations of manual inspection in large scale manufacturing.
Background
The air brake in railway coaches uses compressed air passing through the brake pipe and make changes in the motion. The air brake system is classified as single pipe air brake system and twin pipe air brake system. The system further consists of charging stage, application stage and release stage. In twin pipe system with charging stage, brake pipe (BP) is charged to 5 kg/cm2 pressure which in turn charges control reservoir and auxiliary reservoir through distributor valve. The feed pipe is charged up to 6kg/cm2 which in turn control the reservoir and auxiliary reservoir. At this stage, brake cylinder gets vented to atmosphere through passage in distributor valve. In application stage, when the brakes have to be applied, the pressure in brake pipe has to be dropped by venting air from driver’s brake valve. Reduction in brake pipe pressure positions the distributor valve in such a way that the control reservoir gets disconnected from brake pipe and auxiliary reservoir gets connected to brake cylinder. This results in increase in air pressure in brake cylinder resulting in application of brakes. The magnitude of braking force is proportional to reduction in brake pipe pressure. For releasing brakes, the brake pipe is again charged to 5 kg/cm2 pressure by compressor through driver’s brake valve. The auxiliary reservoir gets isolated from brake cylinder and brake cylinder is vented to atmosphere through distributor valve and thus brakes are released.
The railway coaches are subjected to an elaborate sequence of well-defined tests by the mechanical staff before train used for the subsequent schedule. The various tests include NRV functioning test, brake fading test, leakage test, Bogie isolation test, CR overcharging test, PEASD test, continuity test, parking breaks test, leakage test 2 and manual release test etc. to ensure proper functioning of the brake valve. A technician has to conduct various inspections and tests on all the coaches to check and ensure proper functioning of disc brake as per railways standard. The technician inspects the changes in disc brake status as the air pressure is getting change in the brake pipe from test rig.
Trains will not be used for the next onward journey unless all the tests are conducted and clearance certificate is provided. The proper functioning of airbrake system under various specified test conditions on all the coaches is inspected before the coaches are permitted to be used for the next scheduled trip.
The inspection involves moving the entire rake to a pit line. One important task of the mandatory inspection protocol is the air-brake inspection. The air brake inspection of brake pad status when the pressure is changed in the brake pipe (BP) from the test rig each time. A skilled worker walks under the coaches in the pit line, and manually inspect all the brake pads. The technician walks till the last coach about 600 meters in the pit line to complete one iteration of inspection. It takes about 10-15 minutes to complete one proper walk by a skilled person.
So, there is enormous scope of faultless and scalable brake inspection solution which can save on the time taken than manual inspection.
The Chinese application No. CN105882683A, entitled “Machine vision based technical Inspection & detection system and method for railway trains” relates to a system and method with indoor and outdoor unit. A light rail is placed in between the tracks for data exchange between transmission device and control server underside of rail. But this system includes heavy equipment-based inspection for freight trains only.
The Patent Application No. WO 2004/008067, entitled “Inspection of Railway Vehicles” relates to an apparatus within a box similar to sleeper and connected underside of rail. The apparatus has an optical sensor to monitor wear of wheel trends, brake pads or maintenance related issues etc. The invention includes a box with in sleepers used for inspection of railway vehicle. But the apparatus is more generic and related to maintenance tests for moving railway vehicle.
US Patent Application No. 20100100275A1, entitled “Thermal ageing-based vehicle analysis” provides solution for analysing a vehicle using multi-dimensional infrared image data acquired for the vehicle. A component of the vehicle can be identified within the infrared image data, and the infrared image data for the component can be analysed to determine whether any condition are present on the vehicle. But this proposed system uses thermal image to acquire diagnostic information on passing vehicles.
Therefore, a system and method for time efficient inspection with reduction in manpower requirement, data logging and result storage for future references is highly desired.
Summary
The present invention fulfils the foregoing needs by providing a computer vision-based system includes three sub systems data acquisition, data processing and results visualization. Data acquisition stage includes capturing of image as command from the server reaches Data processing unit (DPU). DPU process includes image aggregation, transmission and running of image processing algorithm. The final stage of visualization includes server to displays the brake inspection results. Image capturing, processing and data transmission happens in parallel manner in all DPUs.
The image processing algorithm in DPU is realized through machine learning approach to Artificial Intelligence (AI) and the design of Graphical User Interface (GUI) for rendering the results of brake tests. The algorithm sets the minimum threshold of lines length to decide the status of disc brake.
Hence a modular design with visual characteristic is leveraged to discriminate between brake (applied/released) inspection of air brake system in railway coaches. A deep learning approach to Artificial Intelligence (AI) and computer vision automate this task that human visual system can do. This approach makes it possible to get deeper insight and enable informed decisions making with computer vision. This scalable extensible modular design can be easily and seamlessly replicated for multiple pit lines.
Brief description of figures
Exemplary embodiments of the present invention will become more fully understood from the description and the accompanying figures, wherein:
Figure1: illustrate network diagram of Brake Monitoring Unit (BMU) System.
Figure2: shows camera placement diagram in the coach
Figure3: shows message flow diagram.
Figure4: shows functional architecture of a Brake Monitoring unit (BMU).
Figure5: shows flow chart of steps of Image processing Algorithm.
Figure6: shows screen snap shots for visualization of test results displayed on server display screen.
Figure 7a: shows image of brake released.
Figure 7b: shows image of brake Applied.
Figure 8(a-d): shows screen snap shot for visualization of result displaying ‘Unknown’, ‘InProgress’, ‘Engaged’, ‘Disengaged’ status.
Detailed description of invention
The foregoing description of the embodiments, the various features, and advantageous details of the invention has been presented for the purpose of illustration. It is not intended to be exhaustive or to limit the invention to the precise form disclosed as many modifications and variations are possible in light of this disclosure for a person skilled in the art in view of the figures, description and claims. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Unless defined otherwise, all technical and scientific terms used herein have the same meanings as commonly understood by person skilled in the art.
The embodiments herein below provide a computer vision based system for air brake inspection of railway coaches in the maintenance pit line. The system includes data acquisition, data processing and rendering of the results. The computer vision based system for air brake inspection runs with image processing algorithm for inspection of brake pad status.
The computer vision based system reduce the time required to conduct the test. The possibility of detecting scenarios beyond average human comprehensions, and data logging and storage for future reference are added benefits. The system deploys infrared image sensing device, ethernet switch as aggregator and computing device for image processing. The processed image from each coach are sent to a server which displays the status of disc brake and also serve as command computer.
In accordance with an exemplary embodiment, a method of air brake inspection on all the coaches on pit line is disclosed. It comprises of three sub systems data acquisition, data processing and results visualization. The method involves image sensing device start capturing the image as command from the server reaches Data processing unit (DPU). Power over ethernet switch (POE) start image aggregation and transmission. And industrial grade compact size computer deploys running of image processing algorithm. Once the image processing is done, the aggregated result of position of all disc break is sent to the server for visualization of test results. The method involves checking the final result of brake testing that brake pads are connected or disconnected with rail wheels.
Figure1 Brake monitoring units (BMU) 100 system architecture comprises three sub systems. Data acquisition 101, Data Processing 102 and result visualization 103.
In Data acquisition 101, each Coach having IP cameras 105 corresponding to each wheel (A01-04) (B01-04). Figure 2 shows the wheel arrangement (A01-04) (B01-04) in coach C01 and the IP cameras are mounted near each wheel across the pit line i.e eight cameras per coach (per BMU). IP cameras 105 are mounted and tilted around their vertical axis to acquire the images of the undercarriage view of the wheel and the brake pad. The cameras 105 are operated in night or infrared mode to facilitate carrying out the tests during night time. This subsystem capture images from all the cameras.
Further in figure 1, data Processing subsystem 102 uses ethernet switch 106 aggregate all the available images and send to Data Processing unit (DPU) 107 for image processing. Power supply unit (PSU) 108 is passing through a 230/110AC to 110v dc adaptor for ethernet switch 106 that supply power to the network cameras 105. PSU 108 is passing through a 230/110AC to 24v dc adaptor to supply power to data processing unit (DPU) 107. Data processing comprises the image processing algorithms realized through AI machine learning approach and the design of GUI for rendering the results.
The aggregated result of position of brake discs in a coach C01 is rendered on server 104 display.
Figure 3 shows the message flow diagram between Brake monitoring units (BMU) client 100 and Brake monitoring units (BMU) server 104. The server 104 sends a request for image capture. The images are captured by the IP cameras 105. BMU 100 acknowledge the request and also send the images obtained from all cameras.
Figure 4 illustrates the Brake monitoring system (BMU) 100 configuration for a multitude of Brake monitoring units (BMU)(1……n ) where n is the number of BMUs. The number of BMUs 401 is equal to the number of coaches under test in the pit line. Each BMU 401 is connected with n-port switch 402 thought Ethernet. The n-port switch connects to the server 403 through ethernet.
Fig 5 describes the steps of image processing algorithm 500. The acquired image 501 is cropped 502 to retain the Area of Interest (AOI) to eliminate the noise back ground. Canny edge algorithm 503 is applied to determine the edges in the image. Hough line transform 504 is applied to the Canny edge detected image to determine straight lines. Further the length of each of the detected lines is computed. Euclidean distance 505 is used to determine the length. The short length lines are eliminated using a threshold 506. Finally, a decision to check if the break is applied 507 or not applied 508 is determined using the threshold 506 of the length of lines present in the image.
Figure 6 provides the screen shot of the visualization of test results displayed on server computer screen with web-based GUI. The vertical axis shows the coach numbers and horizontal axis shows wheel numbers interfaced with IP cameras. The four possible status on server computer screen are Engaged, disengaged, unknown and InProgress. Engaged status describes the interconnection of brake pad with wheel. Disengaged status describes about separation of brake pad from the wheel. The unknown status come to know that there is no response from the coach. InProgress status shows the results are under progress. When the brake is applied/released from server side, IP cameras capture the images and send back to server that displays the real time status of brake pad.
Figure 7a and 7b shows the under carriage of the brake and brake pad. In fig 7a, there is a gap between the wheel and brake pad. So, the brakes are released. And in another fig 7b, there is no gap between the brake pad and the wheel. So, the brakes are applied.
The proposed solution conducted several experiments for inspection of coaches. It has been conducted with two coaches (C01-02) each having eight wheels (A01-04) (B01-04) and eight IP cameras each of 4MP fixed bullet network cameras is used for each coach across each wheel. The images from all the eight wheels in a coach are aggregated by 10 port IEEE802.3 at/af Power over ethernet (POE) switch. Intel Celeron Processor J1900(Quad core 2M Cache, 2Ghz) Industrial PC runs with the IP algorithm on all transmitted images. Once the image processing is done, the aggregated result of all the eight brakes is sent to the server.
The image capturing, processing and data transmission happens in parallel manner.
In first iteration of experiment, when the brakes are “Reset”, the server GUI displays ‘Unknown’ status shown in figure 8(a).
In second iteration of experiment, when the “Brake Applied”, the server GUI displays ‘InProgress’ status for 6 seconds shown in figure 8(b). After 6 seconds, the server GUI displays ‘Engaged’ status shown in figure 8(c).
In third iteration of experiment, when the “Brake Released”, the server GUI displays ‘InProgress’ status for 6 seconds shown in figure 8(b). After 6 seconds, the server GUI displays ‘Disengaged’ status shown in figure 8(d).
Hence, the proposed system provides vision based inspection of air brake system in 6 seconds than manual inspection of 10-15 minutes for entire pit line which is about 600 meters. The proposed system is non-intrusive and more time efficient than manual inspection.