The present invention relates to using a machine vision system to scan passing railroad cars in a train in order to detect a low hanging air hose and/or coupler, thereby avoiding a broken air hose and/or uncoupling which can cause emergency braking or accidents
It is common practice by railroad industries worldwide to utilize an air braking system on both passenger and freight trains. The air brake system is typically composed of a compressed air source and control valves that are normally located on the locomotives and connected through a series of pipes and hoses to the brake valves and brake actuators on each railroad car. The air supply is carried across the span between coupled railroad cars, referred to herein as the “coupler gap”, through a flexible hose/coupler arrangement 26 and 27 in
The invention described herein is designed to measure the height of the air hose assembly in the coupler gap area between each pair of coupled railroad cars and to automatically send a warning if the minimum air hose height is found to be below a certain limit above the top of rail, typically 4-5 inches. The invention relies on a “machine vision” system that utilizes a camera and computer interface to acquire images of the coupler gap area continuously through the train. These coupler gap images are analyzed in real time with a computer program that seeks to identify the air hose in each image and measure the minimum height of the air hose relative to the top edge of the rail in the image. The invention functions automatically and unattended by virtue of a computerized control system.
An additional function of the invention is to measure the height of each railroad car coupler above the top edge of the rail (
The coupler height measurement functions in tandem with the air hose height measurement system requiring no additional equipment. However, the image analysis software program is modified by the addition of an image analysis module that seeks to find the mating couplers in the coupler gap image and measure the height of each coupler shank mid-point above the top of rail. The coupler height value is then compared to the limits imposed by the railroad and a warning issued if coupler height falls outside of this range.
Historically, manual inspection of parked railroad cars has been used to detect a low hanging air hose. If a low hanging air hose catches debris on the ground and brakes, then an emergency braking sequence is initiated. These emergency braking scenarios cost railroads millions of dollars of lost revenue through down time. Also, if the emergency braking mechanisms were to fail, then a lack of braking pressure in a train causes a dangerous situation.
A brief summary of methods/apparatus for remotely detecting whether a specific object is in a safe position follows below.
Prior art consists of devices designed to detect low hanging air hose assemblies by detecting when such assemblies break one or more laser beams aimed across the railroad track to a receiver on the other side of the track. A monitoring system detects the change in the laser receiver output signal when the low hanging air hose blocks the beam and prevents it from impinging on the receiver. These devices are typically installed at a fixed height(s) above the top of the rails and may have single or multiple laser transmitter/receiver pairs. The devices typically include a railroad car wheel detector and railroad car radio identification tag reader such that the interrupting air hose assemblies can be associated with the appropriate pair of coupled railroad cars. The railroad car wheel detector data is also used to locate the coupler gaps where low-hanging air hose assemblies are located as other components in the train, such as wheels and truck frames, also interrupt the laser beams.
U.S. Pat. No. 6,411,215 (2002) to Shnier discloses placing a retro-reflective surface on a target such as a door's locking handle. Then a narrow light beam is focused on the desired position of the retro-reflective surface. If a monitoring device does not sense the reflected light, then an alarm is activated. Unfortunately, the dirt and grime associated with railroad coupling and braking devices would defeat this approach for detecting a low air hose.
U.S. Pat. No. 6,717,514 (2004) to Stein et al. discloses a radio transmitter mounted to a target. A trio of spatially positioned receivers detect an alarm condition when the transmitter is located outside of a safe perimeter. Unfortunately, the installation and maintenance of countless radio transmitters on air hose elements would be costly and prone to failures of the radio transmitters.
U.S. Pat. No. 6,778,092 (2004) to Braune discloses a camera, laser or two cameras connected to an evaluating unit which determines (using software) location and time variables in a safety zone of a robotic machine installed in a factory. The machine can be shut down when predetermined danger conditions exist. This technique is somewhat similar to the present invention, except the logic needed to analyze a passing train is not suggested, nor are solutions to variable outside weather and light conditions suggested.
U.S. Pat. No. 6,812,850 (2004) to Matsuuiya et al. discloses a CCD camera moving in X, Y, Z axes to track a work piece. A protector on the camera includes an antenna and a strain detector, used to prevent a collision of the camera and the work piece.
Pub. No. US2002/0196155 to McNulty, Jr. discloses a fixed laser beam hitting a mirror on a target. A door opening moves the mirror, interrupting the reflected beam, and signaling an alarm.
Pub. No. US2003/0160701 to Nakamura et al. discloses a container contents proximity sensor with a wireless transmitter to detect a terrorist entry into a container.
Salient Systems, Inc. discloses a low hose detector using an optical sensor, a light curtain sensor, to examine the area between railroad cars. Car tag readers also log offending cars. A plurality of lasers are beamed across the track and sensed by receivers. By measuring the height of the lowest beam interrupted by the air hose, a low air hose is detected, as well as all heights of all passing air hoses to within an inch.
GE Transportation Rail® produces a dragging equipment detector that consists of a bar mounted across the tracks. The bar is height adjustable to detect a low air hose by sensing the impact with the bar.
Lynxrail™, www.lynxrail.com, produces a video imaging system for passing railroad cars. It monitors via machine vision algorithms wheel profiles, brake shoe wear, springs, car identification, hand brake and draft gear. No backlit screen or equivalent is known to be used to compensate for daytime, nighttime and ambient weather conditions.
The prior art has the following problems:
1. Proper alignment of the laser transmitter and receiver must be maintained at all times.
2. The laser beam devices must be mounted across and in close proximity to the track and are subjected to severe vibrations that can cause the laser beam to deflect from the receiver, especially when railroad car wheels with flat spots pound the rails.
3. The housings for the laser and receiver interfere with normal track maintenance activities and may pose a trip hazard to railroad workers.
4. The simple beam-break approach is very poor at discriminating shapes and sizes. Broken straps, cables, debris precipitation and other objects can block the laser beam and cause a signal change at the receiver, referred to as a “false positive.”
5. The prior art only detects low hanging air hose assemblies and cannot be used for other detection tasks.
The present invention compensates for all lighting conditions by providing a lit screen or other light uniform background, such as a building or wall, to backlight the space between passing railroad cars. A camera is used to detect a low air hose via machine vision algorithms. A car gap detector can be used to utilize a laser beam to sense when the gap between the cars is properly lined up with the video camera. A wheel counter can identify the car with a defectively low air hose. An image acquisition control computer can send an alarm to a remote location via the internet.
The invention described herein offers significant improvements over the prior art:
1. The imaging camera is located well off the track (27′ from track centerline) in a protective enclosure and is not subjected to vibrations.
2. Being located well off the track, the imaging camera does not interfere with normal track maintenance activities in any way or create a trip hazard to railroad workers.
3. The image analysis software contains sophisticated algorithms to discriminate against any object in the image that does not have the precise characteristic size and shape of an air hose. The analytical technique executed in the software is very proficient at avoiding false positive indications from broken straps, cables, debris, precipitation or other objects that may enter the field of view of the imaging camera.
4. The invention is also useful for measuring the railroad car coupler heights which are also located in the same “coupler gap” areas as the air hose assemblies.
An aspect of the present invention is to provide an air hose height detection system using a camera and a back lit screen, building, or wall to highlight the air hose video profile in all weather and light conditions.
Another aspect of the present invention is to provide a weather resistant enclosure for the camera.
Another aspect of the present invention is to provide three methods of detecting the “coupler gap” between adjacent railroad cars where the air hose assemblies reside.
Another aspect of the present invention is to provide a reliable machine vision algorithm to detect a low air hose.
Another aspect of the present invention is to provide a remote alarm communication sub-system connected to the low air hose detector.
Another aspect of the present invention is to provide a car coupler height detector within the same system as the low-air hose detector.
Other aspects of this invention will appear from the following description and appended claims, reference being made to the accompanying drawings forming a part of this specification wherein like reference characters designate corresponding parts in the several views.
An enclosed camera is mounted to aim perpendicular to a train track at a screen installed on the opposite side of the track. The screen has a light source. The source could be a searchlight aimed at it, or internal bulbs or a searchlight aimed from the back and through the screen. The screen could be replaced with a lightly colored uniform background, such as the side of a building or wall, with illumination for uniform image background at night or in low light conditions.
As a train passes by at perhaps forty miles per hour, the air hose is contrasted against the screen to create a video image captured by a control computer. Custom algorithms reliably define the outline of each air hose and determine a low air hose condition. Optional car coupler finder algorithms also operate.
A camera resides at the side of the track and images the specific coupler gap area of a passing train. An illuminated back light panel or wall provides for consistent contrast and nighttime operation. A sophisticated camera exposure adjustment algorithm compensates for variable ambient lighting conditions. A method of signaling the camera imaging system that a coupler gap is present in the camera field of view and to acquire the current image frame is provided. A “hose finder” algorithm detects hoses in images and finds their lowest point above the top of rail. The “hose finder” algorithm discriminates against cables, straps, and other objects in coupler gap. The top of rail is automatically detected by a machine vision algorithm for hose height measurement.
Before explaining the disclosed embodiment of the present invention in detail, it is to be understood that the invention is not limited in its application to the details of the particular arrangement shown, since the invention is capable of other embodiments. Also, the terminology used herein is for the purpose of description and not of limitation.
Referring first to
When a low air hose is detected by the computer 3, an alert can be sent remotely via phone, internet or microwave link 320.
Referring next to
Referring next to FIGS. 3,4 the camera 5 has a weatherproof housing 1900. The inside environment of the housing 1900 is controlled by an enclosure climate control unit 1901. In
Referring next to
In
In
Referring next to
As the train moves between the camera 5 and back light screen or wall 7 different parts of the cars 25 come into the camera's field of view and are imaged. Because only images of the coupler gaps are desired the invention employs one of three methods to signal the image acquisition and computer control programs when a coupler gap is present in the camera's field of view.
Referring next to
Referring next to
Referring next to
Referring next to
1. The image frame is scanned for the image of an air hose using a vertical line edge finder tool 5001. The tool is initially placed in the upper left corner of the image frame and is then moved horizontally and vertically across the image and the distances between dark/light and light/dark edge transitions are calculated.
2. The number of edge/edge distances falling within the specified range for an air hose width are tallied. Smaller objects such as straps 33,34 are ignored.
3. If the tally of edge/edge distances exceeds a specified target value then the shape is determined to be an air hose assembly image.
4. If the vertical line edge finder tool 5001 completes its scan of the image frame and the tally of edge/edge distances does not exceed the specified target value then a horizontal line edge finder tool 5002 is applied to the center left of the image and the scan repeated. This is necessary because air hoses can have primary vertical or horizontal orientations.
5. Once an air hose shape is found the lowest point of the contiguous air hose shape blob 5003 is found using an image analysis tool.
6. The straight line corresponding to the top edge of the rail 1 is found using an image analysis tool.
7. The vertical distance hmin from the top edge of the rail 1 and the lowest point on the air hose shape blob 5003 is calculated in pixels using an image analysis tool.
8. The image analysis program applies a vertical line edge finder tool 5004 centered horizontally above the lower point of the air hose shape blob 5003 and begins to scan the image frame for coupler images.
9. The coupler vertical line edge finder tool 5004 is moved to the left and the distance between light/dark and dark/light edge transitions calculated.
10. The number of edge/edge distances falling within the specified range for a coupler shank 30 are tallied. Objects larger or smaller than this range are ignored.
11. If the tally of edge/edge distances exceeds a specified target value then the shape is determined to be a left coupler shank 30.
12. If a left coupler is found or the coupler vertical line edge finder tool 5004 completes its scan of the image frame to the left and the tally of edge/edge distances does not exceed the specified target value then the tool is reset to the starting position and a scan to the right is initiated and steps 9-11 repeated.
13. Once left 30 and right coupler 31 are found in the image frame the horizontal centers of the shanks 5005,5006 are identified with an image analysis tool.
14. The vertical distances cL, cR between the top edge of the rail 1 and the left and right coupler center heights are calculated in pixels.
15. Scale factors are applied to the pixel values of the air hose assembly minimum height and the coupler center heights to convert these values to inches above the top of rail. The pixels/inch scale factors are obtained in a calibration procedure in which a scale calibrated in inches is attached to the center of the track and oriented vertically with “0” inches on the scale corresponding to the top of the rails. An image is acquired with the camera 5 and the scaling obtained by comparing the inch readings on the scale to their vertical pixel positions in the camera image frame.
Referring next to
Referring next to
Referring next to
1. The automatic system control (ASC) program (block 1100) continuously monitors the train presence detectors at the edges of the test zone (block 1101).
2. When a train passes over the train presence detectors 2 in
3. The ASC program activates the image lighting system block 1103, coupler gap detector (if so equipped block 1104) and internally signals the image acquisition/analysis (IAA) program to begin operation (block 2100).
4. The IAA program optimizes the camera exposure setting for the current illumination conditions of the camera imaging back light (block 2102).
5. The IAA program monitors the coupler gap detector for a low signal indicating that the first railroad car has arrived at the camera imaging field of view (block 2103).
6. The IAA program begins acquiring images of the railroad cars passing through the camera imaging field of view (block 2104).
7. The IAA program continues to monitor the coupler gap detector signal for a high signal indicating that a coupler gap is in the camera imaging field of view (blocks 2109,2103).
8. When it detects a high signal the IAA scans the coupler gap image (reference FIG. 13)for an air hose (block 2105) using the following technique:
9. When an air hose assembly image blob is found, the IAA program applies an image analysis tool to find the lowest point on the blob 5003 and then measures the vertical distance between the lowest point of the air hose assembly image and the top edge of the rail in pixels(block 2106).
10. The IAA program scans the current image for coupler images 31,32 (block 2107). The scan technique:
11. When coupler images are found, the IAA program uses an image analysis tool to find the vertical center of their shanks 5005, 5006 and then measures the height between the coupler shank mid-point above the top edge of the rail in pixels cL, cR (blocks 2108).
12. The IAA program resets itself and waits until the coupler gap detector signals that another coupler gap is present (block 2109) and then Steps 6-11 are repeated.
13. Concurrent with the IAA program operation, the ASC program continues to monitor and record the wheel voltage pulses and railroad car radio tag identification signals from system 1002 in
14. The ASC program (block 1100) monitors the train presence detector signals (2 in
15. The ASC program (block 1100) converts the air hose assembly minimum height hmin and coupler center height measurements cL and cR from pixels to inches using the previously determined scale factors.
16. The ASC program merges the air hose and coupler height data with the railroad car identification system data and stores in an electronic file for subsequent electronic transmission (block 1108 and
17. The ASC program (block 1100) generates an electronic report file of any railroad cars that were found to have air hose assemblies lower than the minimum limit or with couplers outside of the prescribed height variation limits (block 1109 and
18. The ASC program (block 1100) transmits these files to the appropriate railroad communication network destinations (block 1109).
19. The ASC program (block 1100) deactivates the car radio tag identification reader (11 in
Referring next to
1. Block 1105 (activate coupler gap detector)is eliminated because there is no coupler gap detector to activate.
2. Block 2103 is replaced with block 2103a in
3. Block 2104 is replaced with block 2104a in
4. Program flow proceeds from the “check air hose scan tool coordinates in image” FALSE branch in block 2105 to block 2104a because the “check coupler gap detector state” block in 2103 has been eliminated. Instead the program flow proceeds from the end of the image scan for air hose block 2105 to the replacement block 2104a (
5. Block 2109 is eliminated because there is no coupler gap detector. Instead, program flow proceeds from “check coupler left/right scan direction” RIGHT in block 2107 to block 2104a (
6. The “deactivate coupler gap detector” block is eliminated from block 1110 because there is no coupler gap detector.
Referring next to
1. Block 1105 (activate coupler gap detector)is eliminated because there is no coupler laser/receiver gap detector to activate.
2. Block 2103 is replaced with block 2103a in
3. Block 2104 is replaced with block 2104b in
4. Program flow proceeds from the “check air hose scan tool coordinates in image” FALSE branch in block 2105 to block 2104b because the “check coupler gap detector state” block in 2103 has been eliminated. Instead the program flow proceeds from the end of the image scan for air hose (block 2105) to the replacement block 2104b (
5. Block 2109 is eliminated because there is no coupler gap detector. Instead, program flow proceeds from “check coupler left/right scan direction” RIGHT in block 2107 to block 2104b (
6. Block 1108 is replaced with 1108a in
7. The “deactivate coupler gap detector” block is eliminated from block 1110 because there is no coupler gap detector.
The invention is composed of the components arranged as shown in
1. Automatic computerized image acquisition/control system (3)
2. Charge Coupled Device (CCD) camera with lens in a weatherproof enclosure (5)
3. Frame grabber camera interface in the data collection system (300)
4. Camera/frame grabber electrical cable (310)
5. Railroad car identification system (1002), comprised of a wheel detector (4) and a radio identification tag reader (11)
6. Laser/receiver “coupler gap” detector (1001)
7. Illuminated camera imaging back light or wall(7)
8. Communication link (320)
Below are the preferred configurations of the camera, lens, camera imaging back light and illumination. Other configurations would work as well for different camera-to-track distances, lens focal lengths, larger back lights, etc.
1. Camera: Electronic CCD camera with programmable electronic gain and shutter speed, ⅓″ CCD element, CS lens mount, minimum 30 frames/second scan rate.
2. Lens: Fixed or Variable Focal Length set at 20 mm, CS mount.
3. Camera Orientation and Location: Optical axis perpendicular to railroad track and set back 27′ from track centerline.
4. Illuminated Back Light or wall: 6′ high by 8′ long set back 10′ from track centerline opposite side of track from camera, bottom edge aligned slightly below top edge of rail in FOV, flat white surface coating, illuminated by at least 1000 watts of incandescent lighting.
5. Camera/Railroad Track Orientation: Railroad track should run as close to due east-west as possible at the location of the camera to avoid problems with fore-lighting of the train and back light shadowing in the early morning or late afternoon hours.
Although the present invention has been described with reference to preferred embodiments, numerous modifications and variations can be made and still the result will come within the scope of the invention. No limitation with respect to the specific embodiments disclosed herein is intended or should be inferred. Each apparatus embodiment described herein has numerous equivalents.
Number | Name | Date | Kind |
---|---|---|---|
4412212 | Kolegraff et al. | Oct 1983 | A |
5198800 | Tozawa et al. | Mar 1993 | A |
5331311 | Doctor | Jul 1994 | A |
5660470 | Mench | Aug 1997 | A |
6313753 | Butler | Nov 2001 | B1 |
6411215 | Shnier | Jun 2002 | B1 |
6717514 | Stein et al. | Apr 2004 | B1 |
6778092 | Braune | Aug 2004 | B2 |
6812850 | Matsumiya et al. | Nov 2004 | B2 |
20020196155 | McNulty, Jr. | Dec 2002 | A1 |
20030160701 | Nakamura et al. | Aug 2003 | A1 |
Number | Date | Country | |
---|---|---|---|
20060231685 A1 | Oct 2006 | US |