KR100716033B1 - Sensing System in a Traveling Railway Vehicle for Sensing a Human Body or an Obstacle on a Railway Track - Google Patents

Abstract

The present invention relates to a system for monitoring people and obstacles on a track in a traveling track vehicle.

The present invention provides near-field sensing data generating means for generating near-field sensing data in front of a track, provided at the head of a tracked vehicle, and remote sensing data for generating far-field sensing data in front of the track, provided in the center of the front of the tracked vehicle. Generating means, and a processing computer for fusing data output from said short-range sense data means and data output from said long-range sense data generating means to generate fused data.

  Surveillance Systems, Laser Scanners, Binocular Cameras, Sensor Fusion, World Models

Description

Sensing System in a Traveling Railway Vehicle for Sensing a Human Body or an Obstacle on a Railway Track

1 is a side cross-sectional view of a head of a high speed train or an urban railway vehicle equipped with a monitoring system according to an exemplary embodiment of the present invention.

2 is a front view of a high speed train or a city railroad vehicle head mounted with a monitoring system according to an exemplary embodiment of the present invention.

3 is a view showing the observation range of the 3D laser scanner 1 and / or the binocular camera 3 and the 2D laser scanner 2 used in the present invention.

4 is a diagram schematically illustrating a sensing system according to an exemplary embodiment of the present invention.

5A and 5B conceptually illustrate a world modeling method used in a preferred embodiment of the present invention.

6 to 8 are schematic diagrams illustrating a sensor fusion method according to an exemplary embodiment of the present invention.

Description of the main parts of the drawing

A: means for generating remote sensing data B: means for generating short sensing data

C: safety device

1: 3D laser scanner 2: 2D laser scanner for long range detection

3: 2D laser scanner for near field detection 4. Binocular camera

5: Data Processing Computer 11: 3D Laser Scanner Interface

12: Binocular Camera Interface 13: 2D Laser Scanner Interface

41: sensor fusion and world modeling unit 42: people and obstacle determination unit

43: control signal generator 50: the converted sensor coordinate data

51: sampling space 52: sampling pointer in two-dimensional coordinates

53: Sampling pointer of three-dimensional coordinates 61: Observation range of the 3D laser scanner

62: View range of 2D laser scanner for near field detection

63: 3D laser scanner detection image

64: 2D laser scanner detection for near field detection

66: 2D laser scanner and 3D laser scanner for near field detection

71: binocular camera viewing range

72: 2D laser scanner viewing range for near field detection

73: binocular camera detection image

74: 2D laser scanner detection for near field detection

76: 2D laser scanner fusion image for binocular camera and near field detection

81: binocular camera and 3D laser scanner viewing range

82: 2D laser scanner viewing range for near field detection

83: binocular camera detection image 84: a portion without matching data

85: 3D laser scanner detection image

87: 2D laser scanner detection image for near field detection

88: Fusion image of binocular camera, 3D laser scanner and 2D laser scanner for near field detection

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for detecting people and obstacles on a track, and in particular, when an object (person and obstacle) appears in front of a traveling track vehicle (high speed train, general train, urban railway vehicle, light rail vehicle, etc.) It relates to a system for detecting this using sensors installed at the head of the.

In general, a front obstacle is identified using a sensor such as an infrared sensor, an anti-collision radar sensor, an ultrasonic sensor, a CCD or an IR camera, as a method for identifying a front obstacle of a driving vehicle. These sensors have a problem that they are sensitive to the external environment (electric field, magnetic field, light, etc.). That is, the measured data is distorted due to the influence of the reflectance, color, etc. of the obstacle, or the influence of the external environment such as the electric field or the magnetic field, so that the measured data is different. There is a problem that can not be detected.

In addition, in the case of a traveling track vehicle, accurate distance information on a track and shape information about a person or an obstacle are required to accurately detect people and obstacles on a track. In other words, the detection of people and obstacles on the track requires at least tens of meters of distance performance and the capability of sensors with high distance resolution and distance accuracy of less than 10 cm for the detection of long distance obstacles. It does not satisfy two conditions at the same time. That is, the infrared sensor has a low distance performance, the radar sensor has a good distance performance, but is sensitive to the external environment, thereby reducing the reliability of the distance, and the ultrasonic sensor has a problem that the distance performance and the distance resolution are poor.

The present invention has been made to solve the conventional problems as described above, and to obtain accurate shape information and distance information for people and obstacles in the far and near in front of the track vehicle in the track vehicle running at high speed; It can be used to provide accurate judgment criteria for emergency situations and crash situations for tracked vehicles traveling at high speeds, so that people and obstacles on the track can be detected early and safety of life-saving devices such as air bags and collision warning devices can be used. It is an object of the present invention to provide a system for detecting people and obstacles on a track in a traveling track vehicle, which enables the device to be operated in a timely manner, thereby reducing the injuries and fatalities of human life.

In order to achieve the above object, the present invention provides a short-range sensing data generating means that is installed in the lower end and / or the center of the head of the tracked vehicle to generate near-field sensing data in front of the track; A remote sensing data generating means (A) installed at an upper end or a central portion of the track vehicle to generate remote sensing data in front of the track; By calculating the location and distance information of the objects (persons and obstacles) by fusing the sensed data from the short range sensing data generating means (B) and the remote sensing data generating means (A), respectively, the calculated objects (persons and obstacles) ) A data processing computer for determining whether the object collides with the tracked vehicle from the information and selectively generating a driving signal to the safety device C.

At this time, the remote sensing data generating means (A) is characterized in that any one or both of the 3D laser scanner and the 2D laser scanner for remote detection selectively installed.

In addition, the near-field detection data generating means (B) is characterized in that any one or both of the 2D laser scanner for the near-field detection and the binocular camera selectively installed.

In addition, the data processing computer includes data measured from the remote sensing data generation means A, which is composed of a 3D laser scanner (and / or a 2D laser scanner for remote detection), and a 2D laser scanner for short range detection (and / or a binocular camera). Sensor fusion and world modeling unit for receiving the measured data from the sensor fusion and world modeling; A person and obstacle determination unit that determines a person and an obstacle based on a world model generated by the sensor fusion and world modeling unit; And a control signal generator for generating a control signal to the safety device C in response to the result determined by the person and the obstacle determination unit.

Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

1 and 2 illustrate a cross-sectional side view and a front view of a head of a high speed train or an urban railway vehicle equipped with a monitoring system according to an exemplary embodiment of the present invention.

According to this, the 3D laser scanner 1, which is one of the remote sensing data generating means A, is installed at the center of the head of the tracked vehicle. In this case, in view of the viewing angle and the distance performance of the 3D laser scanner 1, You should look as far as possible on the track so you can see the distance.

In addition, the 3D laser scanner 1 may be installed in the center of the tracked vehicle, or the 2D laser scanner 2 for remote detection may be installed together with the 3D laser scanner 1, and if the 3D laser scanner 1 and the 2D laser scanner 2 for remote detection are installed. If all are installed, they may be located far from each other as long as they are sufficient to ensure their respective viewing angles.

In addition, the 2D laser scanner (not shown) for the remote detection can also observe the far and near by looking at the line as far as possible from the front of the line, similarly to the 3D laser scanner 1.

The binocular camera 4 is used as a short-range sensing data generating means B due to a relatively short distance performance (within 30m), in this case installed in the center of the head portion in consideration of the viewing angle and the distance performance of the binocular camera 4 It's best to look as close to the line as possible so you can see near.

Meanwhile, as shown in FIGS. 1 and 2, a 2D laser scanner for short range detection (3D) is installed at a predetermined height (about 30 cm on the track) from the track at the bottom of the tracked vehicle. The observation direction in (3) should be horizontal in front of the track so that the near field can be observed.

The height at which the near-field detection 2D laser scanner 3 is installed is preferably installed at a height of about 30 cm from the track in consideration of the size of the person and the obstacle to be detected and the viewable range.

In addition, the 3D laser scanner 1, the long-range detection laser scanner (2), the binocular camera (4) and the near-field detection 2D laser scanner (3) is installed inside the tracked vehicle as shown in FIG. The data processing computer 5 is connected via interfaces.

At this time, the 3D laser scanner 1 uses a LAN interface 11 in accordance with the specification of the 3D laser scanner, and the binocular camera 4 is an interface 13 capable of large-capacity communication for image data transmission. (USB or IEEE 1394), the 2D laser scanner for detecting the short distance (3) and the long distance (2), it is usually preferable to use the RS-422 interface 12 capable of communication at a communication speed of 300 (kbps) or more.

As described in detail below, the data processing computer 4 fuses the data transmitted from each sensor to determine the position, size and state of the object appearing in front of the track through the fusion data.

3 shows the viewing range of the 3D laser scanner 1 and / or the binocular camera 4 and the viewing range of the 2D laser scanner 3 for near field detection, and the trajectory as shown at the bottom of FIG. The distance of 30m or more in front of the vehicle is determined by the observation range 31 of the 3D laser scanner 1 and / or the 2D laser scanner 2 for remote detection, and the 2D laser scanner 3 for short-range detection near the vehicle 30m in front of the tracked vehicle. It is preferable to determine the observation range 32 of the binocular camera 4 and the observation range 33 of the binocular camera 4.

The reason is that the highest speed of the present fastest high speed train ("KTX") is about 300KPH, and in the case of the 3D laser scanner (about 10Hz) with the lowest sensor output rate of the present invention, the sensor detection as described below and these Since it takes about 237ms to process the data and generate the control signal for the safety device C, the distance required for the operation of the safety device C is at least 19.7m, so the safety device C is secured. This is because a distance of about 30m is required considering the 50% clearance for operation.

4 is a schematic diagram of a sensing system block according to a preferred embodiment of the present invention.

That is, the data processing computer 4 of the sensing system of the present invention is a short-range detection and the data measured from the remote sensing data generating means composed of a 3D laser scanner 1 and / or a 2D laser scanner for remote detection A sensor fusion and world modeling unit 41 which receives the measured data from the short-range sensing data generating means B including the 2D laser scanner 3 and the binocular camera 4 for sensor fusion and world modeling; A person and obstacle determination unit 42 for determining a person and an obstacle based on the world model generated by the sensor fusion and world modeling unit 41; The control signal generator 43 generates a control signal to a known safety device C in response to the result determined by the person and obstacle determination unit 42.

Thus, the data processing computer 4 can measure data from the 3D laser scanner 1 (and / or the 2D laser scanner 2 for long range detection) and the 2D laser scanner 3 for short range detection (and / or binocular). Upon receiving the measured data from the camera 4, the sensor fusion and the world modeling unit 41 generate the sensor fusion and the world model, and the person and the obstacle determination unit 42 based on the generated world model. By determining the obstacle, the control signal generator 43 generates a control signal such as a signal for operating a safety device C such as a lifesaving device or a collision warning device and / or a signal for sounding a warning sound.

The sensor fusion and world model performed by the sensor fusion and world modeling unit 41 in the data processing computer 4 will be described in detail below.

Before discussing sensor fusion and world modeling according to a preferred embodiment of the present invention, the general characteristics of the sensing and measuring sensor used in the present invention will be described first. In the case of the 3D laser scanner 1, the distance performance is excellent. In addition, the measurement data is accurate and the angle of view can be adjusted so that only the part to be measured can be selected and measured, while the data output rate is higher than that of the binocular camera 4 or the 2D laser scanner 3. The disadvantage is that it drops (up to 10Hz).

On the other hand, in the case of the binocular camera 4, the distance performance is high, the resolution is high, and the data output rate is higher than that of the 3D laser scanner 1 (up to 30 Hz). The disadvantage is that it is difficult to obtain perfect distance data for the observation range.

In addition, the short range (3) and long range (2) 2D laser scanners have the advantage of good distance performance, accurate measurement data, and high data output (up to 75 Hz). The disadvantage is that it can't be.

On the contrary, the 3D laser scanner 1 and the binocular camera 4 can continuously acquire three-dimensional shape information in front of the moving vehicle at regular intervals according to the sensor data output rate.

According to an embodiment of the present invention, it is assumed that the coordinate axis of the binocular camera 4 is the same as the moving vehicle coordinate axis.

That is, the vehicle front traveling direction is the x axis, the vehicle upper direction is the z axis, and the vehicle side direction is the y axis.

Meanwhile, although the coordinate axis of the 3D laser scanner 1 varies depending on the manufacturer of the 3D laser scanner, in the embodiment of the present invention, it is assumed that the coordinate axis coincides with the moving vehicle coordinate axis.

The sensor fusion and world modeling unit 41 in the data processing computer 5 transfers the data received from the 3D and 2D laser scanners 1, 2, 3 or the binocular camera 4 as described below. After converting to a coordinate system and generating a laser scanner world model and a binocular camera world model, a fused world model is generated by fusing the generated world model coordinate data.

In this case, the data received from the binocular camera 4 may be used to generate distance data only when the obstacle resolution is high but the left and right cameras match.

In addition, the person and obstacle determination unit 42 calculates the position of the object and the size of the obstacle from the current moving vehicle by using the fusion world model generated by the sensor fusion and world modeling unit 42 and whether or not the object collides with each other. Generate an estimated time.

 5A and 5B illustrate a concept of a world modeling method used in an embodiment of the present invention.

은 다음 수학식1과 같다.FIG. 5A illustrates a process of converting a sensor coordinate system from a 3D laser scanner or a binocular camera into a driving vehicle center coordinate system.

Is the same as Equation 1 below.

의 크기를 R_L,

의 크기를 R_C라 할때 장애물 좌표 (x₀,y₀, z₀)는 다음 수학식2와 같다therefore,

The size of R _L ,

Obstacle coordinates to the size of the R _C d _{(x 0, y 0, z} 0) is equal to the following expression (2)

Each of these coordinates is represented as a world model through unit sampling.

5B illustrates a method of representing a spatial world model with data received from each sensor.

From the coordinate-converted data 50 obtained using the method as described above, a sampling space 51 is generated in units of 5 × 5 cm with respect to the xy plane among the spatial coordinates represented by x, y and z, and the center point thereof. Is designated by the sampling pointer (5) of the xy coordinate.

Further, the highest altitude data of the coordinate-converted data 50 is obtained as the z-coordinate of the sampling pointer to obtain the sampling pointer 53 of the xyz coordinate.

The sampling pointers 53 of the xyz coordinates acquired in this manner have the representative coordinates of the corresponding sampling space.

Since the present invention has an object in determining objects such as people and obstacles on a track, expressing altitude data in a unit world model of a cuboid as shown in FIG. 5B is sufficient to achieve the object of the present invention. It will be apparent to those skilled in the art that the method of connecting the sampling pointer sets to form a spatial shape is not limited thereto.

In the preferred embodiment of the present invention, considering the characteristics of each sensor as described above and the point that the track vehicle travels on the track, based on the high data output rate and the distance accuracy of the remote (2) and near (2) 2D laser scanners, By utilizing sensor fusion with the 3D laser scanner 1 and / or the binocular camera 4, it is possible to obtain accurate shapes of near and far people and sudden obstacles.

Hereinafter, the sensor fusion employed in the present invention will be described with reference to FIGS. 6 to 8.

6A to 6C show a conceptual diagram of the fusion of the 2D laser scanner 3 and the 3D laser scanner 1 for short range detection according to an embodiment of the present invention.

6A shows the observation range 62 of the 2D laser scanner 3 for near field detection and the observation range 61 of the 3D laser scanner 1 when a person appears in front of the track.

Although the 3D laser scanner 1 is capable of accurate detection at long distances and near distances, the 3D laser scanner 1 has a low data output rate, which is used for near-field detection, especially when a person suddenly appears in front of the track. Is inefficient.

Therefore, in the present embodiment, as shown in FIG. 6B, the 3D laser scanner 1 observes a distance of 30 m or more, and the 2D laser scanner 3 for short distance detection observes a short distance of 30 m or less to integrate the far and near data. have.

In FIG. 6B, reference numeral 63 denotes a detection image by the 3D laser scanner 1, reference numeral 64 denotes a detection image by the 2D laser scanner 3 for near field detection, and reference numeral 66 denotes a 2D laser scanner 3 for near field detection. ) And the result of data fusion of the 3D laser scanner 1.

FIG. 6C is a view showing the generation result of the world model as described above using the data generated through the fusion of the 2D laser scanner 3 for short range detection and the 3D laser scanner 1.

7A to 7C are conceptual views illustrating a fusion of a 2D laser scanner 3 for short range detection and a binocular camera 4 according to another embodiment of the present invention.

Similarly with reference to FIGS. 6A to 6C, in FIG. 7A, the viewing range 72 of the near-field detection 2D laser scanner 3 and the binocular camera 4 when a person appears in front of the track ( FIG. 7B shows a detection image 73 of the binocular camera 4, a near detection image 74 of 30 m or less of the 2D laser scanner 3, a binocular camera 4, and a 2D laser scanner 3. Image 76 representing the result of data fusion of the "

FIG. 7C is a diagram illustrating the generation result of the world model as described above using data generated through the fusion of the near-field detection 2D laser scanner 3 and the binocular camera 4.

8A to 8C show a conceptual diagram of the fusion of a 2D laser scanner 3, a binocular camera 4 and a 3D laser scanner 1 for near field detection according to another embodiment of the present invention.

Similarly as described with reference to Figs. 6A to 6C, Fig. 8A shows the observation range 82 of the near-field detection 2D laser scanner 3, the 3D laser scanner 1 and the binocular when a person appears in front of the track. The observation range 81 of the camera 4 is shown.

8B shows a near-field detection image 83 of 30 m or less of the binocular camera 4, a far-field detection image 85 of 30 m or more of the 3D laser scanner 1, and a near-field of 30 m or less of the 2D laser scanner 3 for near field detection. The detection image 87 and the image 88 showing the result of the data fusion of the binocular camera 4 and the 2D laser scanner 3 are shown.

However, as described above, although the binocular camera has a high obstacle resolution, it is possible to generate a fine world model. However, when the left and right matching ratio is not high, the binocular camera 4 as shown by 84 in FIG. 8B is used. This may result in mismatched parts.

In general, however, since the 3D laser scanner 1 may generate a world model over the entire area, although the detail of the world model may be lower than that of the binocular camera 4, according to the present embodiment, the 3D laser scanner 1 Even if the data of the part not matched by 4) is lost, it can be supplemented by the data of the corresponding part 86 of the 3D laser scanner 1, so that the accurate detection of the object is possible. It can be created.

On the other hand, the detailed description of the person and the obstacle determination performed by the person and obstacle determination unit 42 in the data processing computer 4 is as follows.

The object (obstacle or person) is determined based on the world model generated through the sensor fusion and the world modeling unit 41 described above.

The standard for determining the object is set to 30 cm or more in height and the width is considered in consideration of the resolution of each sensor, but is preferably about 20 cm or more.

In addition, when the detected position of the object is more than 30m from the head of the tracked vehicle, it is determined as a collision possible object, and if it is less than 30m, it is determined as a collision and accidental object.

As described above, when a predetermined output signal is generated from the person and obstacle determination unit 42 by determining the person and the obstacle, the control signal generator 43 in the data processing computer 4 controls the safety device C such as a lifesaving device. Generate a signal for

That is, when it is determined by the person and obstacle determination unit 42 that the detected object is determined to be a collision or accident, the control signal generator generates an operation signal to the known safety device C, and detects the object. Is within the range of 30m to 50m of the tracked vehicle, it is possible to generate a safety signal such as an alarm sound or to generate a safety device (C) operation signal after a predetermined time (t).

Here, the predetermined time t may be obtained by using Equation 3 below.

t (s) = [distance to detected object (m)] / [track vehicle speed (m / s)]

Meanwhile, in the above, the safety device C includes a lifesaving device such as a known air bag and a collision warning device, and thus, detailed drawings, operating principles, and operating states thereof will be omitted.

Although preferred embodiments of the present invention have been described above, it will be apparent to those skilled in the art that various modifications are possible without departing from the spirit and scope of the present invention.

According to the present invention, by providing a sensing system capable of acquiring accurate shape information and distance information of long distance and near people and obstacles in front of a track vehicle in a track vehicle traveling at high speed, It can be used to provide accurate emergency situations and accurate crash standards, helping to early detection of people and obstacles on the track, and ensuring that safety devices such as airbags and crash alarms are activated in a timely manner, significantly reducing personal injury and fatality rates. This can be an effect.

Claims (6)

In the system for detecting people and obstacles on the track that the tracked vehicle, Short-range sensing data generating means installed at a lower end of the head of the track vehicle to generate short-range sensing data in front of the track; Remote sensing data generating means installed at an upper end of the track vehicle and generating remote sensing data in front of the track; By calculating the position and distance information of the object by fusing the sensed data from the short range sensing data generating means and the remote sensing data generating means, and determining whether the object will collide with the tracked vehicle from the calculated object information. And a data processing computer for selectively generating a drive signal with a safety device. The method of claim 1, The remote sensing data generating means includes at least one of a 3D laser scanner and a 2D laser scanner for remote sensing, and the near sensing data generating means includes at least one of a 2D laser scanner and a binocular camera for near sensing. On-track people and obstacle detection systems in driving tracked vehicles. The method according to claim 1 or 2, The data processing computer, A sensor fusion and world modeling unit for performing data sampling based on the data measured from the remote sensing data generating unit and the data measured from the short range sensing data generating unit, and expressing the sampled data set as a world model; And a person and an obstacle determining unit determining a person and an obstacle based on the world model generated by the sensor fusion and the world modeling unit. The method of claim 3, wherein The data processing computer further includes a control signal generator for generating a driving control signal to a safety device in response to a result determined by the person and the obstacle determining unit. . The method of claim 2, A LAN interface is installed between the 3D laser scanner and the data processing computer, an RS-422 interface is installed between the 2D laser scanner and the data processing computer, and a USB or IEEE 1394 is provided between the binocular camera and the data processing computer. Human and obstacle detection system on a track in a traveling track vehicle characterized in that the interface is installed. The method of claim 1, The short range and the long range in the short range sensing data generating means and the long range sensing data generating means are determined based on the traveling speed of the tracked vehicle and the observable range of the short range sensing data generating means and the long range sensing data generating means. A system for detecting people and obstacles on a track in a traveling track vehicle.

Images