JP2013228914A - Railroad crossing detecting device and method for detecting railroad crossing - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a railroad crossing detecting device, even when contrast is lowered due to sunshine, capable of accurately detecting a railroad crossing, and a method thereof.SOLUTION: The railroad crossing detecting device includes: an image pickup unit 10 that photographs a prescribed area and outputs an image; a candidate area extraction unit 21 that extracts a plurality of areas with prescribed color components from the photographed image; a combination creation unit 22 that selects two areas from the plurality of extracted areas and creates one or more combinations; a separation degree evaluation unit 23 that calculates a separation degree of each created combination with respect to a luminance distribution; and a railroad crossing determination unit 24 that determines whether or not an object is a railroad crossing on the basis of the calculated separation degree.

Description

  The present invention relates to a crossing detection device and a crossing detection method for detecting a state of a crossing in front of a vehicle by image processing.

  Conventionally, as a crossing detection device, a device that detects a crossing by performing image processing on image data obtained by imaging a state of a crossing in front of the vehicle is disclosed, and Patent Document 1 is disclosed.

  The image processing apparatus described in Patent Document 1 is an image processing apparatus mounted on a moving body, and continuously captures a scene outside the moving body and continuously forms an image corresponding to the scene. An image capturing unit that processes the image formed by the image capturing unit, continuously extracts a predetermined feature in the scene, and transmits the feature extracted by the feature extracting unit to an external device A transmission unit; and a recording unit that records the feature extracted by the feature extraction unit.

  According to this image processing apparatus, feature information including, for example, a crossing alarm device is extracted from an intermittently captured image, and the extracted feature information is recorded, so that an accurate road condition is recorded with a small capacity. be able to.

JP-A-11-213137

  However, when the contrast is reduced by sunshine, the image processing apparatus described in Patent Document 1 described above cannot accurately separate the neon sign and the like from the level crossing.

  Therefore, the present invention has been proposed in view of the above-described circumstances, and it is an object to provide a crossing detection device and a crossing detection method capable of accurately detecting a crossing even when the contrast is reduced due to sunlight. And

  The present invention includes an imaging unit that captures a predetermined region and outputs an image, a candidate region extraction unit that extracts a plurality of regions having a predetermined color component from the captured image, and two regions from the extracted plurality of regions A combination creating unit that creates one or more combinations by selecting the; a separation degree evaluating unit that calculates a separation degree with respect to the luminance distribution of each created combination; and an object at the level crossing based on the calculated separation degree A crossing determination unit that determines whether or not there is a crossing.

  According to the present invention, two regions are selected from a plurality of extracted regions to create one or more combinations, the degree of separation for the luminance distribution of each created combination is calculated, and based on the calculated degree of separation Thus, it is determined whether or not the object is a level crossing, so that even when the contrast is reduced due to sunlight, the level crossing can be accurately detected.

It is a figure showing an example of vehicles carrying a railroad crossing detection device concerning a 1st embodiment of the present invention. It is a block diagram which shows the internal structure of the crossing detection apparatus which concerns on 1st Embodiment of this invention. It is a flowchart which shows the process sequence by the candidate area | region extraction part of the crossing detection apparatus which concerns on 1st Embodiment of this invention. It is a flowchart which shows the process sequence by the isolation | separation degree evaluation part and level crossing determination part of the level crossing detection apparatus which concern on 1st Embodiment of this invention. It is a flowchart which shows the process sequence by the candidate area | region extraction part of the crossing detection apparatus which concerns on 2nd Embodiment of this invention.

  Hereinafter, a crossing detection device and a crossing detection method according to first and second embodiments to which the present invention is applied will be described with reference to the drawings.

[First Embodiment]
[Configuration of level crossing detection device]
FIG. 1 is a view showing an example of a vehicle equipped with a crossing detection device according to a first embodiment of the present invention. As shown in FIG. 1, the crossing detection device is mounted on a vehicle 1 and includes an imaging unit 10 and a computer 20 and detects a state of a crossing in front of the vehicle.

  The imaging unit 10 captures the state of a predetermined area in front of the vehicle and outputs an image. The state of the predetermined area indicates whether or not the railroad crossing alarm is turned on, that is, whether or not the train is approaching.

  The imaging unit 10 includes a CMOS (Complementary Metal Oxide Semiconductor) or CCD (Charge Couple Device) imaging device, and is configured by a color camera that can acquire a color image (color information) expressed in an RGB color space. be able to. Moreover, the imaging part 10 is attached to the back side of the rear-view mirror of the vehicle 1, for example, and is attached horizontally toward the advancing direction of the vehicle 1.

  FIG. 2 is a block diagram showing an internal configuration of the crossing detection device according to the first embodiment of the present invention. As illustrated in FIG. 2, the computer 20 includes a candidate area extraction unit 21, a combination creation unit 22, a separation degree evaluation unit 23, and a crossing determination unit 24. Specifically, the computer 20 includes a central processing unit (CPU) (not shown) and a memory, and the candidate area extraction unit 21 and the combination creation unit 22 are executed when the CPU executes a crossing detection program stored in the memory. Each process of the separation degree evaluation unit 23 and the level crossing determination unit 24 can be executed.

  In FIG. 2, the candidate area extraction unit 21 extracts a plurality of areas having a predetermined color component from the image captured by the imaging unit 10. The combination creating unit 22 creates one or more combinations of the selected two regions by selecting two regions from the plurality of regions extracted by the candidate region extracting unit 21. The separation degree evaluation unit 23 calculates the separation degree with respect to the luminance distribution of each combination created by the combination creation unit 22. The level crossing determination unit 24 determines whether the object is a level crossing based on the degree of separation calculated by the degree of separation evaluation unit 23.

[Candidate area extraction processing]
Next, the details of the candidate area extraction processing by the candidate area extraction unit 21 will be described with reference to the flowchart shown in FIG.

  First, in step S100, the candidate area extraction unit 21 converts the color space of the color image input from the imaging unit 10 from RGB to HSV (Hue (hue), Saturation (saturation), Value (lightness)). The HSV color space is a color space that expresses lightness and hue separately, and can express a color without being affected by brightness.

  Next, in step S <b> 101, the candidate area extraction unit 21 extracts an area whose hue corresponds to “red” from the HSV color space image. Here, as a condition of “red”, Non-Patent Document 1 (Toshiya Umezawa, Kazuhiko Eguchi: Research on road traffic signal recognition technology from night video, Aichi Institute of Technology Research Report No. 43B 2008) The value corresponding to the “red” of the traffic light described can be referred to.

  According to Non-Patent Document 1, for “red”, the hue H is 0 <H ≦ 10. However, since the area where the brightness is too dark is not suitable as the lighting color, the area of S <0.3 or V <120 is excluded from the areas to be extracted. Note that the set values of the hue H, the hue S, and the lightness V should be appropriately set experimentally, and are not limited to the above-described values.

  Next, in step S102, the candidate area extraction unit 21 individually recognizes the areas extracted in step S301 by the labeling process. Specifically, a process of assigning the same number to pixels in which color portions (or portions having no color) are continuous can be performed to individually identify color regions.

  Next, in step S103, the candidate area extraction unit 21 examines the area of each area detected by the labeling process in step S102, determines whether the area of each area is less than a predetermined value, and determines the area of each area. Is less than the predetermined value, the area is excluded from the areas to be extracted.

  The predetermined value is, for example, the area on the image of the lighting area of the traffic light 20 meters ahead. Since the vehicle always stops once before the crossing, it is only necessary to detect the state of the crossing once while the vehicle is stopped. Therefore, it is sufficient that the predetermined value is 20 m, and if the detection range is expanded more than necessary, it may cause an increase in false detection of an area to be extracted.

  Next, in step S104, the candidate area extraction unit 21 obtains a ratio between the vertical and horizontal lengths of each area, and extracts only areas where the ratio satisfies 0.7 <ratio <1.3. That is, the shape of the region is determined. Here, in order to reduce the processing load, the vertical and horizontal lengths are the difference between the minimum value and the maximum value of the x coordinate and the difference between the minimum value and the maximum value of the y coordinate, respectively. The candidate area extraction unit 21 selects a candidate area having a substantially circular shape having a predetermined color component when the ratio of the vertical and horizontal lengths of each area is 0.7 <ratio <1.3. Can be extracted.

  In addition, the candidate area extraction unit 21 can also extract, as a candidate area, an area that is substantially circular and has an area equal to or larger than a predetermined value.

  Next, in step S105, the candidate area extraction unit 21 stores the candidate area extracted in the current process in a memory for a predetermined time. In step S106, the candidate area extraction unit 21 outputs candidate areas stored in the memory for a predetermined time. In addition, the candidate area extraction unit 21 can also set, as a candidate area, a union of a plurality of areas extracted within a predetermined time, that is, a logical sum of a plurality of areas.

  The combination creating unit 22 performs processing for selecting two related regions from the plurality of candidate regions output from the candidate region extracting unit 21 and creating one combination for all of the plurality of candidate regions. Create one or more combinations. Here, the combination creating unit 22 sets candidate areas separated by a predetermined distance or less from each other as a combination target, and excludes candidate areas separated by a predetermined distance from the combination target.

  In addition, the combination creating unit 22 sets candidate areas that are arranged substantially horizontally or vertically as a combination target, and excludes candidate areas that are not arranged substantially horizontally or substantially vertically from the combination target. Whether or not the two regions are in a substantially horizontal position is determined by whether or not the coordinate values of the y coordinate of, for example, the center pixel in the region are substantially the same. Whether or not the two regions are in a substantially vertical position is determined by whether or not the coordinate values of the x coordinate of the central pixel in the region are substantially the same.

  Then, the two regions combined by the combination creating unit 22 are denoted as a region R1 and a region R2, respectively.

[Separation degree evaluation process and level crossing determination process]
Next, the details of the processes of the degree-of-separation evaluation unit 23 and the crossing determination unit 24 will be described with reference to the flowchart of FIG.

  In step S200, the degree-of-separation evaluation unit 23 acquires images at the same time in the region R1 and the region R2 combined by the combination creation unit 22. In subsequent steps S201 to S04, processing is performed on the images of the region R1 and the region R2 at the same time.

  In step S201, the separation degree evaluation unit 23 calculates the average luminance value μ1 of the region R1 and the average luminance value μ2 of the region R2. Further, in step S202, the degree-of-separation evaluation unit 23 calculates the luminance variance σ1 of the region R1 and the luminance variance σ2 of the region R2.

  Next, in step S203, the separation degree evaluation unit 23 uses the average luminance value μ1 of the region R1, the average luminance value μ2 of the region R2, the luminance variance σ1 of the region R1, and the luminance variance σ2 of the region R2. The degree of separation s is calculated. The degree of separation s is a value calculated based on the ratio between the luminance difference between the region R1 and the region R2 and the sum of the luminance distributions of the region R1 and the region R2, and the degree of separation s is expressed by Equation (1). Can be calculated.

Degree of separation s = inter-class variance / total variance = σ _B ² / σ _T ² (1)
Here, the inter-class variance σ _B ² , the intra-class variance σ _W ² , and the total variance σ _T ² can be expressed by Expressions (2) to (4).

σ _W ² = (N ₁ × σ ₁ ² + N ₂ × σ ^{2 2} ) / (N ₁ + N ₂ ) (2)
However, _{N 1} is the number of pixels in the region R1, _{N 2} is the number of pixels in the region R2, .sigma.1 ² is the variance of the luminance of the region R1, .sigma. @ 2 ² is the variance of the luminance of the region R2.

σ _B ² = {N ₁ × (μ ₁ −μ _T ) ² + N ₂ × (μ ₂ −μ _T ) ² } / (N ₁ + N ₂ ) (3)
However, .mu.1 the average luminance value of the region R1, .mu.2 average luminance value of the region R2, the mu _T is the mean luminance value of the region of the combined region R1 and the region R2.

σ _T ² = σ _B ² + σ _W ² (4)
The degree-of-separation evaluation unit 23 calculates the degree of separation s for each pixel of the target image based on the equations (1) to (4).

Here, as shown in Expression (3), the inter-class variance σ _B ² is a value that changes in accordance with the luminance difference between the region R1 and the region R2, and the average luminance value μ1 of each region R1, R2 , a larger value μ2 is away from the average luminance value mu _T of all the pixels. The total variance σ _T ² is a value that changes according to the sum of the luminance distributions of the regions R1 and R2, and the smaller the variances σ1 and σ2 of the regions R1 and R2, that is, the luminance distribution of the regions R1 and R2 is. The smaller the value, the more uniform.

  Accordingly, the degree of separation s is such that the average luminance value μ1 of the region R1 and the average luminance value μ2 of the region R2 are more distant and the variances σ1 and σ2 of the two are smaller, that is, the luminance distribution of each region R1 and R2. The more uniform the value, the larger the evaluation value.

 Since the degree of separation s is a value between 0 and 1, it is possible to detect a level crossing by setting one threshold without setting various thresholds depending on conditions.

As described above, according to the railroad crossing detection apparatus according to the first embodiment, even if the solar radiation amount changes from a time zone with a large amount of solar radiation such as daytime to a time zone with a small amount of solar radiation such as morning and evening, the solar radiation condition changes. A railroad crossing can be detected without being affected. In a time zone where the amount of solar radiation is small, such as in the morning and evening, the entire image becomes dark and the luminance distribution is shifted overall in the direction of lower luminance values. That is, the average luminance value .mu.1, the value of μ2 is reduced, but since the difference between .mu.1 and μ2 unchanged, interclass variance sigma _B ² does not change. The dispersibility .sigma.1, since hardly changes only involve some change .sigma. @ 2, does not change the total variance sigma _T ^2.

 Accordingly, since the same value is calculated as the degree of separation s even if the amount of solar radiation changes, it is possible to detect a crossing robustly against changes in the solar radiation conditions.

  Next, in step S204, the level crossing determination unit 24 performs threshold processing on the degree of separation s calculated by the degree of separation evaluation unit 23 using a preset threshold value θ. The level crossing determination unit 24 determines that the target area is a level crossing when the degree of separation s is equal to or greater than a preset threshold θ, and does not determine the target area as a level crossing when the degree of separation s is less than the threshold θ.

[Effect of the first embodiment]
Thus, according to the railroad crossing detection apparatus according to the first embodiment, one region is selected from a plurality of extracted regions to create one or more combinations, and the degree of separation with respect to the luminance distribution of each created combination And whether or not the object is a crossing is determined based on the calculated degree of separation. In other words, since the lighting state of the crossing alarm is determined based on the degree of separation, even if the contrast is lowered in the time when the amount of solar radiation is low, such as in the morning and evening, it is affected by changes in the solar radiation conditions. Without this, it is possible to detect a railroad crossing alarm.

  In addition, since a region having a color component corresponding to the lighting color of the railroad crossing alarm device, for example, a “red” region that is the lighting color, is extracted, the processing target can be easily limited.

  In addition, since the union of a plurality of areas extracted within a predetermined time is set as a candidate area, the alarm area can be set regardless of the current on / off status of the crossing alarm that is alternately (exclusively) lit. Can be extracted.

  In addition, since a region having a predetermined color component having a substantially circular shape and an area of the region having a predetermined area or more is extracted as a candidate region, erroneous detection of the candidate region can be reduced.

  In addition, since only candidate areas located within a predetermined distance are combined, erroneous detection of candidate areas can be reduced, and the processing cost can be reduced.

  In addition, since the alarm devices that are alternately lit are installed approximately horizontally or approximately vertically, the candidate regions are limited by detecting regions that are positioned approximately horizontally or approximately vertically as candidate regions. False detection can be reduced.

  In addition, since the degree of separation of the combination of regions with respect to the luminance distribution at the same time is calculated, it is possible to detect a state where one of the two candidate regions combined at the same time is turned off and the other region is turned on. Therefore, it is possible to accurately determine the alarm devices that are alternately turned on.

[Second Embodiment]
[Configuration of Second Embodiment]
The crossing detection device according to the second embodiment shows an example used in a situation where the vehicle is not temporarily stopped before the crossing. The crossing detection device according to the second embodiment corresponds to a country such as the United States where the vehicle is not temporarily stopped before the crossing.

  In addition, since most of the crossing detection apparatus according to the second embodiment is common to the crossing detection apparatus according to the first embodiment, only the configuration different from both will be described here.

[Candidate area extraction processing]
FIG. 5 is a flowchart showing a processing procedure by the candidate area extraction unit of the level crossing detection apparatus according to the second embodiment of the present invention. The flowchart of the level crossing detection apparatus according to the second embodiment shown in FIG. 5 differs from the flowchart of the level crossing detection apparatus according to the first embodiment shown in FIG. 3 only in the process of step S305. Steps S300 to S304 and S306 are the same as the processes of steps S100 to S104 and S106, and thus description thereof is omitted. Here, only the process of step S305 will be described.

  In step S305, the candidate area extraction unit 21 associates the candidate area extracted from the previously captured image with the candidate area extracted from the currently captured image.

[Effects of Second Embodiment]
According to the railroad crossing detection apparatus according to the second embodiment, since the candidate area extraction unit 21 in which the previous candidate area and the current candidate area are associated with each other is provided, when the vehicle moves, Even when the position of the alarm at the level crossing changes, it is possible to reduce erroneous detection of candidate areas.

  The candidate areas are associated by selecting an area having the closest position, shape, and size of the area. Next, the candidate area extraction unit 21 stores the associated area for a predetermined time.

DESCRIPTION OF SYMBOLS 10 Imaging part 20 Computer 21 Candidate area extraction part 22 Combination preparation part 23 Separation degree evaluation part 24 Crossing determination part

Claims (9)

Imaging means for imaging a predetermined area and outputting an image;
Candidate area extraction means for extracting a plurality of areas having a predetermined color component from the image captured by the imaging means;
A combination creating unit that creates two or more combinations by selecting two regions from the plurality of regions extracted by the candidate region extracting unit;
A degree of separation evaluation means for calculating a degree of separation for the luminance distribution of each combination created by the combination creation means;
A crossing determination means for determining whether the object is a crossing based on the degree of separation calculated by the degree of separation evaluation means;
A crossing detection device characterized by comprising:   2. The crossing detection device according to claim 1, wherein the candidate area extraction unit extracts a region having the predetermined color component corresponding to a lighting color of a crossing alarm device.   The crossing detection device according to claim 1 or 2, wherein the candidate area extraction unit extracts a union of areas extracted within a predetermined time as a candidate area.   The candidate area extracting means extracts, as a candidate area, an area in which the shape of the area having the predetermined color component is substantially circular and the area of the area having the predetermined color component is equal to or greater than a predetermined value. The crossing detection device according to any one of claims 1 to 3.   The crossing detection device according to any one of claims 1 to 4, wherein the combination creating unit combines only candidate regions located within a predetermined distance.   The crossing detection device according to any one of claims 1 to 5, wherein the combination creating unit combines candidate regions arranged substantially horizontally or vertically.   The said isolation | separation degree evaluation means calculates the isolation | separation degree with respect to the luminance distribution of the same time of the 1 or more combination produced by the said combination production | generation means, The any one of Claim 1 thru | or 6 characterized by the above-mentioned. Railroad crossing detection device.   The candidate area extraction unit associates a candidate area extracted from a previously captured image with a candidate area extracted from a currently captured image. A crossing detection device according to item 1. An imaging step of imaging a predetermined area by an imaging means and outputting an image;
A candidate region extraction step of extracting a plurality of regions having a predetermined color component from the captured image;
A combination creating step of creating one or more combinations by selecting two regions from the plurality of extracted regions;
A degree-of-separation evaluation step for calculating a degree of separation for the luminance distribution of each created combination;
A crossing determination step for determining whether the object is a crossing based on the calculated degree of separation;
A crossing detection method characterized by comprising:

Images