JP4920073B2 - Method and apparatus for detecting head position of traveling body - Google Patents

Abstract

PURPOSE: A method and device for detecting the head position of a running vehicle capable of high-speed processing is provided to accurately detect the head position of a running vehicle by reducing the affecting factors like light or shadow. CONSTITUTION: A method for detecting the head position of a running vehicle is as follows. An image of head part and background of a running vehicle are taken with an image sensor(S11). The differential picture between is produced based on the taken images(S12). The luminance integration according to the fixed direction of the image is performed according to each image. The projecting brightness distribution is produced. The differential projecting brightness distribution is calculated by using the calculated the difference of two projecting brightness distribution(S13). The location of the head part of the running vehicle is recognized(S16).

Description

  The present invention relates to a technology for detecting the leading position of a traveling body such as a train or an automobile, and more particularly to a technology for detecting the leading position of a traveling body using an image processing technique.

In recent years, trains and subways on the Shinkansen and conventional lines have been equipped with movable fences as one of the safety measures for the purpose of preventing falls from the platform and contact accidents with the train.
As shown in FIG. 1, the movable fence is a partition wall provided so as to isolate a train that stops or passes at the end of the station platform and a passenger on the station platform, and corresponds to a door portion of the stopped train. At the position, a door for the passenger to go up and down is provided. The opening and closing of the door is usually performed based on information from an automatic train operating device (ATO), but it may be operated by a crew member in the event of an abnormality.

However, when installing the above equipment in a station, there is concern that the installation of equipment, such as a new ATO, a management system for information transmitted from those systems, and the like will be enormous. The
Therefore, the applicants of the present application have already developed a technique such as that of Patent Document 1.
Patent Document 1 is a position detection device for detecting the position of a moving body, and is an image sensor installed so as to be able to capture an image of the moving body together with an image of a background of the moving body, and the image sensor. The position (especially the head position) of the moving body is detected by comparing the base image that is an image of the moving body that is captured by the method and the detection image that is an image that includes the moving body. And a position detecting unit including the position detecting unit.

  As described above, opening and closing of the door of the movable fence is mainly performed based on information from the automatic train operation device (ATO), but instead, the train of the train obtained from the position detection device of Patent Document 1 is used. By using the leading position for opening and closing the movable fence, it is possible to construct a movable fence system with a simple method and with reduced installation costs.

JP 2008-298501 A

However, the technique described in Patent Literature 1 is a technique for detecting the head position of a train by comparing a base image in which a train that is a moving body is not captured and a detection image in which the train is captured. Therefore, if there is something that changes in the background part of the image instead of the train itself, that part becomes a disturbance and may cause false detection.
Moreover, the direct reflected light (regular reflected light) of the front illumination light to which the train is attached, the shadow of the train itself, etc. are disturbance light components, and the position of the luminance change moves on the image as the train moves. Such disturbance light components are also different from the base image in the technique described in Japanese Patent Laid-Open No. 2004-133830, and thus produce a clear edge. In addition, since the edge moves together with the moving body, an error is likely to occur in the position detection of the moving body.

Furthermore, as the train approaches the visual field of the image sensor, the luminance distribution of the entire background gradually changes due to the front illumination light at the train head. That is, the luminance gradually changes from the base image in the background portion where the moving object does not exist, and in the technique described in Patent Document 1, a false edge is generated and causes false detection.
That is, in the case of detecting the head position of a train, the technique of Patent Document 1 may not be able to cope with it, and under some circumstances, the head position of the traveling body may be erroneously detected.

  SUMMARY OF THE INVENTION In view of the above problems, the present invention has an object to provide a method and an apparatus for detecting the leading position of a traveling body with as few false detections as possible due to disturbances such as light emitted from the traveling body and shadows of the traveling body. And

In order to achieve the above-described object, the present invention takes the following technical means.
In other words, the traveling body leading position detection method according to the present invention is a detection method for detecting the leading position of the traveling body using image processing, and continuously captures the leading portion and background of the traveling body with an image sensor. Then, based on the captured image, a difference image between two consecutive images or a difference image between two consecutive images is calculated, and the calculated difference image is along a predetermined direction on the image. A difference projection luminance distribution is calculated by performing luminance integration, and a predetermined luminance threshold is applied to the calculated difference projection luminance distribution to recognize the position of the leading portion of the traveling body. .

  Further, another detection method according to the present invention is a detection method for detecting a leading position of a traveling body using image processing, wherein the leading portion and background of the traveling body are continuously captured by an image sensor, The projected luminance distribution is calculated by performing luminance integration along a predetermined direction on the image in each of the two images or two images that are intermittently connected, and the difference between the two calculated projected luminance distributions is calculated. The difference projection luminance distribution obtained is calculated, and a predetermined luminance threshold is applied to the calculated difference projection luminance distribution, thereby recognizing the position of the leading portion of the traveling body.

When performing the luminance integration, it is preferable that the predetermined direction on the image is a direction perpendicular to the traveling direction of the traveling body.
Preferably, the luminance threshold value is a variable value according to the speed of the traveling body.
In addition, in order to eliminate erroneous detection due to a disturbance luminance value existing in the captured image, it is preferable to preliminarily define an application range of a luminance threshold for the distribution waveform.

Note that the disturbance luminance value is a directly reflected light (regular reflected light) of light emitted from the traveling body, a shadow of the traveling body itself, a luminance value that does not actually exist in the image caused by electrical noise, etc. Is mentioned.
Preferably, the application range is set based on movement information of the traveling body and has a variable range width according to the speed of the traveling body.

A leading position detection device for a traveling body according to the present invention is a detection device that detects the leading position of a traveling body using image processing, and an image sensor capable of continuously capturing the leading portion and background of the traveling body; And an image processing unit configured to detect the leading position of the traveling body using the traveling body leading position detection method described above.
The most preferable method for detecting the leading position of the traveling object according to the present invention is a detection method for detecting the leading position of the traveling object using image processing, wherein the leading part and the background of the traveling object are image sensors. Based on the captured image, a difference image between two consecutive images or a difference image between two images that are intermittently connected is calculated. A difference projection luminance distribution is calculated by performing luminance integration along a predetermined direction, and a luminance threshold obtained as a variable value according to the speed of the traveling body is applied to the calculated difference projection luminance distribution. Then, there is a method for recognizing the position of the leading portion of the traveling body.
Also, a detection method for detecting a leading position of a traveling object using image processing, wherein the leading portion and background of the traveling object are continuously captured by an image sensor, and each of two consecutive images or intermittently. In each of the two consecutive images, the projection luminance distribution is calculated by performing luminance integration along a predetermined direction on the image, and a difference projection luminance distribution obtained by taking the difference between the two calculated projection luminance distributions is calculated. By applying a brightness threshold obtained as a variable value according to the speed of the traveling body to the calculated difference projection brightness distribution, the position of the leading portion of the traveling body is recognized.
A method for detecting the head position of the traveling body can also be employed.

  According to the head position detection method and apparatus of the traveling body according to the present invention, erroneous detection due to disturbances such as light emitted from the traveling body and shadows of the traveling body is reduced as much as possible, and the leading position of the traveling body is reliably and accurately. Can be detected.

It is a perspective view which shows the station platform provided with the image sensor. It is the schematic of the system comprised with the head position detection apparatus of a traveling body, and a movable fence. It is a figure for demonstrating the image processing performed with the head position detection method of a traveling body. It is the figure which applied the brightness | luminance threshold value to difference projection brightness distribution. It is a flowchart of the head position detection method of a traveling body. It is a figure explaining the idea for estimating the approximate position of the head of a running body. It is the flowchart which showed the other method of the head position detection of a traveling body.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
A traveling body leading position detection device 1 according to the present invention is installed in a station platform 2 such as a Shinkansen, a conventional line, a subway, and the like, and here, assuming a train 3 traveling on a track as a traveling body, The embodiment which detects the position of the head part of the train 3 which enters the home 2 is shown.
[First Embodiment]
FIG. 1 shows a situation of a station platform 2 where a head position detection device 1 (hereinafter also referred to as a detection device) of a traveling body is installed.

  The station platform 2 is provided with a movable fence 4 as one of safety measures for the purpose of preventing a fall from the station platform 2 and a contact accident with the train 3 as a traveling body. The movable fence 4 is a partition wall provided to isolate the train 3 that stops at the station platform 2 or passes through the station platform 2 and passengers on the station platform 2, and corresponds to the door of the stopped train 3. Is provided with a door 4A for the passenger to go up and down. The door 4A is opened and closed based on “information on the leading position of the train 3” detected by the detection device 1.

The traveling body leading position detection device 1 of the present invention detects the leading position of the train 3 using an image processing technique, and the position of the leading portion of the train 3 is a predetermined position (stopped) on the station platform 2. It is detected whether or not the position L) has been reached.
As shown in FIGS. 1 and 2, the detection apparatus 1 includes an image sensor 5 that captures the head of a train 3 that enters and stops at a station platform together with the station platform 2 and the track 10 that are backgrounds. The image sensor 5 is composed of a CCD camera (video camera or the like) and, for example, continuously outputs images at 30 frames per second.

The image sensor 5 is installed at a position about 3 meters above the ground, such as under the roof 6 installed on the station platform 2, on the leading side of the station platform 2. The image sensor 5 is provided with an image pickup lens that can pick up the head of the train 3 and the background thereof, and an automatic exposure mechanism that optimizes the exposure amount according to time and weather.
The picked-up image output from the image sensor 5 is taken into the image processing unit 7 composed of hardware such as a personal computer or a DSP.

In this image processing unit 7, a frame memory is provided, and a two-dimensional image (640 pixels × 480 pixels) captured at a predetermined time interval (1/30 or 1/15 seconds) is accumulated. Image processing for detecting a head position described later is performed on the two-dimensional image stored in the frame memory.
Furthermore, in the case of the present embodiment, it is determined whether or not the head of the train 3 has stopped correctly at the stop position L determined in the station platform 2 based on the detected head position of the train 3. When sent to the door control unit 8, the door control unit 8 issues a signal for opening the door 4 </ b> A to the movable fence 4 when the train 3 is correctly stopped at the stop position L.

Note that the crew may open the door 4 </ b> A of the movable fence 4 based on the detected leading position of the train 3.
By the way, the detection apparatus 1 has the traveling body sensor 9 for detecting the approach of the train 3 on the near side in the traveling direction of the train 3 from the position where the image sensor 5 is installed.
The traveling body sensor 9 detects the presence or absence of an object in front of the sensor, such as an ultrasonic sensor or an infrared sensor. For example, the traveling body sensor 9 is installed under the roof 6 in the center of the station platform 2, and train 3 Before entering the imaging range (field of view) of the image sensor 5 is arranged so as to detect entry of the train 3 to the platform. By causing the image sensor 5 to operate based on the signal from the traveling body sensor 9 (the signal for entering the train 3), the possibility of detecting an approach other than the train 3 and performing a false detection is greatly reduced.

By using a Doppler sensor using ultrasonic waves or infrared rays as the traveling body sensor 9, not only the train 3 but also the speed of the entering train 3 can be obtained.
Next, based on the flowcharts of FIGS. 3, 4, and 5, details of the “position detection processing of the head of the train 3” performed in the image processing unit 7 will be described.
The train 3 head position detection method performed in the image processing unit 7 has an algorithm for detecting the train 3 head position using image processing, and the image sensor 5 continuously detects the train 3 head position and background. By calculating the difference image between two consecutive images based on the captured image, and performing luminance integration along the direction perpendicular to the traveling direction of the train 3 with respect to the calculated difference image. A projected luminance distribution (hereinafter also referred to as a differential projected luminance distribution) is calculated, a position S that begins to exceed a predetermined luminance threshold is detected for the calculated projected luminance distribution, and the detected position S is detected as the train 3. It is recognized as the position of the head part.

Specifically, first, it is assumed that the traveling body sensor 9 detects that the train 3 has entered the station platform 2 in S10 of FIG.
After detecting the approach of the train, in S11, the image sensor 5 continuously captures images of the head portion of the train 3 including the background of the station platform 2 and the track 10 and the like. The captured image is transferred to the image processing unit 7 and stored in a frame memory in the image processing unit 7. The interval of imaging by the image sensor 5 is arbitrary, but 1/30 seconds to 1 second is preferable in order to reliably detect the entrance of the train 3 to the station platform 2.

(A-1) to (a-4) in FIG. 3 are screens in which the train 3 enters the station platform 2 continuously, or screens that are intermittently captured (for example, images skipped by one frame). FIG.
(A-1) is a situation where the traveling body sensor 9 detects the approach of the train 3 but the train 3 does not enter the imaging range of the image sensor 5. (A-2) and (a-3) are images of the situation where the train 3 enters the image sensor 5. The train 3 enters from the left to the right of the image (toward the coordinate origin along the X coordinate of the image). H in the image indicates a portion (regular reflected light H) illuminated by the headlight of the train 3.

The speed of the train 3 decelerates while transitioning from the situation (a-2) to the situation (a-3). (A-4) is an image of the situation in which the train 3 is correctly stopped at the stop position L of the platform. In the transition from the situation (a-3) to (a-4), the speed of the train 3 is very low.
Next, in S12, a difference image is created by taking the difference between the two captured images based on the obtained captured image. Specifically, the luminance value of the pixel existing at the coordinates (x1, y1) on the image (a-1) and the luminance value of the pixel existing at the coordinates (x1, y1) on the image (a-2). A difference image is obtained by calculating the difference from the value for all the pixels. The difference image between (a-1) and (a-2) is (b-1) in FIG. 3, and the leading portion of the train 3 and the regular reflected light H appear in the (b-1) image.

  (B-2) in FIG. 3 is a difference image between the image of (a-2) and the image of (a-1). In this difference image, the top of the train 3 that has advanced to the center of the screen, and this There is specularly reflected light H emitted from the train 3. The regular reflection light H exists at the left end of the image. The main body (right side of the image) following the head of the train 3 is a thin (low brightness) image because there is no significant change between the image (a-2) and the image (a-1). Include. Note that the specularly reflected light H existing in (a-1) is a negative value in the difference image and is not displayed.

(B-3) in FIG. 3 is a difference image between the image of (a-3) and the image of (a-4). In this difference image, only the head part of the train 3 stopped at the stop position L is shown. Exists. In the transition from the situation of (a-3) to (a-4), the speed of the train 3 is very low, so the leading image of the train 3 in (b-3) is very thin (low brightness) It has become a thing.
Thereafter, in S13, for each of the difference images (b-1) to (b-3), a projection luminance distribution is calculated by performing projection in the Y direction. Specifically, the projection luminance distribution can be obtained by adding the luminance values of a plurality of pixels arranged along the Y coordinate at each X coordinate (by integrating the luminance values). The direction along the Y coordinate is a direction perpendicular to the direction in which the train 3 moves in the captured image.

Since the luminance distribution calculated in this way is a projection luminance distribution related to the difference image, it is hereinafter referred to as a “difference projection luminance distribution”.
The difference projection luminance distribution obtained by the above processing is shown in FIGS. 3 (c-1) to 3 (c-3), and the difference images shown in FIGS. 3 (b-1) to 3 (b-3) respectively. Respectively.

  For example, in the difference projection luminance distribution shown in (c-1) of FIG. 3, the convex waveform on the side close to the coordinate origin is due to the regular reflected light H of the headlight and is a disturbance luminance value. The convex waveform on the side far from the origin is the luminance due to the leading portion of the train 3. With respect to the convex waveform caused by the leading portion, the edge thereof is slightly blurred due to blur due to movement of the train 3 or defocus due to the performance limit of the imaging lens. For this reason, the convex waveform at the head part rises slightly before the train head part.

  In (c-2) of FIG. 3, the train 3 enters the station platform 2, and the convex waveform caused by the regular reflected light H and the convex waveform caused by the leading portion of the train 3 move to the origin side. In addition, since the brightness value of the head portion is low due to the decrease in the speed of the train 3, the peak value of the convex waveform at the head portion of the train 3 decreases, and the peak of the convex waveform due to the regular reflected light H decreases. It is close to the value.

  In (c-3) of FIG. 3, since the regular reflection light H from the headlight moves out of the field of view of the image sensor 5, the convex waveform caused by the regular reflection light H disappears, and Only convex waveform. The luminance value at the head portion becomes extremely low due to the speed reduction accompanying the stop of the train 3, and the peak value of the convex waveform at the head portion of the train 3 becomes a small value. In some cases, the peak value of the convex waveform may be lower than the peak value of the convex waveform caused by the regular reflected light H of the headlight.

In order to accurately detect the head position of the train 3 even under such circumstances, as shown in FIG. 4, a “dynamic brightness threshold variable according to the speed of the traveling body” is set for the difference projection brightness distribution. The convex waveform resulting from the head part of the train 3 is detected, and the position of the train 3 on the image is recognized.
Specifically, in S14, the speed of the train 3 that has entered the station platform 2 is detected by using the traveling body sensor 9 and the like, and a brightness threshold value that becomes low brightness when the speed of the train 3 decreases is set. decide. Various methods can be used for determining the brightness threshold. For example, a constant multiple of the speed of the train 3 at the current time is used as the brightness threshold, or a constant multiple of the average speed of the train 3 up to the current time is used. It may be a luminance threshold value. An exponent multiple of the average speed of the train 3 up to the current time may be used as the brightness threshold. 4A to 4C show the same differential projection luminance distributions as FIGS. 3C-1 to 3C-3.

Thereafter, in S15, as shown in FIG. 4A, the luminance threshold value TH1 (high luminance value) is applied to the difference projection luminance distribution. Then, the position P1 at which the luminance threshold TH1 starts to be exceeded becomes the leading portion of the convex waveform caused by the leading portion of the train 3.
In S16, the start position of the train 3 can be recognized by detecting the position where the exceeding starts (the intersection point P1 of the brightness threshold TH1 and the convex waveform). As described above, with respect to the convex waveform at the head part, the edge is slightly blurred due to blur due to movement of the train 3 or out-of-focus due to the performance limit of the imaging lens, and rises slightly before the train head part. Therefore, the intersection of the brightness threshold TH1 and the convex waveform is closer to the accurate train head position.

  However, as is clear from FIG. 4B, the brightness threshold is lowered from TH1 to TH2 as the speed of the train 3 decreases, so that the regular reflection of the headlights rather than the convex waveform caused by the head of the train 3 is achieved. A convex waveform caused by the light H is detected. Therefore, in this embodiment, in S14, the range in which the luminance threshold is applied, in other words, the range in which the convex waveform at the train head is searched for in the difference projection luminance distribution is limited (a window is provided). A method for obtaining the brightness threshold TH (TH1 to TH3) will be described later.

The search range (application range) is indicated by L1 to L3 in FIG. 4. In describing the range, the right side of the X coordinate (on the opposite origin side) is the starting point, and the left side of the X coordinate (the origin side) is the end point. Call.
With respect to the starting point of the application range of the luminance threshold, in the first obtained difference projection luminance distribution (FIG. 4A), the maximum value of the X coordinate of the projection luminance distribution (the rightmost side of the X coordinate) is set. Next, in the obtained differential projection luminance distribution (FIG. 4B), the position of “the intersection P1 of the luminance threshold TH1 and the convex waveform” obtained in FIG. 4A is used as the starting point of the application range of the luminance threshold. Yes. In the differential projection luminance distribution (FIG. 4C) obtained thereafter, the position of the “intersection P2 of the luminance threshold TH2 and the convex waveform” obtained in FIG. 4B is set as the starting point of the application range of the luminance threshold.

As for the end point of the application range, the distance between the start point and the end point is set so as to become shorter as the speed of the train 3 decreases.
For example, in FIG. 4A, the application range is almost the total length L1 = L _max of the projected luminance distribution, whereas in FIG. 4B where the speed of the train 3 is slow, L2 (<L1 In FIG. 4C where the train 3 is almost stopped, L3 (<L2). A method for obtaining the application range L (L1 to L3) will also be described later.

As described above, the convex waveform (disturbance luminance value) due to the regular reflection of the headlight is made outside the detection range by reducing the range to which the luminance threshold for detecting the head of the train 3 is applied according to the decrease in speed. Therefore, erroneous detection due to the disturbance luminance value can be reliably prevented.
By executing the above S10 to S16 in real time (every time an image is taken), the top position of the train 3 on the picked-up image is determined, and then in S17, the top position of the train 3 and the “train” on the image are displayed. It is determined whether or not “3 stop position L” matches. In addition, stop determination calculates the position of the train 3 with respect to the station platform 2 based on the head position of the train 3 on a captured image, and is the position and the actual stop position L on the station platform 2 matched? It may be determined whether or not.

  As shown in FIG. 2, when it is recognized that the train 3 is stopped at a predetermined position, the information is transmitted to the crew, and the crew may open the door 4 </ b> A of the movable fence 4. Is sent to the door control unit 8 provided separately, and the door control unit 8 is movable when “the train 3 has stopped correctly at the stop position L”. A signal for opening the door 4 </ b> A may be issued to the fence 4.

  In summary, the head position detection device 1 of the traveling body according to the present invention has a dynamic brightness threshold and a dynamic application range (window) for the difference projection brightness distribution obtained based on the difference image between frames. By adopting, it is possible to reduce the number of false detections due to changes in the background situation of captured images and disturbances such as light emitted from the train 3 as much as possible, and to detect the leading position of the traveling body reliably and accurately. Become.

When the movable fence 4 laid on the station platform 2 is operated using the leading position of the train 3 accurately determined by the leading position detecting device 1 of the traveling body of the present invention, the safety of the movable fence 4 It can greatly contribute to the further improvement of sex.
Incidentally, the luminance threshold TH (TH1 to TH3) and the application range L (L1 to L3) used when performing the processes of S10 to S17 can be variously determined in accordance with the actual field situation. The determination is based on “sensor information” as follows.

That is, as a method of calculating the application range L (L2), the application range L (L2) is obtained based on a predicted moving amount on the imaging screen calculated from the speed v of the train 3 measured by the traveling body sensor 9. The movement amount prediction value Lv is obtained by Expression (1).

Lv = αvτ (1)

Here, α is a parameter that connects the real space of the train and the position on the screen, and is obtained in advance by calibration (calibration work) when the image sensor 5 is installed. Also, τ is the elapsed time between imaging frames. Many commercially available image sensors 5 have 1/15 to 1/30 seconds, and τ may be set according to the frame rate of the image sensor 5 actually used.

Note that equation (1) assumes constant speed motion of the train 3. However, since there is a possibility that the vehicle is actually accelerating / decelerating, the width is slightly widened and the applicable range L2 is expressed by equation (2). calculate.

L2 = β × Lv (2)

Here, β is a constant. The appropriate amount of β may be obtained from actual measurement data. Note that the application range L2 may be calculated by Expression (3).

L2 = Lv + γ (3)

Here, γ is a constant. The appropriate amount of γ may be obtained from actual measurement data.

Further, as a method for calculating the brightness threshold TH (TH2), it is preferable to use the equation (4) having the speed v of the train 3 measured by the traveling body sensor 9 as a variable.

TH2 = K × (vτ) ^ξ (4)

Here, τ is an elapsed time between imaging frames, and K is a constant term. ξ may be a constant or a speed function ξ (v). If ξ = 1, TH2 is simply proportional to speed.

Note that the value of the brightness threshold TH decreases as the speed of the train 3 decreases, and a noise signal is erroneously detected. Therefore, the minimum value TH3 of the brightness threshold is determined in advance, and when the result of Expression (4) is smaller than TH3, TH3 is adopted as the brightness threshold.
On the other hand, the luminance threshold TH and the application range L can be determined based on “information obtained from an image”.

  FIG. 6 shows a “movement diagram of the leading position of the train 3 that has entered the station platform 2” created based on the image captured by the image sensor 5. In this figure, the horizontal axis is time, and the vertical axis is the head position on the train screen. Since the train 3 decelerates while proceeding from the right to the left on the imaging screen, and eventually stops, in FIG. 6, the leading position of the train 3 moves to the smaller value on the vertical axis as time passes.

If such a movement map (transition diagram) of the head position is used, the movement amount Lv on the screen in the next image frame can be predicted from the change in the previous train head position. For example, it can be predicted as shown in Equation (5).

Lv = Σ _n (P _i −P _i-1 ) / n (5)

Equation (5), the head position P _i obtained by the i-th frame, (i-1) -th obtains a difference between the head position P _i-1 obtained in the frame image, the difference of n This is performed for the section, and is divided by n to obtain an average value of the inter-frame movement amount n times. The search section L2 can be calculated by using this formula (5) instead of the formula (1). Note that the application range L1 to be applied first is L _max as described above.

Further, when the application range L is determined based on “information obtained from an image”, the luminance threshold value TH2 is obtained by substituting Lv obtained in Equation (5) for vτ in Equation (4), Equation (4). It can also be calculated by
[Second Embodiment]
Next, a second embodiment of the head position detection method and apparatus according to the present invention will be described.

The second embodiment is greatly different from the first embodiment in that the procedure for obtaining the difference projection luminance distribution is different.
That is, as shown in FIG. 7, after detecting the train approach by the traveling body sensor 9 (S20), the image sensor 5 continuously displays images of the head part of the train 3 including the background such as the station platform 2 and the track 10. An image is taken (S21). The captured images are (a-1) to (a-4) in FIG.

Next, in S22, a projection luminance distribution is calculated for each of the images shown in (a-1) to (a-4) of FIG. Specifically, the projection luminance distribution can be obtained by integrating the luminance values in the direction along the Y coordinate in each X coordinate (projecting in the Y direction).
Thereafter, in S23, the difference projection luminance distribution is calculated by taking the difference in the projection luminance distribution between the two images. Specifically, a difference projection luminance distribution that is a difference between the projection luminance distribution of FIG. 3A-1 and the projection luminance distribution of FIG. 3A-2 is calculated. This difference projection luminance distribution is substantially the same as the difference projection luminance distribution in the first embodiment shown in FIG.

  Similarly, a difference projection luminance distribution that is a difference between the projection luminance distribution of FIG. 3A-2 and the projection luminance distribution of FIG. 3A-3 is calculated. This difference projection luminance distribution is substantially the same as the difference projection luminance distribution in the first embodiment shown in FIG. A difference projection luminance distribution which is a difference between the projection luminance distribution of FIG. 3A-3 and the projection luminance distribution of FIG. 3A-4 is calculated. This difference projection luminance distribution is substantially the same as the difference projection luminance distribution in the first embodiment shown in FIG.

  The processing of S24 to S27 is performed on the obtained difference projection luminance distribution, but the processing of S24 to S27 is substantially the same as the processing of S14 to S17 of the first embodiment, and thus description thereof is omitted. . The operational effects are also substantially the same as in the first embodiment, and the number of false detections due to changes in the background state of the captured image and disturbances such as light emitted from the train 3 is reduced as much as possible, and the leading position of the traveling body is reliably and accurately Can be detected.

  As an apparatus for realizing the head position detection method of the second embodiment, the same hardware (image sensor 5, image processing unit 7, and traveling body sensor 9) as in the first embodiment can be employed. Note that the image processing of the second embodiment is an algorithm in which the processing amount of a two-dimensional image having a large amount of data is small as a whole because the projection luminance distribution is obtained at the initial stage of the image processing. Therefore, it is a method for detecting the leading position of a traveling body that can perform high-speed processing and has a low calculation load.

By the way, it should be thought that embodiment disclosed this time is an illustration and restrictive at no points. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.
For example, although the train 3 is illustrated as the traveling body, the traveling body may be an automobile, an airplane, or the like, or a transported article in a factory or the like. Although it has been described that the door 4A of the movable fence 4 is opened and closed based on the detection result of the detection device 1, the detection result of the detection device 1 may be used for other purposes without any problem.

  Also, the diagrams used in the description of the embodiment (for example, FIG. 3 and FIG. 4) are conceptual diagrams for accurate explanation, and describe fine noise or steep spikes that usually appear in images and waveforms. It should be noted that it is not. Although the actual captured image has various noises, spikes, and background textures that can be disturbances, they can be removed using existing image processing techniques and affect the essence of the present invention. is not.

DESCRIPTION OF SYMBOLS 1 Head position detection apparatus of traveling body 2 Station platform 3 Train 4 Movable fence 4A Door 5 Image sensor 6 Roof 7 Image processing part 8 Door control part 9 Running body sensor 10 Track H Headlight regular reflection light L Stop position

Claims (6)

A detection method for detecting a leading position of a traveling object using image processing,
The head and background of the traveling body are continuously captured by an image sensor,
Based on the captured image, a difference image between two consecutive images or a difference image between two consecutive images is calculated,
With respect to the calculated difference image, a difference projection luminance distribution is calculated by performing luminance integration along a predetermined direction on the image,
A travel characterized by recognizing the position of the leading portion of the traveling body by applying a brightness threshold obtained as a variable value according to the speed of the traveling body to the calculated difference projection brightness distribution. Body head position detection method. A detection method for detecting a leading position of a traveling object using image processing,
The head and background of the traveling body are continuously captured by an image sensor,
In each of two consecutive images or each of two images that are intermittently connected, a projection luminance distribution is calculated by performing luminance integration along a predetermined direction on the image,
Calculating a difference projection luminance distribution obtained by calculating a difference between the two calculated projection luminance distributions;
A travel characterized by recognizing the position of the leading portion of the traveling body by applying a brightness threshold obtained as a variable value according to the speed of the traveling body to the calculated difference projection brightness distribution. Body head position detection method.   The method for detecting the leading position of a traveling body according to claim 1 or 2, wherein when performing the luminance integration, a predetermined direction on the image is a direction perpendicular to a traveling direction of the traveling body. In order to eliminate erroneous detection due to disturbance luminance values existing in the captured image, in any one of claims 1 to 3, characterized in that in advance define the scope of the luminance threshold value for the difference projection luminance distribution The head position detection method of the described traveling body. 5. The head of the traveling body according to claim 4 , wherein the applicable range is set based on movement information of the traveling body and has a variable range width according to the speed of the traveling body. Position detection method. A detection device for detecting a leading position of a traveling body using image processing,
An image sensor capable of continuously capturing the head and background of the traveling body;
An image processing unit configured to detect the leading position of the traveling body using the traveling body leading position detection method according to any one of claims 1 to 5 ,
A head position detection device for a traveling body, comprising: