JP6139729B1 - Image processing device - Google Patents

Abstract

To detect not only a person but also a thin string-like object in real time.
An image processing apparatus according to an embodiment includes first image acquisition means, first motion detection means, second image acquisition means, second motion detection means, and signal output means. The second image acquisition means acquires a plurality of second images in which only the area around the door is shown with respect to the plurality of images continuously captured by the camera. The second motion detection means compares the plurality of acquired second images in units of blocks, and detects the motion of the object in the area around the door. The signal output means outputs a door opening instruction signal for opening the door of the car when at least one of the first movement detecting means and the second movement detecting means detects a block with movement. Outputs to the car control device that controls the door opening and closing of the car door.
[Selection] Figure 1

Description

  Embodiments described herein relate generally to an image processing apparatus.

  In recent years, various techniques have been devised to prevent people and objects from being caught in elevator car doors. For example, a technique has been devised that detects a user moving toward an elevator using a camera and extends the door opening time of the elevator door. In addition, a technique has been devised that detects a user at a landing using a camera and controls opening / closing of the elevator door based on the distance from the user to the elevator.

  In such a technique, since it is necessary to detect a user in real time, an analysis process is performed at high speed. For this reason, the detection accuracy is slightly lowered, and high-speed analysis processing is realized.

JP 2008-273709 A

  However, the above-described technique can detect a person having a certain size such as a person with high accuracy. However, since the detection accuracy is lowered, for example, a thin string-like object such as a string connecting pets is accurately detected. There is an inconvenience that it cannot be detected well.

  According to this inconvenience, even if a thin string-like object is likely to be caught in the elevator door, the elevator door can be closed as it is without being able to detect the thin string-like object. There is a risk of serious accidents.

  The problem to be solved by the present invention is to provide an image processing apparatus capable of detecting not only a person but also a thin string-like object in real time.

An image processing apparatus according to an embodiment includes first image acquisition means, first motion detection means, second image acquisition means, second motion detection means, and signal output means. The first image acquisition means acquires a plurality of first images having a resolution lower than that at the time of shooting, regarding the plurality of images continuously shot by the camera while the door of the car door is opened and closed. The first motion detection means detects a motion of a person on the landing by comparing a plurality of first images acquired by the first image acquisition means in units of blocks. The second image acquisition means acquires a plurality of second images in which only the area around the door is shown with respect to the plurality of images continuously captured by the camera. The second motion detection means detects a motion of an object in an area around the door by comparing a plurality of second images acquired by the second image acquisition means in block units. The signal output means is a door opening instruction for opening the door of the car when a block with movement is detected by at least one of the first movement detection means and the second movement detection means. A signal is output to a car control device that controls the opening and closing of the door of the car. The second image has a higher resolution than the first image.

FIG. 1 is a diagram illustrating a configuration of an elevator user detection system including an image processing apparatus according to an embodiment. FIG. 2 is a view for explaining a door peripheral area in the embodiment. FIG. 3 is a diagram for explaining temporal changes in the door peripheral area in the embodiment. FIG. 4 is a view for explaining a door peripheral area different from FIG. 2 in the embodiment. FIG. 5 is a view showing an example of an image taken by the camera in the embodiment. FIG. 6 is a view showing a state in which the image in the embodiment is divided into blocks. FIG. 7 is a diagram for explaining a detection area in real space in the embodiment. FIG. 8 is a diagram for explaining a coordinate system in the real space in the embodiment. FIG. 9 is a diagram for explaining motion detection by image comparison in the embodiment, and is a diagram schematically showing a part of an image taken at time t. FIG. 10 is a diagram for explaining motion detection by image comparison in the embodiment, and schematically shows a part of an image taken at time t + 1. FIG. 11 is a flowchart illustrating an example of a procedure of user detection processing in the embodiment. FIG. 12 is a flowchart showing an example of a procedure of motion detection processing in the embodiment. FIG. 13 is a flowchart showing an example of the procedure of position estimation processing in the embodiment. FIG. 14 is a flowchart showing an example of a procedure for boarding intention estimation processing in the embodiment. FIG. 15 is a diagram showing a change state of the foot position with the intention to board in the embodiment. FIG. 16 is a flowchart illustrating an example of a string-like object detection process procedure according to the embodiment.

Hereinafter, embodiments will be described with reference to the drawings.
FIG. 1 is a diagram illustrating a configuration of an elevator user detection system including an image processing apparatus according to an embodiment. Note that, here, a single car will be described as an example, but a plurality of cars have the same configuration.

  A camera 12 is installed above the entrance of the car 11. Specifically, the camera 12 is installed in a curtain plate 11 a covering the upper part of the entrance / exit of the car 11 with the lens portion facing the landing 15. The camera 12 is a small surveillance camera such as an in-vehicle camera, and has a wide-angle lens and can continuously shoot images of several frames (for example, 30 frames / second) per second. When the car 11 arrives at each floor and opens, the state of the landing 15 is photographed including the state near the car door 13 in the car 11.

  The shooting range at this time is set to L1 + L2 (L1 >> L2). L <b> 1 is a shooting range on the landing 15 side and is, for example, 3 m from the car door 13 toward the landing 15. L2 is a photographing range on the car 11 side, and is, for example, 50 cm from the car door 13 toward the back of the car. L1 and L2 are ranges in the depth direction, and the range in the width direction (direction orthogonal to the depth direction) is at least larger than the lateral width of the car 11.

  In the landing 15 on each floor, a landing door 14 is installed at the arrival entrance of the car 11 so as to be freely opened and closed. The landing door 14 opens and closes by engaging with the car door 13 when the car 11 arrives. The power source (door motor) is on the car 11 side, and the landing door 14 simply opens and closes following the car door 13. In the following description, it is assumed that when the car door 13 is opened, the landing door 14 is also opened, and when the car door 13 is closed, the landing door 14 is also closed.

  Each image (video) captured by the camera 12 is subjected to preprocessing by the image preprocessing device 20 and then analyzed in real time by the image processing device 30. In FIG. 1, for convenience, the image preprocessing device 20 and the image processing device 30 are taken out from the car 11, but actually, the image preprocessing device 20 and the image processing device 30 are the cameras 12. At the same time, it is housed in the curtain 11a. In FIG. 1, the image preprocessing device 20 and the image processing device 30 are illustrated as separate units, but the functions of the image preprocessing device 20 may be mounted on the image processing device 30.

  The image preprocessing apparatus 20 includes a resolution reduction processing unit 21 and a door peripheral image extraction unit 22. The resolution reduction processing unit 21 reduces the resolution of each image (captured image) captured by the camera 12 and generates a low-resolution image. As a method for reducing the resolution of the captured image, for example, there is a method of thinning out pixels constituting the captured image. The generated low resolution image is sent to the image processing apparatus 30. By reducing the resolution of the image in this way, the image size (image capacity) can be reduced, and the analysis processing in the image processing apparatus 30 can be executed at high speed (in real time).

  When the door peripheral image generation unit 22 starts to close the car door 13, the door peripheral image generation unit 22 extracts the door peripheral area from each image captured by the camera 12, and an image (hereinafter, referred to as the extracted door peripheral area) is displayed. (Denoted as a door peripheral image) without reducing the resolution. The generated door peripheral image is sent to the image processing device 30.

  As shown in FIG. 2, the door peripheral area E <b> 0 indicates a predetermined area near both ends of the car door 13. Specifically, the door peripheral area E0 is an arbitrary distance M1 from both ends of the car door 13 to the door closing direction of the car door 13 (however, the distance M1 is a distance shorter than the opening width of the car door 13). For example, 100 mm. The distance M2 in the depth direction of the door peripheral area E0 (the direction orthogonal to the door closing direction of the car door 13) is, for example, the door on the platform 15 side of the landing door 14 from the door surface on the car 11 side of the car door 13. It is set to a value larger than the distance to the surface (door thickness).

  As described above, the door peripheral area E0 has an arbitrary distance M1 from both ends of the car door 13 with respect to the door closing direction of the car door 13, and thus changes with the door closing of the car door 13. Specifically, the door peripheral area E0 changes as shown in FIGS. 3 (a) to 3 (d). First, when the door closing of the car door 13 is started, the door peripheral area E0 is set in the vicinity of both ends of the car door 13 as shown in FIG. Thereafter, as the door closing of the car door 13 proceeds, the two door peripheral areas E0 approach each other as shown in FIG. 3B, and when the door closing of the car door 13 further proceeds, these door peripheral areas E0 As shown in FIG. 3C, they are superimposed (that is, the door peripheral area E0 is gradually reduced). Finally, when the car door 13 is closed, the door peripheral area E0 disappears as shown in FIG.

  Here, it is assumed that the door peripheral area E0 is a rectangular area having a distance of M1 in the door closing direction of the car door 13 and a distance of M2 in the depth direction as shown in FIG. However, the present invention is not limited to this, and the door peripheral area E0 may further include a part of the car 11. Specifically, the door peripheral area E0 may be L-shaped as shown in FIG. In this way, by making the door peripheral area E0 L-shaped, the detection range in the detection process executed by the string-like object detection unit 33 described later can be expanded, and detection in consideration of the height in the real space The process can be executed. However, since the detection processing by the string-like object detection unit 33 using an image having a low resolution has a large processing load, a part of the car 11 included in the door peripheral area E0 affects the processing speed of the detection processing. It is assumed that the size does not reach.

  The door peripheral area E0 may be irradiated with light from a light source provided in the vicinity of the camera 12. According to this, the accuracy of detection processing (accuracy for detecting a string-like object) executed in the string-like object detection unit 33 described later can be improved.

  The image processing apparatus 30 includes a storage unit 31, a user detection unit 32, and a string object detection unit 33. The storage unit 31 sequentially stores low-resolution images and door peripheral images sent from the image preprocessing device 20 and data necessary for processing by the user detection unit 32 and the string-like object detection unit 33. Has a buffer area for temporarily storing.

  The user detection unit 32 detects the presence / absence of a user who intends to get on the vehicle by paying attention to the movement of the person / thing closest to the car door 13 among the low-resolution images sent from the image preprocessing device 20. (judge. If this user detection part 32 is divided functionally, it will be comprised by the motion detection part 32a, the position estimation part 32b, and the boarding intention estimation part 32c.

  The motion detection unit 32a detects the motion of a person / thing by comparing the luminance of each low-resolution image in units of blocks. The “movement of a person / thing” here refers to the movement of a person on the hall 15 or a moving body such as a wheelchair. In the present embodiment, the movement of a person / thing is detected by paying attention to the luminance of the image. However, the method for detecting the movement of the person / thing is not limited to this. The movement of a person / thing may be detected by using it.

  The position estimation unit 32b extracts the block closest to the car door 13 from the blocks with motion detected for each low-resolution image by the motion detection unit 32a, and the center of the car door 13 in the block (the door opening) The coordinate position in the hall direction is estimated as the user's position (foot position).

  The boarding intention estimation unit 32c determines whether or not the user has a boarding intention based on the time series change of the position estimated by the position estimation unit 32b.

  The string-like object detection unit 33 pays attention to the movement of small objects in each door peripheral image sent from the image preprocessing device 20, and an object (for example, a pet) that may be pinched by the car door 13. The presence or absence of a string-like object such as a string to be connected is detected (determined). The “movement of a fine object” referred to here is a swaying (movement) of a string-like object that may be pinched by the car door 13.

  The car control device 40 is connected to an elevator control device (not shown), and transmits and receives various signals such as hall calls and car calls to and from the elevator control device. The “call to hall” is a call signal registered by operating a hall call button (not shown) installed at the hall 15 on each floor, and includes information on the registered floor and the destination direction. The “car call” is a call signal registered by operating a destination call button (not shown) provided in the car room of the car 11, and includes information on the destination floor.

  In addition, the car control device 40 includes a door opening / closing control unit 41. The door opening / closing control unit 41 controls the door opening / closing of the car door 13 when the car 11 arrives at the landing 15. Specifically, the door opening / closing control unit 41 opens the car door 13 when the car 11 arrives at the landing 15 and closes the door after a predetermined time. However, when a person who intends to get on is detected by the user detection unit 32 of the image processing device 30 while the car door 13 is opened and closed, or when the car door 13 is closed, the string-like object of the image processing device 30 is used. When a string-like object is detected by the detection unit 33, the door opening / closing control unit 41 prohibits the door closing operation of the car door 13 and maintains the door open state, or the car door 13 in the middle of the door closing. Is opened again.

FIG. 5 is a diagram illustrating an example of an image photographed by the camera 12. E1 position estimation region in the figure, y _n denotes the Y coordinate of the foot position of the user is detected. FIG. 6 is a diagram illustrating a state in which images (low-resolution images, door peripheral images) stored in the storage unit 31 are divided in units of blocks. An image obtained by dividing an image into a grid of one side W _block is called a “block”. However, the number of pixels included in one block is assumed to be constant. That is, one side W _block of a _block obtained by dividing a door peripheral image having a higher resolution than that of a low resolution image in a grid pattern is larger than one side W _{block of a} block obtained by dividing a low resolution image in a grid pattern. Shorter.

  As shown in FIG. 6, the user detection unit 32 divides the low-resolution image stored in the storage unit 31 into blocks of a predetermined size, detects blocks with movement of people and objects, and blocks with the movement It is determined whether or not the user is willing to board. Similarly, in the string-like object detection unit 33, as shown in FIG. 6, the door peripheral image stored in the storage unit 31 is divided into blocks of a predetermined size, and a block with a small movement is detected. The presence / absence of an object that may be pinched by the door 13 is determined.

  In the example of FIG. 6, the vertical and horizontal lengths of the blocks are the same, but the vertical and horizontal lengths may be different. Further, the blocks may be uniform in size over the entire image, or may be non-uniform in size, such as shortening the length in the vertical direction (Y direction) toward the top of the image. Thus, the foot position estimated later can be obtained with a higher resolution or a uniform resolution in the real space (if it is divided uniformly on the image, the farther away from the car door 13 is, the smaller the resolution is in the real space). .

  FIG. 7 is a diagram for explaining a detection area in the real space, and FIG. 8 is a diagram for explaining a coordinate system in the real space.

  In order to detect the movement of a user who intends to get on the board from a low-resolution image, first, a movement detection area is set for each block. Specifically, as shown in FIG. 7, at least a position estimation area E1 and a boarding intention estimation area E2 are set. The position estimation area E1 is an area for estimating a part of the user's body coming from the landing 15 toward the car door 13, specifically the position of the user's feet. The boarding intention estimation area E2 is an area for estimating whether or not the user detected in the position estimation area E1 has a boarding intention. The boarding intention estimation area E2 is included in the position estimation area E1 and is an area for estimating the user's foot position. That is, in the boarding intention estimation area E2, the user's stepping position is estimated and the user's boarding intention is estimated.

  In the real space, the position estimation area E1 has a distance L3 from the center of the car door 13 toward the landing, and is set to 2 m, for example (L3 ≦ shooting range L1 on the landing side). The lateral width W1 of the position estimation area E1 is set to a distance equal to or greater than the lateral width W0 of the car door 13. The boarding intention estimation area E2 has a distance L4 from the center of the car door 13 toward the boarding direction, and is set to 1 m, for example (L4 ≦ L3). The lateral width W2 of the boarding intention estimation area E2 is set to be approximately the same distance as the lateral width W0 of the car door 13.

  The lateral width W2 of the boarding intention estimation area E2 may be larger than W0. Further, the boarding intention estimation area E2 may be a trapezoid excluding a blind spot in a three-sided frame instead of a rectangle in real space.

  Here, as shown in FIG. 8, the camera 12 has a horizontal direction with respect to the car door 13 provided at the entrance of the car 11 in the X axis, and from the center of the car door 13 to the landing 15 (with respect to the car door 13). The vertical direction) is taken as the Y axis, and the height direction of the car 11 is taken as the Z axis. In each image (low-resolution image, door peripheral image) photographed by the camera 12, the portion of the position estimation area E1 and the boarding intention estimation area E2 shown in FIG. The movement of the foot position of the user moving in the direction of the hall 15 from the center of the vehicle, that is, the Y-axis direction is detected.

This is shown in FIGS.
9 and 10 are diagrams for explaining motion detection by image comparison. FIG. 9 schematically shows a part of an image taken at time t, and FIG. 10 schematically shows a part of an image taken at time t + 1.

  P1 and P2 in the figure are user image portions detected as having motion on the photographed image, and are actually a collection of blocks detected as having motion by image comparison. A block Bx having a motion closest to the car door 13 is extracted from the image portions P1 and P2, and the presence or absence of a ride is determined by following the Y coordinate of the block Bx. In this case, the distance in the Y-axis direction between the block Bx and the car door 13 can be determined by drawing equidistant lines (horizontal lines at equal intervals parallel to the car door 13) as indicated by dotted lines in the Y-axis direction.

In the example of FIGS. 9 and 10, the detection position of the block Bx of the most close motion to the car door 13 is changed to y _n → y _n-1, it can be seen that the user approaches the car door 13 .

Next, the procedure of the user detection process executed by the user detection unit 32 will be described in detail with reference to the flowchart of FIG.
When a predetermined time elapses after the car door 13 of the car 11 is fully open (or when a door closing button (not shown) provided in the car 11 is pressed), the door opening / closing control unit 41 closes the door. The operation is started (step S1). At this time, the photographing operation of the camera 12 is continuously performed from when the car door 13 is opened. The resolution reduction processing unit 21 of the image preprocessing device 20 acquires images continuously captured by the camera 12 and executes a process for reducing the resolution of these images. As a result, an image having a lower resolution than the image captured by the camera 12 is generated (step S2). The image processing device 30 acquires the low-resolution images generated by the image preprocessing device 20, and sequentially stores these images in the storage unit 31 (step S3), and executes the user detection process in real time (step S3). S4).

  The user detection process is executed by a user detection unit 32 provided in the image processing apparatus 30. This user detection process is divided into a motion detection process (step S4a), a position estimation process (step S4b), and a boarding intention estimation process (step S4c).

(A) Motion Detection Process FIG. 12 is a flowchart showing the procedure of the motion detection process in step S4a. This motion detection process is executed by the motion detection unit 32a which is one of the components of the user detection unit 32. Here, the case where the luminance of an image is used to detect the movement of a person / thing will be described.

  The motion detection unit 32a reads out each low-resolution image stored in the storage unit 31 one by one, divides each image into a grid, and calculates an average luminance value for each block obtained thereby (step A1). At this time, the motion detection unit 32a holds, as an initial value, an average luminance value for each block calculated from the first image in time series in a buffer area (not shown) in the storage unit 31 (Step S31). A2).

  When the second and subsequent low-resolution images are read, the motion detection unit 32a reads the average luminance value for each block of the current image and the average luminance value for each block of the previous image held in the buffer area. Are compared (step A3). As a result, when there is a block having a luminance difference equal to or larger than a preset value in the current image, the motion detection unit 32a determines the block as a block with motion (step A4).

  When determining the presence or absence of motion for the current image, the motion detection unit 32a holds the average luminance value for each block of the image in the buffer area for comparison with the next image (step A5).

  Thereafter, in the same manner, the motion detection unit 32a repeatedly determines the presence or absence of motion while comparing the luminance values of the low resolution images stored in the storage unit 31 in block units in time series.

(B) Position estimation process FIG. 13 is a flowchart showing the procedure of the position estimation process in step S4b. This position estimation process is executed by the position estimation unit 32b which is one of the components of the user detection unit 32.

  The position estimation unit 32b checks a block with motion in the current image based on the detection result of the motion detection unit 32a (step B1). As a result, when there is a block with movement in the position estimation area E1 shown in FIG. 7, the user detection unit 32 extracts a block closest to the car door 13 from among the blocks with movement (step) B2).

  Here, as shown in FIG. 1, the camera 12 is installed at the upper part of the entrance / exit of the car 11 toward the landing 15. Therefore, when the user is heading from the landing 15 toward the car door 13, the right or left foot portion of the user may be reflected in the foreground of the captured image, that is, in the block on the car door 13 side. High nature. Therefore, the position estimation unit 32b obtains the Y coordinate (the coordinate in the direction of the landing 15 from the center of the car door 13) of the block with the movement closest to the car door 13 as data of the user's foot position, and stores it in the storage unit 31. It is held in a buffer area (not shown) (step B3).

  In the same manner, the position estimation unit 32b obtains the Y coordinate of the block with the movement closest to the car door 13 for each image as data of the user's foot position, and holds it in the buffer area. Note that such a foot position estimation process is performed not only in the position estimation area E1 but also in the boarding intention estimation area E2.

(C) Ride intention estimation process FIG. 14 is a flowchart showing the ride intention estimation process in step S4c. This boarding intention estimation process is executed by the boarding intention estimation unit 32c, which is one of the components of the user detection unit 32.

  The boarding intention estimation unit 32c smoothes the user's foot position data of each image held in a buffer area (not shown) in the storage unit 31 (step C1). As a smoothing method, for example, a generally known method such as an average value filter or a Kalman filter is used, and detailed description thereof is omitted here.

  When the data at the foot position is smoothed and there is data whose change amount is equal to or greater than the predetermined value (YES in step C2), the boarding intention estimation unit 32c excludes the data as an outlier (step C3). The predetermined value is determined by the standard walking speed of the user and the frame rate of the captured image. Further, outliers may be found and excluded before smoothing the data at the foot position.

  FIG. 15 shows a change state of the foot position. The horizontal axis indicates time, and the vertical axis indicates the position (Y coordinate value). When the user walks from the landing 15 toward the car door 13, the Y coordinate value of the user's foot position gradually decreases with time.

  For example, in the case of a moving body such as a wheelchair, the data changes linearly as shown by a dotted line, but in the case of a user, the left and right feet are detected alternately, so the data change curved as shown by a solid line become. Further, when some noise enters the detection result, the amount of instantaneous change in the foot position increases. Such foot position data having a large change amount is excluded as an outlier.

  Here, the boarding intention estimation part 32c confirms the motion (data change) of the foot position in the boarding intention estimation area E2 shown in FIG. 7 (step C4). As a result, if the movement (data change) of the user's foot position toward the car door 13 in the Y-axis direction in the boarding intention estimation area E2 can be confirmed (YES in step C5), the boarding intention estimation unit 32c determines that the user has an intention to board (step C6).

  On the other hand, when the movement of the user's foot position toward the car door 13 in the Y-axis direction within the boarding intention estimation area E2 cannot be confirmed (NO in step C5), the boarding intention estimation unit 32c It is determined that the user does not intend to board (step C7).

  In this way, the block with the movement closest to the car door 13 is regarded as the user's foot position, and the presence / absence of the user's intention to board is estimated by tracking the temporal change of the foot position in the Y-axis direction. be able to.

  Returning to FIG. 11, when a user with an intention to board is detected (YES in step S <b> 5), a door opening instruction signal is output from the image processing device 30 to the car control device 40. When the door opening / closing control unit 41 of the car control device 40 receives the door opening instruction signal during door closing, the door opening / closing operation of the car door 13 is interrupted and the door opening operation (reopening) is performed again (step S6).

  Thereafter, each time the door closing operation is started again by the door opening / closing control unit 41, the process returns to step S1 and the same process is repeatedly executed.

  Next, a string-like object detection process executed by the string-like object detection unit 33 will be described in detail with reference to the flowchart of FIG. This string-like object detection process is executed in parallel with the user detection process shown in FIG. Similarly to the motion detection process described above, this process will be described on the assumption that the brightness of the image is used to detect an object that may be pinched by the car door 13.

  When a predetermined time elapses after the car door 13 of the car 11 is fully open (or when a door closing button (not shown) provided in the car 11 is pressed), the door opening / closing control unit 41 closes the door. The operation is started (step S11). At this time, the photographing operation of the camera 12 is continuously performed from when the car door 13 is opened. The door peripheral image generation unit 22 of the image preprocessing device 20 acquires images continuously captured by the camera 12, and extracts the door peripheral area E0 from these images. As a result, a door peripheral image (high resolution) in which only the door peripheral area E0 is shown is generated (step S12). The image processing apparatus 30 acquires (high-resolution) door peripheral images generated by the image preprocessing apparatus 20, and sequentially stores these images in the storage unit 31 (step S13), while performing string object detection processing in real time. Run with.

  The string-like object detection unit 33 in the image processing device 30 reads the door peripheral images stored in the storage unit 31 one by one, divides each door peripheral image into a grid, and the average luminance value for each block obtained thereby Is calculated (step S14). At that time, the string-like object detection unit 33 holds, as an initial value, an average luminance value for each block calculated from the first image in time series in a buffer area (not shown) in the storage unit 31. (Step S15).

  When the second and subsequent door peripheral images are read, the string-like object detection unit 33 reads the average luminance value for each block of the current image and the average for each block of the previous image held in the buffer area. By comparing with the luminance value, it is determined whether or not there is an object that may be pinched by the car door 13 (step S16). Specifically, if there is a block having a luminance difference equal to or greater than a preset value in the current image as a result of the comparison, the string-like object detection unit 33 determines the block as a block with motion. It is determined that there is an object that may be caught between the car doors 13. That is, an object that can be caught between the car doors 13 is detected.

  When an object that may be caught by the car door 13 is detected (YES in step S16), a door opening instruction signal is output from the image processing device 30 to the car control device 40. When the door opening / closing control unit 41 of the car control device 40 receives the door opening instruction signal while the door is closed, the door closing operation of the car door 13 is interrupted and the door opening operation (reopening) is performed again (step S17).

  Thereafter, every time the door closing operation is started again by the door opening / closing control unit 41, the process returns to step S11 and the same process is repeatedly executed.

  According to the embodiment described above, the image processing apparatus 30 performs the user detection process for detecting the person at the hall 15 using the low-resolution image, but does not reduce the resolution. A string-like object detection process for detecting an object that may be caught by the car door 13 using a door peripheral image in which only the peripheral area E0 is shown is provided. According to this configuration, since the area is limited to the door peripheral area E0 and a high-resolution image that can detect a string-like object is used, the processing load is light and not only a person but also a thin string-like shape is used. This object can also be detected in real time.

  In addition, although some embodiment of this invention was described, these embodiment is shown as an example and is not intending limiting the range of invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

  DESCRIPTION OF SYMBOLS 11 ... Car, 12 ... Camera, 13 ... Car door, 14 ... Landing door, 15 ... Landing place, 20 ... Image pre-processing apparatus, 21 ... Resolution reduction processing part, 22 ... Door periphery image generation part, 30 ... Image processing apparatus , 31, storage unit, 32, user detection unit, 32 a, motion detection unit, 32 b, position estimation unit, 32 c, ride intention estimation unit, 33, string object detection unit, 40, car control device, 41, door open / close Control unit, E0 ... door peripheral area, E1 ... user position estimation area, E2 ... boarding intention estimation area.

Claims (4)

First image acquisition means for acquiring a plurality of first images having a resolution lower than that at the time of shooting, with respect to the plurality of images continuously shot by the camera while the door of the car door is opened and closed;
A first motion detecting means for comparing a plurality of first images acquired by the first image acquiring means in units of blocks and detecting a motion of a person on the landing;
Second image acquisition means for acquiring a plurality of second images in which only the area around the door is displayed with respect to the plurality of images continuously captured by the camera;
A second motion detection means for comparing a plurality of second images acquired by the second image acquisition means in block units and detecting a motion of an object in the area around the door;
When a block with motion is detected by at least one of the first motion detection means and the second motion detection means, a door opening instruction signal for opening the door of the car is sent to the car Signal output means for outputting to the car control device for controlling the opening and closing of the door of the door ,
The image processing apparatus according to claim 1, wherein the second image has a higher resolution than the first image . The second image acquisition means includes
2. The image according to claim 1, wherein a plurality of second images in which only an area around the door is displayed are acquired with respect to a plurality of images taken while the door of the car door is closed. Processing equipment.   The area around the door has a distance shorter than the opening width of the door with respect to the door closing direction of the car door, and has a distance longer than the thickness of the door with respect to the door closing direction. The image processing apparatus according to claim 1. The block closest to the door of the car is extracted for each first image from the blocks with motion detected by the first motion detection means, and the coordinates of the landing direction from the center of the door in the block Position estimation means for estimating the position as the position of a person;
Boarding intention estimating means for estimating the presence or absence of the person's boarding intention based on a time-series change in the position of the person estimated by the position estimating means; and
The signal output means includes
When the person estimated to have boarding intention by the boarding intention estimation means is at the landing, or when a block with motion is detected by the second movement detection means, the door opening instruction signal is controlled by the car. The image processing apparatus according to claim 1, wherein the image processing apparatus outputs the image to an apparatus.

Images