JPH11282999A - Instrument for measuring mobile object - Google Patents

Abstract

PROBLEM TO BE SOLVED: To count the number of mobile objects with precision while considering the transition of movement in the respective objects by forming two-dimensional X-T images and two-dimensional Y-T images, analyzing at least one of a movement direction and a movement speed concerning the objects, based on the Y-T images and counting the number of them, based on the X-T image through the use of an analysis result. SOLUTION: An X-T image forming part 34 aligns X-axis binary images in a time base direction by time sequence order and forms the two-dimensional X-T images. A congestion degree judging part 36 judges the congestion degree of persons on a passage based on the X-T image. A Y-T image forming part 42 aligns Y-axis binary images in the time base direction by time sequence order and forms the two-dimensional Y-T images. A direction discriminating part 46 discriminates the movement direction of the mobile object by the inclining direction of a cluster and a speed discriminating part 48 judges the movement speed by the inclining size of the cluster. A separation by direction part 52 separates the X-T images by direction, based on a map by direction and an individual counting part 54 counts the number of passing persons within a processing time.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving object measuring apparatus, and more particularly, to an apparatus for measuring the number of people at a place where people come and go, such as a passage.

[0002]

2. Description of the Related Art The number of people is measured for various purposes in passages where people circulate, such as the stairs of a station or a ticket gate. At that time, counting the number of persons manually is extremely complicated. For this reason, a moving object measurement device that measures the number of people by image processing has been proposed. Products, vehicles,
The moving object measuring device is also used when automatically counting the number of animals and the like.

[0003] Japanese Patent Application Laid-Open No. 8-123935 discloses a conventional moving object measuring device. In this device, an image of an observation area is acquired by a camera. One-dimensional counting lines are set in the image area. The counting line functions as an observation line for monitoring the passage of the moving object. Here, moving object images are sequentially extracted by calculating the difference between the one-dimensional image representing the background on the counting line and the one-dimensional image including the moving object image on the counting line. Then, for a two-dimensional object image formed by aligning the moving object images at each time in the time axis direction, the number of persons is counted by, for example, dividing the total area by a reference value. Further, in this conventional device, a counting line for direction determination is set above and below the counting line. Then, the moving direction and speed of the moving object are determined based on the passing position and the passing time of the moving object on those counting lines.

[0004] Japanese Patent Application Laid-Open No. 5-266196 also discloses a person counting apparatus based on image processing.

[0005]

However, in the apparatus proposed in Japanese Patent Application Laid-Open No. H8-123935, when the congestion degree is high, another person simultaneously or near the upper and lower counting lines. There is a possibility of passing at the time, and in that case, a problem is pointed out that it is difficult or inaccurate to determine the moving direction and the moving speed. Further, even when the congestion degree is low, it is considered that the same problem as described above occurs when a person stops on the counting line or another person overtakes another person. Note that this conventional document does not disclose any method of measuring the number of people in consideration of the moving speed, switching of the measurement method according to the degree of congestion, and the like.

The present invention has been made in view of the above-mentioned conventional problems, and an object of the present invention is to accurately measure the number of moving objects in consideration of the transition of the movement of each moving object.

Another object of the present invention is to measure the number of moving objects by an appropriate method in consideration of the degree of congestion.

It is another object of the present invention to measure the number of moving objects in consideration of a moving speed.

[0009]

(1) In order to achieve the above object, the present invention is to sequentially convert an X-axis binary image representing a moving object passing through an X-axis window set along the X-axis. A first image forming means for acquiring and aligning the X-axis binary images at each time in the time axis direction to form a two-dimensional X-T image; and a plurality of images arranged and set across the X-axis window. For each Y-axis window, a Y-axis binarized image representing the moving position of the moving object is sequentially obtained, and the Y-axis binarized images at each time are aligned in the time axis direction to form a two-dimensional YT image. A second image forming means, a movement analyzing means for analyzing at least one of a moving direction and a moving speed of each moving object based on the Y-T image, and the X Counting means for counting the number of moving objects based on the T image. And butterflies.

According to the above configuration, an X-axis binarized image representing the existence (or displacement) of the moving object on the X-axis window is obtained, and the X-axis binarized images are aligned in chronological order. A two-dimensional X-T image is formed. One axis of the X-T image corresponds to the longitudinal direction (X-axis) of the X-axis window, and the other axis corresponds to the time axis (T-axis).

On the other hand, a plurality of Y-axis windows are set to cross the X-axis window, and a Y-axis binarized image representing the position (or displacement) of the moving object on each Y-axis window is obtained. A two-dimensional YT image is formed by aligning the axial binarized images in chronological order. That YT
One axis of the image corresponds to the longitudinal axis (Y axis) of the Y axis window, and the other axis corresponds to the time axis (T axis).

Incidentally, it is desirable that the time axes of both the XT image and each of the YT images coincide with each other, so that the two images are formed at a predetermined timing (a fixed period or a timing at which the movement is stopped). It is desirable to control. Note that the X axis is desirably orthogonal to the main movement direction, and the Y axis is desirably parallel to the main movement direction.

As described above, Y-T in Y-axis window units
When an image is obtained, the movement of the moving object is recognized on each Y-T image, and information such as the moving direction and the moving speed is acquired using the movement. Therefore, at each position on the X-axis window, even if the moving direction of each moving object is different and changes with time, according to the above YT image, the moving object is focused on the motion of each moving object. The direction and the like can be faithfully determined. Of course, if it is assumed that all the moving objects move in a certain direction, it is not always necessary to determine the moving direction.

When the movement analysis is performed as described above, the analysis result is used in counting the number of objects on the X-T image. Various methods can be applied as the counting method. However, if the counting process is performed using at least one of the moving direction and the moving speed, the counting accuracy can be improved particularly in a state where the congestion degree is high. When the moving speed is obtained for each moving object, the counting process may be performed using a representative value (for example, an average value), or the counting process may be performed in consideration of the moving speed of each moving object. May go.

The apparatus according to the present invention is desirably used for measuring the number of people at places where people circulate in two directions (forward and reverse directions), such as the stairs of a station. That is, according to the present invention, the number of people can be counted at high speed and with high accuracy in a place where the degree of congestion fluctuates or a dense state is formed.

(2) In the present invention, preferably, imaging means for imaging an observation area including the X-axis window and the plurality of Y-axis windows, and an adjacent frame for each image acquired by the imaging means A difference calculation means for performing a difference calculation between the two to obtain a difference image; and a binary image creation means for creating the X-axis binary image and the Y-axis binary image based on the difference image. Including.

According to the above arrangement, an X-axis binarized image and a Y-axis binarized image are taken out based on the image picked up by the image pickup means, and each binarized image is extracted using a single image pickup means. There are advantages that can be obtained. Here, the imaging means is preferably provided above the observation area (preferably above the center of the X-axis window). If such setting is made, variations in the height (for example, height) of the moving object and the distance between the moving objects The measurement error due to the overlap of can be reduced as much as possible.

For example, when an image is taken using a camera or the like, the appearance of the image of the moving object differs depending on the distance of each image pickup position from immediately below the camera. In this case, weighting or weighting according to the distance is used. It is desirable to perform image correction.

Of course, dedicated imaging means for the X-axis window and each Y-axis window may be individually provided. As such an imaging means, a CCD element or the like can be used. Alternatively, the captured image may be recorded in a recording device, and the number of moving objects may be counted by processing the reproduced image.

The difference image is an image in which only the portion where the displacement has occurred is represented as an image, and only the displacement of the moving object can be imaged, and the background and the stationary object (including the stationary object itself)
Can be eliminated. Note that it is desirable to apply, for example, a binarization process to the difference image in order to separate and extract the displacement of the moving object from the background (noise).

(3) In the present invention, preferably, the binarized image creation means cuts out an X-axis window image corresponding to the X-axis window from the difference image,
Means for projecting the axis window image on the X axis and performing binarization processing to create the X axis binary image, and cutting out a Y axis window image corresponding to each of the Y axis windows from the difference image, Means for projecting the Y-axis window image on the Y-axis and performing binarization processing to create the Y-axis binary image.

The essence of the binarized image lies in the extraction and clarification of the moving object (or its displacement). Therefore, each window image is acquired as an image from which only the moving object can be discriminated by means of lighting and imaging. If such a window image is used, it is considered that the window image substantially corresponds to the binarized image, and the window image may be used as it is for image formation. In general, in order to eliminate noise and local brightness differences, and to accurately extract a moving object, it is necessary to apply a projection process and a binarization process to each window image to create a binarized image. desirable. Each binarized image is preferably configured as a one-dimensional image (line image) or a band-like image having a certain width.

(4) In the present invention, preferably, the movement analysis means includes movement trajectory recognition means for recognizing a movement trajectory of each moving object in the YT image, and the moving object is determined based on a tilt direction of the movement trajectory. Direction determining means for determining the moving direction of the camera.

In the YT image, a change over time in the position of the moving object observed in each Y-axis window is represented as a moving locus, and the moving direction is determined from the inclination direction of the starting locus. In this case, if the moving locus does not become a straight line due to a change in the moving speed, the moving direction may be determined from the inclination direction of the straight line approximating the moving locus.

According to the above processing, the moving speed is obtained for each moving object passing through each Y-axis window. For example, in the case where the moving direction of the moving object in the specific Y-axis window is reversed within the time axis range of the Y-T image, the reversal can be grasped by the above processing.

(5) In the present invention, preferably, the counting means is configured to control the X-axis according to the determined moving direction.
Classifying means for classifying an object image included in a T image for each moving direction, direction-specific counting means for counting the number of moving objects for each moving direction based on the XT image after the classification,
including.

As described above, since the moving direction of each moving object is determined, the number of moving objects can be measured while distinguishing the moving direction of each moving object on the XT image.

(6) In the present invention, preferably, the movement analysis means includes a movement trajectory recognition means for recognizing a movement trajectory of each moving object in the YT image, and the movement analysis means determines the inclination of the movement trajectory. Speed determining means for determining the moving speed of the moving object.

According to the above arrangement, the inclination of the movement locus in the YT image corresponds to the movement speed, and the movement speed is calculated from the magnitude of the inclination. When the moving speed changes in the middle, the moving speed may be obtained from the inclination of the moving trajectory before and after the change, or when the moving trajectory can be linearly approximated, the moving speed may be obtained as an average value of the speed. Good. Further, the moving speed may be calculated for only one or a plurality of representative moving trajectories in response to a request for the number of objects.

(7) In the present invention, preferably, the counting means includes means for variably setting a unit area based on the determined moving speed, and means for setting a total area of the object image in the XT image to the unit area. Division counting means for counting the number of moving objects based on the result of the division by the area.

For example, when the congestion degree is high, it becomes difficult to individually count each object image in the XT image. Therefore, the number of moving objects is obtained by dividing the total area of the object image by the reference area per moving object. In this case, if the reference area is variably set according to the speed (for example, the average speed) of the moving object, the number of objects can be counted with relatively high accuracy even in a dense state.

(8) In order to achieve the above object, the present invention sequentially obtains an X-axis projected binary image for observing a moving object passing through an X-axis window set along the X-axis. XT image forming means for forming a two-dimensional XT image by aligning the X-axis projected binary images at each time in the time axis direction, and congestion determining a congestion degree based on the XT image And a counting means for counting the number of moving objects based on the XT image in accordance with a counting method according to the determined congestion degree.

According to the above configuration, the counting method can be switched according to the degree of congestion, and appropriate counting can be performed according to the situation. For example, when the congestion degree is low, a counting method faithful to the actual passing fact is selected, and when the congestion degree is high, a counting method capable of counting even if the moving objects are dense is selected. If the counting methods are flexibly switched in this way, the advantages of each counting method can be fully utilized.

(9) In the present invention, preferably, the congestion degree is determined based on the total area of the object images included in the XT image. Since the X-T image reflects the number of passing moving objects within a certain period, the congestion degree is determined based on the number.

(10) In the present invention, desirably, the counting means counts the number of object images included in the XT image when the congestion degree judging means judges that the congestion degree is low. The individual counting means for calculating the number of moving objects, and when the congestion degree determining means determines that the congestion degree is high, dividing the total area of the object images included in the X-T image by a reference area. And division counting means for calculating the number of moving objects by

According to the above configuration, the individual counting method is applied when the congestion degree is low, that is, when the possibility of connection of each object image is low, and when the congestion degree is high, that is, the connection of each object image is possible. In the case where the performance is high, a division counting method using the area as a calculation object is applied for convenience. When the individual counting method is applied, the size (area) of each object image may be evaluated according to the moving speed. For example, when there is an object image that is too large even in consideration of the moving speed, it can be recognized as a connection of the object images, and a division operation method may be applied to the portion. However, when a person is carrying a large luggage, another image processing may be combined and applied so as not to erroneously determine that the two are two persons.

(11) In the present invention, preferably, the individual counting means and the division counting means count the number of moving objects for each moving direction. In this configuration, it is desirable to determine the moving direction based on the above-described YT image.

As described above, according to the present invention, it is not necessary to set or tune complicated parameters, and it is not necessary to use expensive equipment, and even if the congestion degree is high, the counting can be performed with relatively high accuracy. It can be performed. Therefore, there is an advantage that it is possible to provide a low-cost and highly practical moving object measurement device.

[0039]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows a preferred embodiment of the moving object measuring apparatus according to the present invention, and FIG. 1 is a block diagram showing the entire configuration. The moving object measurement device according to the present embodiment is, for example, a people measurement device provided at a ticket gate or stairs of a station. Of course, the present invention is also applicable to a case where an object other than a person is measured.

In FIG. 1, an image input unit 10 is a means for taking in an image of an observation area. In this embodiment, an image captured by the camera 12 is converted into a digital signal by the A / D converter 14. Thus, the signal from the camera 12 may be directly input,
For example, an image captured by the camera 12 may be temporarily stored in the recording device 16 and a signal read from the recording device 16 may be fetched.

It is desirable that the camera 12 be installed just above the observation area through which a person passes. By arranging the cameras 12 in this manner, the flow of people can be observed from above, so that the influence of concealment and variation in physique (such as height) can be reduced.

FIG. 2 shows an image 100 of the observation area captured by the camera 12. In FIG. 2, reference numeral 106 indicates a position immediately below the camera, and reference numerals 108 and 110 indicate a left end 108 and a right end 110 of the passage 102, respectively. Such coordinates are registered by the user through a parameter input unit 30 described later. In FIG. 2, the X axis is orthogonal to the main flow direction,
The Y axis is parallel to the main flow direction.

It is desirable that the camera 12 is arranged at a position as high as possible. If such an arrangement is adopted, it is possible to prevent overlap between persons as much as possible and to reduce measurement errors. By the way, the device of the present embodiment has a passage 1 shown in FIG.
02 is intended to measure the number of people who circulate in the forward and reverse directions, and the movement of the person in the direction orthogonal to the circulation direction is not detected. By extending the principle of the present invention, it is also possible to measure a person moving in various directions.

When the object to be imaged is not a flat passage but a stair, for example, the camera 10 is installed above or diagonally above the stair. In any case, as described above, it is desirable to perform imaging from a direction in which people do not overlap as much as possible.

The parameter input section 30 shown in FIG. 1 is composed of, for example, a keyboard and a mouse, and is a means for allowing a user to input an imaging condition.
Of course, if the imaging conditions are known in advance, it is not always necessary to provide such a parameter input unit 30. The input parameters include the position just below the camera 106, the left end 108 and the right end 110 of the passage, or the width of the passage as shown in FIG. 2, and the angle of view and installation height of the camera 12 used for imaging. Can be In inputting those parameters, for example, a captured image is displayed on a display, and as shown in FIG. 2, coordinates corresponding to each parameter are clicked using a pointing device such as a mouse to input a user input. You may. By using such a user interface, the burden on the user can be reduced, and the complexity of setting the initial parameters can be greatly reduced. If the camera is permanently installed, it is not necessary to input such parameters every time each measurement is performed, and it is sufficient to register them in advance. Even when the number of people is measured by temporarily installing a camera, the input of parameters is simple as described above, and there is an advantage that the measurement can be executed immediately.

Returning to FIG. 1, A / D
An image as a digital signal output from the converter 14 is input. The inter-frame difference calculation unit 18 is a unit that obtains a difference image by performing a difference calculation between frames. The difference image is an image representing the movement, that is, the displacement of the person. According to such a difference calculation, it is possible to pay attention to the movement of a person, and it is possible to eliminate, for example, when a person is stationary, performing redundant measurement. There is also an advantage that separation from the background can be performed easily and reliably.

The binarization processing section 20 is means for executing a binarization process on the difference image. According to such binarization, it is possible to reliably extract only the displacement of the moving object while eliminating noise and the like. The threshold in that case is set by the threshold setting unit 22. The threshold value setting unit 22 is set according to the brightness of each part in the image in order to cope with a local difference in brightness in the image 100 shown in FIG. For this reason, the image output from the A / D converter 14 is desirably referred to by the threshold value setting unit 22. Then, the threshold setting unit 22 sets an appropriate threshold based on such an image. In this case, it is desirable that the threshold value be variably set corresponding to each coordinate on the X-axis in accordance with the setting of an X-axis window 112 described later along the X-axis direction.

The difference image subjected to the binarization processing as described above is temporarily stored in the memory 24. The camera 12 continuously captures the image 100, and the memory 24 sequentially stores sequentially acquired difference images. Of course, when the X-axis binarized image forming unit 32 and the Y-axis binarized image forming unit 40 each have a storage unit, it is not necessary to provide the memory 24 independently.

The processing period control unit 26 controls an X-
This is a means for setting a formation period of a T image and a YT image, that is, a processing period. Such a processing period is set as a fixed period or as a timing at which circulation of a person on the passage 102 is interrupted. Therefore, a timer 28 is connected to the processing period control unit 26, and a difference image is fetched as needed. When the processing period control unit 26 outputs the refresh signal, the content stored in the memory 24 is cleared at that timing, and the difference image is stored in the next processing period. Of course, the difference images stored in the memory 24 are sequentially sent to an X-axis binarized image forming unit 32 and a Y-axis binarized image forming unit 40, which will be described later, at a predetermined timing.

By the way, the timer 28 measures, for example, 10 seconds, that is, the image processing and the number of persons described below are executed at intervals of 10 seconds as long as the circulation of the person is not interrupted. Prior to the description, an X-axis window (counting window) and a Y-axis window (direction / speed detection window) will be described.

FIGS. 3 and 4 conceptually show such an X-axis window 112 and a Y-axis window 114. These windows are virtually set in the observation area. In FIG. 3, the X-axis window 112 is set along a direction (the X-axis direction in each drawing) orthogonal to the main flow direction of the person (the Y-axis direction in each drawing). That is, the X-axis window 112 is set as a rectangular area extending in the X-axis direction. Its width W1
Is set to be at least equal to or larger than the detection range in which the number of people is measured. In consideration of the image processing speed of the apparatus shown in FIG. 1 and the height of the camera 12, the height H1 is set as a size at which a passing person appears in one frame, preferably a plurality of frames. Of course, when detecting the number of articles flowing at a constant low speed on the belt conveyor, for example, the height H1 of the X-axis window 112 may be substantially one pixel. However, when the number of people is measured, the height H1 is set as described above in consideration of the predicted maximum passing speed of the person and the like.

The X-axis window 112 is automatically set in accordance with the input of each parameter as shown in FIG. 2, and the center coordinates of the X-axis window 112 correspond to the position 106 directly below the camera.

In FIG. 4, a plurality of Y-axes cross each position of the X-axis window 112, ie, along the Y-axis direction.
The axis window 114 is automatically set. The Y-axis window 114 is basically set so that it is orthogonal to the X-axis window 112 and that the center coordinate of the Y-axis window 114 in the Y-axis direction matches the Y-axis center coordinate of the X-axis window 112. In the example shown in FIG. 4, n Y-axis windows 114 are set, and each Y-axis window 114
Are in contact with each other. Here, if the number of the Y-axis windows 114 is increased, the measurement accuracy can generally be improved accordingly, but the processing load increases. On the other hand, the number of the Y-axis windows 114 is determined in consideration of the processing capability of the apparatus. desirable.

The width W in the X-axis direction of each Y-axis window 114
2 is determined from the length of the X-axis window 112 in the X-axis direction and the number n of the Y-axis windows 114, and the height H2 in each Y-axis direction is appropriately set according to the expected maximum moving speed of the person. You. If the height H2 is increased, a movement trajectory to be described later can be acquired over a longer time or a longer distance. However, the processing load increases correspondingly, so the height H2 is set according to the processing capacity of the system. Is desirable.

For example, in an image of a staircase or the like, the distance from the camera is different between the upper part and the lower part of the staircase, and the appearance of the moving object (person) in the image is different. Therefore, as will be described later, image correction and weighting may be performed according to such a difference in appearance, or weighting may be performed by making the Y-axis window 114 a trapezoidal region.

Returning to FIG. 1, each difference image stored in the memory 24 is sequentially input to the X-axis binary image forming unit 32. Then, the X-axis binarized image forming unit 32 first cuts out an X-axis window image corresponding to the X-axis window 112 shown in FIG. 3, projects it on the X-axis, and performs a binarization process. An X-axis binarized image 122 (FIG. 5) is formed.

FIG. 5 conceptually shows the processing contents. As shown in (A), the X-axis window 112 is automatically set in the observation area using the parameters input by the parameter input unit 30. The X-axis window image cut out from the difference image includes the difference image 120 of the moving object. That is, the difference image 120 is an image representing the displacement of the person between the frames. As shown in (B), the X-axis window image is projected on the X-axis,
As a result, a frequency distribution of luminance values on the X axis is created. By binarizing the frequency distribution using a threshold value, an X-axis binarized image 122 as shown is formed. The X-axis binarized image 122 includes an object image 124 corresponding to the difference image 120 described above. By the way, code 12
Reference numeral 6 denotes a background portion other than the object image 124.

Returning to FIG. 1, the XT image forming section 34
This is a means for forming a two-dimensional XT image by aligning the X-axis binarized image 122 created as described above in the time axis direction in chronological order. FIG. 6 shows an X-T image 130. As described above, this image is created by connecting the X-axis binarized images 122 shown in FIG. 5B in time series. On the X-T image 130, the object images 124 are connected in the time axis direction, and a cluster 132 is formed by them. This cluster 132
Is positioned as an object image representing its displacement during the period when a person passes through the X-axis window 112.

By the way, as shown in FIG.
200A and 202A, a person 200 located immediately below and a person 202 at a position displaced therefrom.
The appearance from the camera 12 is different as shown in FIG. Therefore, it is desirable to perform correction corresponding to the difference in appearance. The X-T image forming unit 34 shown in FIG. 1 also has a function of performing such correction, and FIG. 8 shows the details of the correction. That is, since the difference in appearance as described above can be grasped in accordance with the parameters set in the parameter input unit 30 and the position of the cluster, correction of each cluster is performed so as to eliminate such difference in appearance. Will be As shown in FIG. 8, no special correction is performed on the cluster located at the center on the X axis, and X
As the shaft is displaced left and right in the axial direction, the area of the cluster is corrected as indicated by A in reference numeral 130 according to the degree of displacement. Of course, the correction as shown in FIG. 8 may be performed on the X-T image 130 itself. However, if weighting is performed based on the difference in appearance in counting the number of people to be described later, substantially the same correction as described above is performed. Can do it.

The XT image formed as described above is sent to the direction separating section 52 described later and also to the congestion degree judging section 36 in FIG. Congestion degree judgment unit 3
Reference numeral 6 denotes a passage 10 based on the X-T image in the present embodiment.
2 is a means for determining the degree of congestion of a person on 2. For example, the congestion degree may be determined from the ratio of the total area of the cluster to the entire area of the XT image. Note that the congestion degree may be determined based on the detection duration of a moving object observed in the X-axis window 112 or the like.

The counting method selecting section 38 is a means for selecting each counting method to be described later according to the determined congestion degree.
By switching the counting method according to the degree of congestion, it is possible to accurately measure the number of people according to the situation. This will be described later in detail.

The difference image is sequentially input from the memory 24 to the Y-axis binary image forming section 40 shown in FIG. The Y-axis binarized image forming unit 40 cuts out a Y-axis window image corresponding to the Y-axis window from the difference image, projects it on the Y-axis, and performs a binarization process to thereby convert the Y-axis binarized image. 134 (FIG. 9).

FIG. 9 conceptually shows the processing contents of the Y-axis binary image forming section 40. As shown in (A), a plurality of Y-axis windows 114 are automatically set across the X-axis window 112 in the observation area.
In setting such a Y-axis window 114,
Reference is made to the various parameters described above. The Y-axis binarized image forming unit 40 cuts out the Y-axis window image for each Y-axis window 114, and projects the Y-axis window image on the Y-axis, thereby obtaining the frequency distribution of the luminance value on the Y-axis. Form. By binarizing the frequency distribution using a threshold value, Y as shown in FIG.
An axis binarized image 134 is formed. Here, the object image 136 is included in the Y-axis binarized image 134, and the object image 136 corresponds to the difference image 120 included in the difference image. Incidentally, reference numeral 138 indicates a background portion other than the object image 136. Therefore, n Y-axis windows 114 are set for one difference image, and a Y-axis binary image 134 is formed for each Y-axis window 114.

Returning to FIG. 1, the YT image forming section 42
This is a means for forming a two-dimensional YT image by aligning the Y-axis binarized image 134 formed for each Y-axis window as described above in the time axis direction in chronological order.

FIG. 10 shows a YT image 140. By aligning the individual Y-axis binarized images, FIG.
The object images 136 shown in (B) are connected in chronological order, and they constitute a cluster 142. Cluster 1
Reference numeral 42 denotes a linear pattern formed when a person passes through the Y-axis window at a constant speed. On the other hand, when the moving speed of the person changes in the Y-axis window 114, the pattern is bent in accordance with the speed change. It is configured as a shape or a curved pattern.

In each case, the direction of inclination of the cluster 142 corresponds to the direction of movement of the person, and the degree of inclination of the cluster 142 corresponds to the degree of movement speed of the person. According to the Y-T image 140, the movement trajectory can be obtained by focusing on the movement of each person for each Y-axis window.

The moving trajectory extracting section 44 shown in FIG.
The labeling is performed by, for example, assigning a label to each adjacent pixel on the Y-T image 140 shown in (1), thereby separating and extracting each cluster 142, that is, the movement trajectory. Each cluster thus extracted is input to the direction discriminating unit 46 and the speed discriminating unit 48.

The direction judging section 46 is means for judging the moving direction of the moving object from the tilt direction of the cluster, and the speed judging section 48 is means for judging the moving speed from the magnitude of the tilt of the cluster.

FIG. 11 shows the concept of direction discrimination and speed discrimination. As shown in FIG.
A cluster 142 is formed on 0 for each moving object passing therethrough. An approximate straight line 146 is obtained by connecting, for example, the area centroid points of the cluster elements to the cluster 142.
Then, the direction determining unit 46 determines the moving direction for each moving object from the inclination direction of the approximate straight line 146.
In addition, the speed determining unit 48 determines the moving speed of each moving object according to the magnitude of the inclination of the approximate straight line 146. Here, as shown in FIG. 11A, if the slope of the approximate straight line 146 is negative, the moving direction is determined to be down,
On the other hand, if the slope is positive, it is determined to be up. The angle θ between each of the approximate straight lines 146 and the T axis (or Y axis) is obtained by calculation, and the moving speed is determined from the θ as described above.

The direction-specific map generating section 50 shown in FIG. 1 aligns the direction-specific patterns 148 shown in FIG. 11B in the X-axis direction, thereby forming the direction-specific map 15 shown in FIG.
This is a means for creating 0. That is, direction-specific pattern 1
Reference numeral 48 denotes a pattern representing the direction of each moving object that has passed through a specific Y-axis window during the processing period set by the processing period control unit 26. For example, on the stairs of the station, the distribution direction of people changes momentarily according to the arrival and departure of a train, and the direction-specific pattern 148 reflects such ever-changing flow of people for a specific Y-axis window. It is a pattern. Therefore, the Y-axis window 11
By arranging and aligning the direction-specific patterns 148 obtained for each Y-axis window along the arrangement direction of 4, ie, along the X-axis direction, the above-described direction-specific map 150 can be obtained. The direction-specific map 150 is an image in which a change in the distribution direction of a person within a certain processing period is represented at each part on the X-axis.

The direction separating section 52 shown in FIG.
The XT image 130 output from the image forming unit 34 and the directional map 15 output from the directional map creating unit 50
0 is input. And the direction-specific separation unit 52
Performs a process of separating the X-T image 130 for each direction based on the direction map 150. FIG. 12 shows the details of the processing. When the direction-specific map 150 and the X-T image 130 are integrated as shown in FIG.
As shown in (1), an up image 130A and a down image 130B are generated. That is, for each cluster 132 included in the X-T image 130, it is discriminated whether it is the upward direction or the downward direction. According to such processing for each direction, there is an advantage that the number of people can be measured for each direction.

In FIG. 1, the up image 130A and the down image 130B separated as described above are
And the division count unit 56. Their counting section 5
Reference numerals 4 and 56 function alternatively according to the degree of congestion, and their operations are controlled by the counting method selection unit 38. In the present embodiment, when the low congestion degree is determined, the operation of the individual counting unit 54 is selected by the counting method selection unit 38. On the other hand, when the high congestion degree is determined, the operation of the division counting unit 56 is selected by the counting method selection unit 38.

The operation of each counting section will be described below. In the present embodiment, the evaluation is performed in two stages of the low congestion degree and the high congestion degree. However, the congestion degree may be evaluated in three or more stages. Then, it is desirable to select an appropriate counting method according to the degree of congestion.

The individual counting section 54 is a means for counting the number of clusters 132 in each of the up image 130A and the down image 130B shown in FIG. This takes into account the possibility that each moving object is likely to be independently clustered during off-peak hours. In addition, in the division counting described below, the total area of the cluster is divided by the reference area per person, but when there is no traffic, there is a high possibility that the moving speed of each person is different, and in this case, If a uniform reference area is used, the counting error increases. In that sense, it is desirable to perform individual counting during low congestion.

However, when a plurality of people pass vertically or horizontally, a counting error may occur. Therefore, in the present embodiment, processing is performed to determine how many persons each cluster corresponds to, based on the width and the area thereof. The threshold for determining the width may be based on the width of the person observed in the X-axis window, and the threshold for determining the area may be determined based on the moving speed of each cluster determined by the speed determining unit 48. The staying time of the moving object in the axis window can be estimated and determined by multiplying the estimated time by the width of the person.

FIG. 13 shows an example of the operation of the individual counting section 54. The operation shown in FIG.
30A and the down image 130B.

In S101, clusters are extracted one by one from the image. In S102, a reference area is calculated based on the moving speed of the cluster determined by the speed determination unit 48 and the above-described width of the person (reference width). S103
Then, it is determined whether or not the width of the cluster is equal to or smaller than a predetermined value.
At 104, it is determined whether the cluster area is equal to or smaller than the reference area. If the width is less than the predetermined value and the cluster area is less than the reference area, it can be recognized as one person,
At S105, counting is performed. On the other hand, if NO is determined in at least one of S103 and S104, it is determined in S106 how many clusters correspond. For example, the number of people is estimated by dividing the area of the cluster by the reference area. In S107,
The estimated number of people is counted. In S108, it is determined whether or not the next cluster exists. If there is a next cluster, the steps from S101 are repeatedly executed. As a result, the number of passers in the processing period is calculated for each of the ascending image 130A and the descending image 130B.

In FIG. 1, a division counting section 56 is means for estimating the number of persons by dividing the total area of the clusters in the up image 130A and the down image 130B by the reference area. Here, the reference area used in that case is determined from the width of the moving object and the staying time of the moving object in the X-axis window, as described above. Here, the staying time may be calculated for each cluster. However, in general, when the vehicle is crowded, it can be considered that the moving object is moving at a uniform speed in the same direction on the passage. Therefore, it is desirable to calculate the average moving speed of the cluster for each direction, and determine the reference area using the average moving speed.

FIG. 14 shows an example of the operation of the division counter 56. By the way, this operation is performed for the up image 130
This is executed for each of the A and the downstream image 130B.

In S201, the average moving speed is calculated as described above. That is, in the ascending image 130A and the descending image 130B, the average moving speed is calculated by averaging the moving speed obtained for each cluster. In S202, a reference area is calculated from the average moving speed and the reference width thus obtained. And S
At 203, the up image 130A or the down image 130B
Is calculated and the total area is divided by the reference area to calculate the number of passers in the processing period.

The counting result of the individual counting unit 54 and the counting result of the division counting unit 56 are input to the output control unit 58 in FIG. That is, the number of people passing by each direction is input to the output control unit 58 for each processing period. The output control unit 58 creates a table, a graph, and the like based on the number of people passing by each direction in each processing period,
It is output to the display 60. Of course, it may be configured to be printed out. Therefore, the display 60 displays the number of people passing through each processing period as a graph such as a numerical value or a histogram, and it is possible to automatically perform a traffic volume survey in each time zone on the stairs of the station, for example.

The processing period is set by the processing period control unit 26 as described above, and the above-described operations of the image processing and the number of people measurement are executed in units of the processing period.
That is, the data ranges of the X-T image 130 shown in FIG. 6, the Y-T image 140 shown in FIG. 10, and the directional map 150 shown in FIG. I have.

The number measuring device shown in FIG. 1 can also be constituted by connecting the image input unit 10 to a general-purpose computer, for example. In that case, the function of each configuration shown in FIG. 1 is substantially realized by software. Of course, for example, the XT image forming process and the YT image forming process can be executed in parallel,
Special hardware may be configured to enable such parallel processing.

The apparatus shown in FIG. 1 can be used for purposes other than counting people. For example, the present invention is applicable to the measurement of the number of articles flowing on a belt conveyor, the number of vehicles flowing on a highway, and the like. Various image processing conditions or measurement conditions may be appropriately set according to the nature of the measurement target.

According to the above-described embodiment, since there is no need to perform complicated parameter adjustment, there is an advantage that the number of people can be easily measured using, for example, a temporary camera. In addition, there is an advantage that the direction and speed of each person can be calculated by simple processing while paying particular attention to the movement trajectory of each person, and highly accurate person measurement can be realized by using them. Especially,
Real-time processing can be performed at low cost. Furthermore, since the counting method can be switched according to the congestion degree, there is an advantage that the appropriate number of people can be measured according to various situations.

[0087]

As described above, according to the present invention,
The number of moving objects can be accurately measured in consideration of the transition of the movement of each moving object. Further, according to the present invention, it is possible to appropriately measure the number of moving objects in consideration of the degree of congestion.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a preferred embodiment of a people counting apparatus according to the present invention.

FIG. 2 is a diagram showing a coordinate relationship in an observation area.

FIG. 3 is a conceptual diagram showing an X-axis window (a counting window).

FIG. 4 is a diagram showing the concept of a Y-axis window (direction / speed detection window).

FIG. 5 is a diagram illustrating a principle of forming an X-axis binary image.

FIG. 6 is an explanatory diagram showing an XT image.

FIG. 7 is a diagram for explaining a difference in appearance between a position immediately below a camera and a position displaced from the camera.

FIG. 8 is a diagram illustrating correction of a cluster according to a difference in appearance.

FIG. 9 is a diagram illustrating a principle of forming a Y-axis binary image.

FIG. 10 is an explanatory diagram showing a YT image.

FIG. 11 is a diagram showing a determination of a moving direction and a moving speed based on a cluster and a map for each direction.

FIG. 12 is a diagram showing an integration process of an XT image and a map for each direction.

FIG. 13 is a diagram illustrating an operation of an individual counting process.

FIG. 14 is a diagram illustrating an operation of a division counting process.

[Explanation of symbols]

10 image input unit, 12 cameras, 18 inter-frame difference calculation unit, 20 binarization processing unit, 22 threshold value setting unit, 26 processing period control unit, 30 parameter input unit,
32 X-axis binarized image forming unit, 34 X-T image forming unit,
36 congestion degree determination unit, 38 counting method selection unit, 40 Y
Axis binarized image forming unit, 42 YT image forming unit, 44 moving trajectory extracting unit, 46 direction discriminating unit, 48 speed discriminating unit,
50 map creating unit for each direction, 52 separating unit for each direction, 54
Individual counting unit, 56 division counting unit.

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Satoshi Watananami 8-10-16 Shimorenjaku, Mitaka-shi, Tokyo Inside Secom Co., Ltd.

Claims (11)

[Claims] 1. An X-axis binary image representing a moving object passing through an X-axis window set along the X-axis is sequentially acquired,
The X-axis binarized image at each time is aligned in the time axis direction to obtain a two-dimensional X
First image forming means for forming a T image; and a plurality of Ys arranged and set across the X-axis window.
For each axis window, a Y-axis binarized image representing the moving position of the moving object is sequentially acquired, and the Y-axis binarized images at each time are aligned in the time axis direction to form a two-dimensional YT image. 2, an image forming unit, a movement analyzing unit that analyzes at least one of a moving direction and a moving speed of each moving object based on the YT image, and the X- A moving object measuring device, comprising: counting means for counting the number of moving objects based on a T image. 2. An apparatus according to claim 1, wherein said imaging means captures an observation area including said X-axis window and said plurality of Y-axis windows, and a frame adjacent to each image acquired by said imaging means. A difference calculation means for performing a difference calculation between the two to obtain a difference image, and the X-axis binary image and the Y value based on the difference image.
A moving object measuring device, comprising: a binarized image creating unit that creates an axis binary image. 3. The apparatus according to claim 2, wherein the binarized image creating unit cuts out an X-axis window image corresponding to the X-axis window from the difference image, and converts the X-axis window image into an X-axis window image.
Means for creating the X-axis binarized image by projecting onto the axis and performing a binarization process; cutting out a Y-axis window image corresponding to each of the Y-axis windows from the difference image; Means for creating the Y-axis binarized image by projecting the image on the Y-axis and performing binarization processing. 4. The apparatus according to claim 1, wherein the movement analysis unit includes: a movement trajectory recognition unit configured to recognize a movement trajectory of each moving object in the YT image; A moving object measuring device, comprising: direction determining means for determining a moving direction of the moving object from a tilt direction. 5. The apparatus according to claim 4, wherein the counting unit is configured to classify an object image included in the X-T image according to the determined moving direction for each moving direction; A moving object measuring device, comprising: a direction-specific counting unit that counts the number of moving objects for each moving direction based on an X-T image. 6. The apparatus according to claim 1, wherein the movement analysis unit includes: a movement trajectory recognition unit configured to recognize a movement trajectory of each moving object in the YT image; A moving object measuring device for judging a moving speed of the moving object from a magnitude of the inclination. 7. The apparatus according to claim 6, wherein the counting unit variably sets a unit area based on the determined moving speed, and sets the total area of the object image in the X-T image as the unit. A moving object measuring device, comprising: division counting means for counting the number of moving objects based on the result of dividing by the area. 8. An X-axis binary image representing a moving object passing through an X-axis window set along the X-axis is sequentially acquired,
The X-axis binarized image at each time is aligned in the time axis direction to obtain a two-dimensional X
Image forming means for forming a -T image; congestion degree judging means for judging congestion degree based on the X-T image; and a counting method according to the judged congestion degree, based on the X-T image. And a counting means for counting the number of moving objects. 9. The moving object measuring device according to claim 8, wherein the congestion degree is determined based on a total area of the object images included in the XT image. 10. The apparatus according to claim 8, wherein the counting unit counts the number of object images included in the XT image when the congestion degree determination unit determines that the congestion degree is low. Individual counting means for calculating the number of moving objects by counting; and when the congestion degree determination means determines that the congestion degree is high, the total area of the object images included in the X-T image is divided by a reference area. And a division counting means for calculating the number of moving objects. 11. The moving object measuring apparatus according to claim 10, wherein the individual counting means and the division counting means count the number of moving objects for each moving direction.