JP2000053361A - Passenger monitoring device for man-conveyer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize operation of a man-conveyer without passenger residence by setting a residence detecting region for detecting passenger residence for an image inputted by an image input means, and by calling attention with a guide means when an object exists in the residence detecting region more than a designated time. SOLUTION: During operation of a man-conveyer, an image processing device 3a is started, takes in an image at a doorway with a camera 1a, sets a residence detecting region based on this image, and records the information. It photographs a background image to record it, and performs differential process between the present image and the recorded background image, and converts the differential process result to a binary image with an image conversion circuit 104. Using the binary image after noise removal, a histogram process circuit 105 creates x-axis and y-axis projection distributions. Then, it judges extent of traffic on the man-conveyer from area of a step region, calculates the in-region residence time, and when residence's occurrence is judged, performs attention guidance with a guiding device 22a.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a passenger monitoring device for a conveyor, and more particularly to a monitoring device for detecting, guiding, displaying, and controlling the behavior of passengers.

[0002]

2. Description of the Related Art Conventionally, this kind of technology is disclosed in
As disclosed in -319762, passengers on the conveyor and passengers near the entrance of the conveyor are imaged,
2. Description of the Related Art There is known an image in which images including passengers are arranged in chronological order, and a phenomenon such as a fall is detected by detecting the disturbance. Further, as described in JP-A-9-301664,
It is known to determine the number of passengers by image processing and to detect the stop of the passengers at the entrance / exit.

[0003]

The above prior art monitors the flow of passengers in a conveyor on the basis of image data, detects the occurrence of a fall accident from disturbances in the flow of passengers, and issues an alarm or the speed of the conveyor. It attempts to prevent the accident from spreading by controlling it, but does not consider the passengers getting out of the handrail. The latter is based on image processing, which determines the number of passengers and performs optimal passenger guidance for detecting a stop of the passenger at the entrance / exit, but does not disclose specific means for realizing the same. . Furthermore, the latter prior art does not consider the wandering of passengers near the entrance. SUMMARY OF THE INVENTION The present invention has been made in view of the above points, and has as its object to process a video from a camera, thereby enabling high-accuracy detection of stagnation such as prowl, standing talk, and mischief near an entrance / exit. It is to provide a device.

[0004]

In order to achieve the above object, a passenger monitoring apparatus for a passenger conveyor according to the present invention provides a stay detection area for detecting a stay of a person in an input image input by an image input means. Area setting means for setting
A time measuring means for measuring the elapsed time of the stay of the person detected by the area setting means, an object extracting means for extracting an object in the stay detection area, and an object extracting means for extracting the object extracted by the object extracting means for a predetermined time. The above is achieved by the configuration including the stay detection means for detecting presence in the stay detection area and the guide means for calling attention based on the detection result.

[0005]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is an overall configuration diagram of the present invention.

The cameras 1-a and 1-b are connected to a man conveyor 2-
The entrance and exit of a and 2-b are photographed. Assuming that it is a rising conveyor, the camera 1-a photographs the main body of the conveyor 2-a, the entrance of 2-a and the exit of 2-b, and the camera 1-b captures the main body of 2-b and 2-1-2. The images taken by the cameras 1-a and 1-b by photographing the entrance of b are input to the image processing devices 3-a and 3-b, respectively. The image processing devices 3-a and 3-b include an image processing unit 100 that processes the captured image, a CPU 110 that controls the entire apparatus and processes the result of the image processing unit 100, and interactively sets a processing area and the like. It comprises an interface unit 111 for exchanging information with the mouse 112 and the conveyor control device 21. Hereinafter, the operation of the image processing apparatus 3-a will be mainly described, but the operation of the image processing apparatus 3-b will be the same. The video taken by the camera 1-a is input to the image processing device 3-a. The image processing device 3-a takes in the captured image signal, converts it into density data of, for example, 256 gradations by the A / D converter 101, and stores it in the image memory 102. In the case of a color camera, the image data is stored in the image memory 102 for each of R, G, and B color components by three sets of A / D converters. Further, the image processing unit 100 uses the data of the image memory 102 to perform a spatial product-sum operation, an inter-image operation,
Binarization, shape conversion, histogram processing, normalized correlation operation, labeling (numbering), etc., are respectively performed by the spatial product-sum operation circuit 115, the inter-image operation circuit 103, and the pixel conversion circuit 1.
04, shape conversion circuit 116, histogram processing circuit 10
5, normalized correlation operation circuit 117, labeling circuit 118
And the processing result, the input image, etc.
The video signal is converted into a video signal by the A
8 is displayed. The histogram memory 106 is a memory for storing a projection distribution, a density frequency distribution, and the like of an image. The CPU 110 uses the memory 109 to store the image
Based on the processing result of 0, the presence / absence of a passenger embarking event is determined, and information of the event occurrence is transmitted via the interface unit 111 to the conveyor control device 21-a or 21-b where the event has occurred. The man-conveyor control device 21 stops the man-conveyor and performs guidance and display by the guidance device 22 according to the content of the sent event.

FIG. 2 is an explanatory diagram of stay detection according to the present invention.
The figure which looked at the escalator main body and the vicinity of the entrance / exit from the camera which set the center axis of the visual field of the camera near the center between the handrails on both sides of the escalator and above the entrance / exit is shown. There are a left handrail 23-L and a right handrail 23-R of the escalator, a floor board 27 and a tread board 28 which constitute a platform and an exit of a man conveyor. The passenger 20 is moving on the floorboard along with the shadow 25 as a movement trajectory 29.
The presence or absence of the shadow 25 depends on the surrounding lighting conditions.

In order to detect the stay of passengers, the installer or the observer of the present monitoring apparatus uses a mouse 112 or the like to set a stay detection area 24 near the floor plate 27 at the entrance / exit.
A step area 30 is set near the step (stepboard) 28. Note that the stay detection area 24 and the step area 3
0 can process the captured image and automatically set it from the area information of step (stepboard) 28.

FIG. 3 is a processing flow chart showing the operation of one embodiment of the present invention. First, when the image processing apparatus 3 is started, initialization such as clearing of the image memory 102 and the histogram memory 106 is performed (200), and an image is captured from the camera 1 (205). Using the mouse 112 on the basis of this image, for example, the operator interactively sets the stagnation detection area 24, or the apparatus automatically sets the stagnation detection area 24 based on the position information and the like, and stores the information. It is stored in the memory 109 (210). Next, the background image is captured, stored in the image memory 102 (215), and monitoring is started. Note that a background image without a person or the like can be automatically generated by repeatedly processing an input image.

After the start of monitoring, first, the current image is fetched (220), and a difference process is performed between the image and the background image stored in the image memory 102 in advance.
3, the result of the difference is converted into a binary image composed of "1" and "0" values by the pixel conversion circuit 104, and noise is removed by contraction / expansion processing using the shape conversion circuit 116 (225). ). Further, at time t-1 (previous time) of the step area 30 to obtain the traffic volume of the conveyor.
Processing (inter-frame difference processing) between the two images at time t and the current time (current time) is executed by the inter-image calculation circuit 103, and the difference result is expressed by the pixel conversion circuit 104 as “1” and “0”. The image is converted into a binary image composed of the following, and noise is removed by contraction / expansion processing using the shape conversion circuit 116 (230).

Hereinafter, a stay detection information generation process 240 and a stay determination process 250 which are the core of the present invention will be described with reference to FIGS.
Explanation will be made using 0.

In order to detect a stagnant person such as prowl, standing talk, mischief, etc., it is sufficient to track each object and determine whether or not the object is present in the stagnant detection area for a specified time or more. Since the object is relatively narrow and close to the camera, the size of the object greatly changes depending on the posture of the person or the like, and when the flow of the person becomes crowded, the persons overlap in the field of view of the camera. Therefore, it is difficult to accurately track individual objects. In addition, the wandering of the passenger occurs when the traffic volume is small, but it is considered that it does not occur when the traffic volume is large because the passenger cannot prowl. When the traffic volume is large, there is a possibility that pranks and standing talks may occur in places such as the outside of the handrail that do not significantly affect the flow of people. In this naughty or standing story, it is assumed that people rarely change position. It is relatively easy to detect staying when traffic is light, but it is not easy to detect staying when traffic is heavy. Therefore, in order to detect the stay when the traffic volume is large, it is conceivable to divide the stay detection area and change the target area of the stay detection processing according to the traffic volume.

FIG. 4 shows that the stay detection area 24 is replaced with the stay detection area W.
(Whole) 24-W, a stay detection area L near the handrail 23-L (left) 24-L, and a stay detection area R near the handrail 23-R (right) 24-R are shown.

FIG. 1 shows an embodiment in which the stay detection area 24 is divided into a plurality of areas, and the target area of the stay detection processing is changed according to the traffic volume, so that the stay detection can be performed accurately even in a congested state or in a normal state. 5 will be described.

First, the histogram processing circuit 105 uses the binary image after noise removal obtained in step (225).
Then, the x-axis projection distribution and the y-axis projection distribution shown in FIG. 6 are created (242). The x-axis projection distribution and the y-axis projection distribution are stored in the histogram memory 106. The spatio-temporal overlap information creation (244) is realized as follows. A portion where the frequency of the x-axis projection distribution and the y-axis projection distribution stored in the histogram memory 106 is equal to or larger than the threshold th-h is extracted and stored in the spatiotemporal overlap information table 51;
Spatio-temporal overlap information is created using the table 51.

The spatiotemporal overlap information creation (244) will be specifically described with reference to FIG. FIGS. 7A and 7C show time t−
The frequency of the x-axis projection distribution at 1 and time t is a threshold
The portion over th-h is schematically shown, and the horizontal axis represents x-axis coordinates. FIG. 7B shows the spatiotemporal overlap information tables 51-a and 51-b corresponding to FIG. 7A. FIG.
As shown in FIG. 7A, assuming that the object of label 1 exists at the position of coordinates x3 to x4 and the object of label 2 exists at the position of coordinates x5 to x6 at the current time t, as shown in FIG.
As shown in (b), the spatiotemporal overlap information table 51-b
The start position of label 1 is x3, and the end position is x
4. The appearance time is t, the existence start position of the label 2 is x5, the existence end position is x6, and the appearance time is t. On the other hand, at the previous time t-1, the object of the label 1 is moved from the coordinate x1 to the coordinate x2.
As shown in FIG. 7B, the existence start position of the label 1 in the spatiotemporal overlap information table 51-a is x1, the existence end position is x2, and the appearance time is t-, as shown in FIG.
It is 1. The two sets of tables 51-a, 51-
The spatio-temporal overlap information is updated using b. With reference to the start position and end position of the object in the spatiotemporal overlap information table 51-b at the current time t, refer to the start position and end position of the object in the spatiotemporal overlap information table 51-a at the previous time t-1 for each label as a reference. And check for overlap. When the position overlaps with the object at the previous time as in label 1, the appearance time of the object in the table 51-a at the previous time is referred to and compared with the appearance time of the table 51-b at the current time. Then, the old time t-1 is written in the appearance time column. When the position does not overlap with the object at the previous time as in the label 2, the current time t becomes the appearance time as it is. As described above, temporal and spatial overlap information of the object is created. Note that time t
FIG. 7C shows the correspondence between objects at −1 and at time t.
Can be classified into any of appearance, continuous, confluence, separation, and disappearance. Next, the histogram processing circuit 10
5 is used to determine the area of the object for each stay detection area (2
46).

Next, the stay determination process (250) will be described. First, the step area 30 obtained in step 230
Is smaller than or equal to the threshold th-as, the traffic of the conveyor is determined (251). In a normal case where the traffic volume is not so large, referring to the appearance time of the spatiotemporal overlap information table 51 generated in step (244) for the entire stay detection area (W), and using the current time to stay in the area The time is calculated, and it is determined whether the time is equal to or longer than the specified time th-tW (252). If the time is equal to or longer than the specified time th-tW, the area equal to or larger than the threshold value th-apW is further determined using the area information obtained in step (246).
It is determined whether it is not less than W (253). Threshold th-
If the area equal to or larger than apW is equal to or larger than the designated time th-tW, it is determined that a stagnation has occurred, the passenger is cautioned by a microphone or the like (260), and the process returns to step (220) to perform the above-described processing. repeat. On the other hand, if it is less than the designated time, it is determined that no stagnation has occurred, and step (22)
Returning to 0), the processing described above is repeated.

On the other hand, when there is a large traffic volume, such as an escalator at a station at the time of commuting, where passengers come continuously, the above-mentioned spatiotemporal overlap information and area information are used for the entire stay detection area. If the occurrence of stagnation is determined, an erroneous determination will be made. Also, if the traffic volume is large,
It is very unlikely that stagnation will occur near the center of the vehicle, and even if it occurs, it will be near the handrails on both sides. Therefore, when the traffic volume is large, the processing is performed by limiting the stay detection area to be determined to the stay detection areas L and R near the handrails on both sides. That is,
With reference to the appearance time in the spatiotemporal overlap information table 51 generated in step (244) for the stay detection area L and the stay detection area R, the stay time in the area is calculated using the current time, and the time is designated. It is determined whether or not the time is equal to or longer than the time th-tL (254). If it is equal to or longer than the specified time th-tL, it is further determined whether or not the area equal to or larger than the threshold value th-apL is equal to or longer than the specified time th-tL using the area information obtained in step (246) (255). If the area equal to or larger than the threshold value th-apL is equal to or larger than the designated time th-tL, it is determined that stagnation has occurred, and caution is given to the passenger with a microphone or the like (2).
Along with 60), the process returns to step (220) and repeats the above-described processing. On the other hand, if it is less than the designated time, it is determined that no stagnation has occurred, the process returns to step (220), and the above-described processing is repeated.

Next, as another embodiment, the stay detection area 2
4 is divided into a plurality of areas, the target area of the stagnation detection processing is changed according to the traffic volume, and the processing content is changed for each target area, so that the stagnation detection can be accurately performed even in a congested state and in a normal state. An example is shown in FIG. FIG. 8 shows a stay detection information generation process 340 corresponding to the stay detection information generation process 240 and the stay determination process 250 in FIG.
Only the stay determination processing 350 is shown, and the other parts in FIG.

In the stagnation detection information generation process 340, the x-axis projection distribution and the y-axis projection distribution of the stagnation detection area are created (34).
2), creation of spatio-temporal overlap information (344) and extraction of the area of the stagnation detection area (346) are described above. Creation of the x-axis projection distribution and y-axis projection distribution of the stagnation detection area in FIG.
This process is the same as the process of creating spatiotemporal overlap information (244) and extracting the area of the stay detection area (246).

The difference from FIG. 5 is that the binary overlap information obtained in step (225) is the spatial overlap information of objects that are not completely continuous in time but are almost continuously spatially overlapped. Is obtained by accumulating (348)
That is.

The image accumulation will be described with reference to FIG.
An image 63 is obtained by adding 62 the binary image 60 obtained in step (225) and the cumulative image 61 stored in the image memory 102 in advance using the inter-image calculation circuit 103. On the other hand, the binary image 60 is inverted 64 using the pixel conversion circuit 104 to obtain an inverted image 65. Inter-image calculation circuit 1
03 is used to subtract 66 the inverted image from the image 63 66
Then, an image 67 is obtained, and the inverted image 6 is obtained from the image 67.
5 is subtracted 66 to obtain a cumulative image 61. By the above processing, the object portion is incremented by 1, and the portion other than the object is incremented by -2. According to this processing, the cumulative value gradually increases when the object portion is continuously present at substantially the same position. However, the cumulative value of the moving object or the portion of the noise generated at random hardly increases, and the cumulative value becomes zero. . The value of n when -n is applied to a portion other than the object depends on the traffic density of the moving object, but n is preferably 1 to 3.

Next, the stay determination process (350) will be described. First, the step area 30 obtained in step 230
Is smaller than or equal to the threshold th-as, it is determined whether the traffic volume of the conveyor is small (351). In the normal case where the traffic volume is not so large, the current time is determined by referring to the appearance time of the spatiotemporal overlap information table 51 generated in step (344) for the entire stagnation detection area (W) as in FIG. Is used to calculate the residence time in the area, and the time is calculated as the designated time th−
It is determined whether it is not less than tW (352). Specified time
If it is not less than th-tW, it is further determined whether or not the area not less than the threshold value th-apW is not less than the specified time th-tW by using the area information obtained in step (346) (353).
If the area equal to or greater than the threshold value th-apW is equal to or greater than the specified time th-tW, it is determined that a stagnation has occurred, the passenger is cautioned with a microphone or the like (260), and the process returns to step (220). The above-described processing is repeated. On the other hand, if it is less than the designated time, it is determined that no stagnation has occurred, and the process returns to step (220) and repeats the above-described processing.

On the other hand, when there is a large traffic volume such as an escalator at a station at the time of going to work, the possibility of stagnation near the center of the floorboard 27 is extremely low.
Even if it occurs, it is considered to be near the handrails on both sides. In addition, when the traffic volume is large, it is not possible to wander, and it is considered that the staying person is continuously present at almost the same position. Therefore, when the traffic volume is large, the stagnation detection area to be determined is limited to the stagnation detection area L and the stagnation detection area R near both handrails. (Spatial overlap information)
Is used to determine whether or not it is staying. That is, step (348) is performed for the stay detection area L and the stay detection area R.
Using the accumulated image (spatial overlap information) obtained in
It is determined whether an object having an area not less than th-aaL exists for not less than a specified time th-taL (354). If the object having the area equal to or larger than the threshold value th-aaL is equal to or larger than the specified time th-taL, the area equal to or larger than the threshold value th-apL is further specified using the area information obtained in step (346). It is determined whether or not −tL or more (355). If the area equal to or greater than the threshold value th-apL is equal to or greater than the specified time th-tL, it is determined that a stagnation has occurred, and the passenger is cautioned with a microphone or the like (260), and the process returns to step (220). The above-described processing is repeated. On the other hand, if it is less than the designated time, it is determined that no stagnation has occurred, and the process returns to step (220) and repeats the above-described processing.

The above description is based on a plurality of stay detection areas 2
4 is set in advance, and the processing target area 24
Have been described, and the processing contents are changed according to the processing target area 24. However, considering that the flow of passengers (traffic flow) differs depending on the installation location of the conveyer, an area 24 that is hardly affected by continuous passengers is individually set depending on the installation location of the conveyer. It is also considered difficult to optimally set the area 24 that is not significantly affected by continuous passengers. Therefore, as another embodiment, the stay detection area 24 is fixed,
FIG. 10 shows an example in which the processing content is changed depending on the traffic flow. The entire processing flow is the same as that in FIG. The difference from FIG. 8 is that, as shown in FIG. 11, the processing content is changed depending on the traffic amount for the stagnation detection area 24 set at one place near the entrance of one man conveyor. In particular,
The staying detection information generation process 440 is the same as the staying detection information generation process 340 in FIG. 8, and steps (442) to (448) are performed in steps (342) to (3) in FIG.
This is the same processing as 48).

Next, the stay determination process (450) will be described. First, the step area 30 obtained in step 230
Is smaller than or equal to the threshold th-as, the traffic volume of the conveyor is judged (451). In a normal case where the traffic volume is not so large, referring to the appearance time of the spatiotemporal overlap information table 51 generated in step (444) for the entire stagnation detection area 24 as in FIG. Calculate the residence time in the area, and calculate the time
It is determined whether or not it is not less than W (452). Specified time th
If it is not less than -tW, it is further determined whether or not the area not less than the threshold value th-apW is not less than the specified time th-tW using the area information obtained in step (446) (45).
3). The area over the threshold value th-apW is the specified time th-tW
If it exists, it is determined that a stay has occurred, and a cautionary guidance is given to the passenger with a microphone or the like (260), and the process returns to step (220) to repeat the above-described processing. On the other hand, if it is less than the designated time, it is determined that no stagnation has occurred,
Returning to step (220), the processing described above is repeated.

On the other hand, when there is a large traffic volume such as an escalator at the station at the time of commuting, the passengers come continuously, and the accumulated image (spatial overlap) obtained in step (448) for the entire stay detection area 24 is targeted. Information) using the threshold th-aaL
It is determined whether an object having the above area exists for the specified time th-taL or more (454). If the object having the area equal to or greater than the threshold value th-aaL is equal to or greater than the designated time th-taL, the threshold value is further determined using the area information obtained in step (446).
It is determined whether the area not less than th-apL is not less than the designated time th-tL (455). If the area equal to or larger than the threshold value th-apL is equal to or larger than the designated time th-tL, it is determined that the stagnation has occurred, and the passenger is cautioned with a microphone or the like (26).
With (0), the process returns to step (220), and the above-described processing is repeated. On the other hand, if it is less than the designated time, it is determined that no stagnation has occurred, and the process returns to step (220) and repeats the above-described processing.

When a cumulative image is obtained in the whole area 24 to detect a staying person, a state where passengers are always present occurs when traffic is extremely large, such as when leaving or leaving a station, and the cumulative image is thresholded. If the value exceeds the value and stagnation does not occur, the possibility of erroneously detecting that stagnation has occurred increases. Therefore, in order to distinguish between a moving continuous passenger and an almost stationary occupant, attention is paid to the fact that the brightness and color of the passenger's clothes are different from each other, and a binary image, which is input information of an accumulated image, is used as a background. An accumulated image is obtained not for an image obtained by binarizing the absolute value of the difference result (an image in which the entire object region is extracted) but for an area having a different brightness from a background area or an area having a different color. In this case, the area of the stagnant who is almost stationary is accumulated, but the area of the moving continuous passenger is not continuously extracted due to the difference in clothes, so the effect of -n described in FIG. Are hardly accumulated. Specifically, the difference in brightness will be described as an example. The background subtraction result is classified into three regions, a background region, a region brighter than the background (+), and a region darker than the background (-), and is brighter than the background (+).
The region and the (−) region darker than the background are separately accumulated, and separate accumulated images are obtained. Then, using the (+) cumulative image and the (-) cumulative image, it is determined whether an object having an area greater than or equal to the threshold value th-aa exists for a specified time th-ta or more, and one of the cumulative images is determined. If the conditions are satisfied, it is determined that there is a staying person, and an attention guide (260) is given.
If both accumulated images do not satisfy the determination condition, it is determined that there is no resident, and the process returns to step (220) to repeat the process. In the case of color, a cumulative image is obtained for each of several color ranges, and a staying person is detected in the same manner as described above.

Here, another method of measuring the traffic volume will be described. The binary image of the step area 30 obtained in step (230) is input to the labeling circuit 118, and the number of objects is obtained by numbering the objects. You can also judge that there are many. Further, the binary image of the stay detection area 24 obtained in step (225) is input to the labeling circuit 118 instead of the step area 30, and the number of objects is obtained by numbering the objects. th-
When n is greater than or equal to n, it can be determined that the traffic flow is large. Further, the stagnation detection area 2 obtained in step (225)
4 is input to the histogram processing circuit 105,
The area of the object is obtained, and when the area is equal to or larger than the threshold th-a, it can be determined that the traffic flow is large. Further, when both the area of the area and the number of objects are each equal to or larger than the threshold value, it can be determined that the traffic flow is large.

Further, as a method of extracting the object region of the stagnation detection region 24, there is a method of using an inter-frame difference of the object region of the stagnation detection region 24 instead of the method using the background difference described in step (225). Here, FIG.
2, a method of using a normalized correlation operation will be described. The normalization correlation operation itself is described in the literature (supervised by Hideyuki Tamura: Introduction to Computer Image Processing, pp. 149-15)
0, Soken Shuppan Co., Ltd. H6 / 4). First, the stagnation detection area 24 is divided into a plurality of mesh areas 40, and the mesh area 4 of each of the mesh areas 40 of the background image 42 and the input image 43 is determined using the normalization correlation circuit 117.
A normalized correlation value between 0-a and 40-b is determined. The background portion has a high correlation value because the pattern is substantially the same between the background image 42 and the input image 43, and the object portion 20 has a low correlation value since the pattern differs between the background image 42 and the input image 43. The brightness of the shadow portion 25 of the object differs between the background image 42 and the input image 43, but the pattern is almost the same, so that the correlation value increases. Therefore, an area where the normalized correlation value is smaller than the threshold value th-cr is a binary image 4 shown in FIG.
The object area is represented by the black area 41 of FIG. The binary image 45
As can be seen from the above, it is possible to extract the object region excluding the shadow by the normalized correlation operation. By making the mesh finer, the object region represented by the black region 41 can be made closer to the shape of the object 20, but the processing time slightly increases. FIG. 12C shows the input image 4 for reference.
Binary image 4 obtained by binarizing the difference between 3 and background image 42
4 is shown.

According to the present invention, a staying object can be extracted with high accuracy by extracting an object using a background difference or an inter-frame difference of a stay detection area or a normalized correlation operation. Further, by detecting the staying object using the area information of the extracted object, the staying object can be detected with high accuracy.

Furthermore, by detecting the staying object using the spatial overlap information of the extracted object, the staying object can be detected with high accuracy.

Furthermore, spatial overlap information (cumulative image) of the extracted object is obtained based on information on the difference in brightness or color from the background image of the object, and a staying object is detected using the information. Thus, the staying object can be detected with high accuracy.

Furthermore, by detecting the staying object using the spatiotemporal overlap information of the extracted object, the staying object can be detected with high accuracy.

Still further, by changing a stagnation detection area to be processed with respect to a plurality of stagnation detection areas set in advance by the area setting means according to the traffic volume of the conveyor, and detecting a stagnant object, Accurate detection of staying objects.

Further, by obtaining the traffic volume of the man conveyor using the area information of the step area and / or the area information of the stay detection area, the staying object can be detected with high accuracy.

Furthermore, by obtaining the traffic volume of the man conveyor using the area information and the number of objects of the step area or the area information and the number of objects of the stay detection area, the staying object can be detected with high accuracy.

Furthermore, by changing the processing contents of a plurality of stagnation detection areas set in advance by the area setting means according to the stagnation detection areas to be processed and detecting the stagnated objects, the stagnation object is detected with high precision. Objects can be detected.

Furthermore, by detecting the staying object by changing the processing contents of the stay detection means according to the traffic volume, the staying object can be detected with high accuracy for the entire stay detection area preset by the area setting means. it can.

[0040]

According to the present invention, an object can be extracted with high accuracy by extracting an object using the background difference or inter-frame difference of the stay detection area or the normalized correlation operation.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram of the present invention.

FIG. 2 is an explanatory diagram of the present invention.

FIG. 3 is a flowchart showing the flow of a monitoring process according to the present invention.

FIG. 4 is a diagram illustrating an example of setting a step area and a plurality of stay detection areas.

FIG. 5 is a flowchart showing a flow of processing for generating stay detection information and determining stay;

FIG. 6 is an explanatory diagram of creating an x-axis projection distribution and a y-axis projection distribution.

FIG. 7 is an explanatory diagram of creating spatiotemporal overlap information.

FIG. 8 is a flowchart showing a flow of processing for generating stay detection information and determining stay based on another embodiment of the present invention.

FIG. 9 is an explanatory diagram of creating spatial overlap information (cumulative image).

FIG. 10 is a flowchart showing a flow of processing of stay detection information generation and stay determination according to another embodiment of the present invention.

FIG. 11 is a diagram illustrating an example of setting one stay detection area.

FIG. 12 is an explanatory diagram of object extraction by a normalized correlation operation.

[Explanation of symbols]

1-a, 1-b camera, 2-a, 2-b man conveyor, 3-a, 3-b image processing device, 21-a, 21-
b: man conveyor control device, 22-a, 22-b: guide device, 23: handrail, 24: stay detection area, 10
0: image processing unit, 108: monitor, 109: memory, 1
10 CPU, 111 interface unit, 112 mouse.

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Tadaaki Kitamura 7-1-1, Omikacho, Hitachi City, Ibaraki Prefecture Inside Hitachi, Ltd. Hitachi Research Laboratory, Ltd. (72) Inventor Shigehisa Sakimura 7-1, Omikamachi, Hitachi City, Ibaraki Prefecture No. 1 Inside Hitachi, Ltd.Hitachi Research Laboratories (72) Inventor Toshimitsu Tobita 1070 Ma, Hitachinaka City, Ibaraki Prefecture Inside Mito Plant, Hitachi Ltd. (72) Inventor Kenichi Yamashita 1070 Ma, Hitachinaka City, Ibaraki Stock Inside the Mito Plant of Hitachi, Ltd. (72) Inventor Hisao Chiba 1070, Ma, Hitachinaka-shi, Ibaraki F-term in Mito Plant of Hitachi, Ltd.F-term (reference) DC04 DC14 DC23 DC25 DC32 DC34

Claims (9)

[Claims] 1. A passenger monitoring device for a conveyer for photographing a passenger on a conveyer and monitoring the condition of the passenger using an image input by an image input unit for inputting an image including the passenger, wherein the image input unit Area setting means for setting a stay detection area for detecting stay of a person in an input image input; time measuring means for measuring elapsed time of stay of a person detected by the area setting means; Object extraction means for extracting an object in the area, stay detection means for detecting that the object extracted by the object extraction means is present in the stay detection area for a predetermined time or more, based on the detection result A passenger conveyor monitoring device for man-conveyors that has guidance means to call attention. 2. The method according to claim 1, wherein said object extracting means comprises:
A passenger monitoring device of a man conveyor that extracts an object using a background difference, an inter-frame difference, or a normalized correlation operation. 3. The apparatus according to claim 1, wherein the stay detection means comprises:
A passenger monitoring device of a man conveyor for detecting a staying object using area information of the extracted object extracted by the object extracting means. 4. The apparatus according to claim 1, wherein the stay detection means comprises:
A passenger monitoring device of a man conveyor that detects a staying object by using spatial overlap information of an extracted object extracted by the object extracting means. 5. The stagnation detecting means according to claim 4,
A passenger monitoring device for a man-conveyor that obtains spatial overlap information of an extracted object based on information on a difference in brightness or a color of an object from a background image, and detects a staying object using the spatial overlap information. 6. The apparatus according to claim 1, wherein the stay detecting means includes:
A passenger monitoring device of a man conveyor that detects a staying object using spatiotemporal overlap information of the extracted object extracted by the object extracting means. 7. The method according to claim 1, wherein the stay detecting means includes:
A passenger conveyor monitoring device for a man conveyor that changes a stay detection region to be processed according to a traffic volume of the man conveyor, for a plurality of stay detection regions set in advance by the region setting means. 8. The apparatus according to claim 1, wherein the stay detecting means includes:
A passenger monitoring device of a man-conveyor that changes the processing content according to a stay detection area to be processed, for a plurality of stay detection areas preset by the area setting means. 9. The method according to claim 1, wherein the stay detecting means includes:
A passenger monitoring device of a man-conveyor that changes the processing content according to the traffic volume for the entire stay detection area preset by the area setting means.