JPH0686295A - Surveillance camera equipment - Google Patents

Abstract

PURPOSE:To surely record the face of an entering person, and to improve a crime prevention function and a crime deterrent effect by opening a door by a door opening/closing device, only when it is decided that the face is visible. CONSTITUTION:An image is inputted from a TV camera 1, a human body is detected from in the image, and the part of a face is fetched. If the part of the face is detected, a decision processing for whether the face is visible or not is executed. As a result, when it is decided that the face is visible, a signal for opening a door is sent to a door switch 5, the door is opened, and an image in which the face is reflected is recorded in an image recorder 7. On the other hand, in the case the part of the face cannot be detected, and it is not decided that the face is visible, the number of times of repetition of a repeat counter is checked, and when its number of times is smaller than a prescribed value, a voice is synthesized by a speech synthesizing device 8 and outputted to a loud-dpeaker 9. Also, a voice is uttered so as to take off a hat, a helmet or sunglasses, so that the face part can be detected.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a surveillance camera.

[0002]

2. Description of the Related Art A large number of surveillance cameras have been installed for crime prevention in financial institutions such as stores and banks. They are free to enter the store and take pictures of the people who enter. However, in general, criminals try to prevent their faces from being seen, and in many cases the face of the subject is not clearly captured by the camera, so they do not play a role as surveillance cameras.

[0003]

As described above, the conventional surveillance camera has a crime deterrent effect, but is not sufficiently effective when a crime actually occurs. The present invention has been made in view of the above conventional problems, and an object of the present invention is to provide a surveillance camera device that accurately performs the role of a surveillance camera.

[0004]

In order to solve the above-mentioned problems, the present invention has a door opening / closing means, an image inputting means, a face part detecting means, and a means for judging whether a face is clearly visible or not. It is a camera device.

[0005]

According to the present invention, the face portion is obtained from the input image, and it is determined whether or not the face is visible in the portion.
The door is opened by the door opening / closing means only when it can be determined that the face is visible. Furthermore, all the faces of people who enter inside will be clearly recorded, and in the event of a crime, an image of the face of the criminal will be obtained.

[0006]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 shows an embodiment of the present invention.
This system comprises a television camera 1, an A / D converter 2, a microprocessor circuit 3, an automatic door sensor 4, a door opening / closing device 5, a door 6, an image recording device 7, a voice synthesizing device 8 and a speaker 9. Various determination processes are realized by the microprocessor circuit 3. Hereinafter, the details of each unit will be described along with an operation example of the present invention.

Door 6, automatic door sensor 4, door switch 5
Constitutes a commonly used automatic door. Any type of automatic door may be used.
However, when the person 10 stands outside the door to enter the target space, even if the automatic door sensor 4 detects the person, unlike the normal automatic door, the door 6 is directly passed through the door opening / closing device 5.
Will not open.

When the automatic door sensor 4 detects an outside person, the signal is sent to the microprocessor circuit 3.
FIG. 2 shows a flowchart of the processing in the microprocessor circuit 3.

First, the repetitive counting counter is reset (set to 0) for the reconfirmation process described later (step 201). Then, an image is input from the TV camera 1, a person is found in the image, and a face portion is taken out (step 202). The processing method of this part by the microprocessor will be described later. If the face portion is detected (YES in step 203), it is determined whether or not the face is visible (step 204). The processing method of this portion will also be described later. If it is determined that the face is visible (step 2
(05 YES), a signal for opening the door is sent to the door opening / closing device 5, and the door 6 is opened (step 206). And
The image showing the face is recorded in the image recording device 7. The recording device may be an analog video tape recording device, a digital image recording device, or any device.

If the face portion cannot be detected (NO in step 203) or if it is not determined that the face is visible (NO in step 205), the repeat counter is checked first. If the counter is less than a predetermined number (for example, if the number of repetitions is only one, the predetermined number is 1,
Therefore, if the counter is 0) (step 208 YES),
The voice synthesizer 8 synthesizes a voice and outputs it to the speaker 9 (step 209). The content of the voice is, for example, "I'm sorry
I'm sorry if you can see the person with the mark, but if you wear sunglasses, can you remove it? " In this embodiment, as shown in FIG. 3, as the television camera 1, a small camera is mounted so as to be embedded in the door. The attachment part is arranged in or near a mark) (for example, in the case of a shop, the mark of the shop). Then, the repetitive counting counter is incremented by one (step 210), and the process returns to the face part detecting process. Instead of the voice synthesizer 8, the voice recorded on the tape or the memory may be reproduced. As a case where a face or the like cannot be detected, even if a person is detected by the automatic door sensor 4, there is a case where the person does not intend to go inside and immediately goes there. In such cases,
If the face is not detected and the instruction of the clerk described below is moved to the processing for opening the clerk, the burden on the clerk is increased. Therefore, in order to prevent this, there is a person from the automatic door sensor 4 during the processing in the microprocessor circuit 3. If there is no such signal, the process is interrupted, the repeat counting counter is reset, and the initial state is restored. Since this does not enter the flow of the flowchart, it is written on the outside of the flow as (step 299) in FIG.

When the repetition count counter is equal to or more than the predetermined number (step 208 NO), an alarm sound is output to the speaker 11 under the capacity 13 (step 211). Clerk 1
Reference numeral 3 determines whether or not the user is a suspicious person by looking at the image of the TV camera 1 or another TV camera 14 on the monitor 14 at hand, or by directly looking at the door if the environment allows it. That is, it is determined whether the person outside the door intentionally makes the face hard to see or the processing of the device is incorrect. Then, the determination result is transmitted to the device by the hand switch 15 (step 212). When it is determined that the device has made an error (step 212A), the door 6 is opened (step 206). At the same time, an instruction to record the image of the person outside the door is issued and recorded in the recording device 7 (step 20).
7) and the process ends. FIG. 4 shows an example of the hand switch 4.

If no person is present when the clerk looks at the monitor or the outside, the reset button is pressed with the switch at hand (step 212B). In this case, without opening the door,
The process ends. This happens when there is something wrong with the process.

When there is a suspicious person whose face is hidden, the clerk first presses the alarm button of the switch 15 at hand.
Press once (step 212C). Then reconfirm. For this, the voice synthesizer 8 may be operated again, or a clerk may use the microphone 16 to output a sound to the speaker 9 through the amplifier 17 and speak to an outside person, or may speak directly through the door. . If it is determined that he is not a suspicious person, he instructs to open the door with the hand switch, and goes to step 212A (step 213A). If it is still not possible to judge, or if it is judged to be suspicious without reconfirmation, the reconfirmation process is skipped and the alarm button is pressed again (step 213B). If this is done, it is safe to send an emergency call to the security company or the police (step 214). It is even safer to tell the security company that pressing the first button on the alert may be urgent. For example, if the alarm button is pressed once, but then the button for opening the door is not pressed for a while, there is a possibility of emergency. This may be done so that the operation status of the button is transmitted to the security company, or simply, if the alarm button is operated once and then the other buttons are not pressed for a while, the alarm button will be activated. The switch may be made so that it is treated as if it were pressed twice. It should be noted that in case the alarm button is pressed by mistake, the reset button should be pressed to return to the initial state.

Regarding the number of repetitions, the predetermined number may be zero.
In this case, there is no confirmation by voice or the like, and the process proceeds immediately by the judgment of the clerk. It is convenient to be able to easily change the number of repetitions according to the customer's visit frequency and time. For example, in FIG. 4, it is possible to allow the clerk to change the setting by using a push button under the opening / closing door in the switch. It should be noted that the illumination 18 is attached as necessary so that the target image can be properly input.

In the above description of the embodiment, a store such as a convenience store is taken as an example. However, in the case of an unmanned store (automatic deposit withdrawal machine) of a financial institution, etc., there is no store clerk nearby, so the part of the store clerk's judgment The image of the TV camera 1 or the monitor 14 may be sent to a security company or the like, and the determination may be performed there. Although the overall operation of the embodiment has been described above, the processing of the face part detection and the part for determining whether the face is clearly visible, which is omitted in the above description, will be described below.

First, the processing in the face part detection (step 202) will be described with reference to FIG. Automatic door sensor 4
When the signal of is received, the image is taken by the television camera 1 at two adjacent times, and the digital image is input through the A / D converter 2 (steps 501 and 502). FIG. 6A shows an example of an image at a temporary point for explanation. Then, the absolute value of the difference is obtained for each pixel of both images (step 503). Since the target person is moving a little between the two points,
As shown in (b), a large difference occurs around the contour.
And the part where the absolute value of the difference is larger than an appropriate threshold is 2
It is extracted by digitization (step 504). In the following description, the point extracted by binarization is set to the pixel value 1 for convenience of description.

Next, the projection on the X axis is obtained for the binarized result (step 505). That is, the number of pixels in the binarized portion is counted for each vertical column of the image.
Then, as shown in FIG. 7A, a mountain is formed in the projection result in the part where the person is present, so that both ends thereof are detected (step 506). Specifically, a continuous portion whose projection value is equal to or larger than an appropriate threshold value is obtained from the projection. On this occasion,
Even if there is a small gap (point below the threshold value), it may be determined to be continuous. If there is no such mountain over a certain width (NO in step 507), the result that the human being cannot be detected and therefore the face cannot be detected is output (step 508), and the process ends.

If there is a mountain (step 507YE
S), as shown in FIG. 7B, the image is examined laterally line by line between both ends of the mountain, and the leftmost 1 position and the rightmost 1 position are examined.
Is calculated, and the difference between the positions is determined as the width of the target object (step 509). In addition, after the width reaches the both ends in the lower portion, the width is set between the both ends.

Then, the width is checked from above to find the portion where the value changes (step 510). For an ordinary person, the width is as shown in FIG. Therefore, an appropriate position (B in FIG. 8) before the width begins to increase is determined, and the position between the upper end and A is set as the face position. However, if the upper end is examined from the top and the portion where 1 appears first is used as it is, it is easily affected by noise, so it is set to the first position where 1 portion appears continuously after that.

In the case of a person with long hair, the width value does not increase from the remarkable minimum value as shown in FIG.
As in (b), the width may change gently. In this case, an appropriate position C just before the width is saturated
Is the bottom of the face. In the above description, the part using the word "appropriate" is determined in advance by examining how it appears in a human image. Depending on the installation environment of the camera, a fixed value may be sufficient. Alternatively, those values may be determined as a function of the size of the object in the image (determined by W and H shown in FIG. 8). The functional relationship may be analytically or experimentally determined and stored as a table for W and H for use.

When the width fits the model as shown in FIGS. 8A and 8B and the lower end is obtained (step 511YE).
S), assuming that the face can be detected, the position of the face is output and the process ends (step 512). Otherwise (NO in step 511), it is determined that the face cannot be detected, and step 508 is performed.
Proceed to.

In this embodiment, the position of the face is properly obtained, but as a simplification method, after obtaining the mountain from the projection, the width of the mountain and W and H shown in FIG. May be obtained, the lower end of the face may be determined as a function thereof, and the face portion may be cut out. This functional relationship may be obtained in advance and tabulated as in the case described above. In this case, if the object is detected, the position of the face is obtained, and at this stage, there is no judgment that the face cannot be detected (the object may not be detected). However, if the face portion is not correctly detected, it is determined in the next process that the face is not visible, so there is no problem in the overall operation.
Next, the processing method of the part that determines whether the face is visible will be described. FIG. 9 shows a flowchart of the processing.

First, the position near the center of the face is obtained (step 1101). The vicinity of the center refers to the part including the eyes and mouth. The approximate position of the contour of the face is obtained from the values of the left and right ends of each line of the image when the width is obtained in step 509 of the face detection process and the upper and lower ends of the face determined by the face detection process. . The relative positional relationship of the central part of the face with respect to the contour of the face is experimentally obtained in advance,
The relationship is used to determine the central part of the face. As an example of a specific method, consider a rectangle RT passing through the upper and lower ends, the leftmost end, and the rightmost end of the face as shown in FIG. 10, and experimentally determine the relative position of the rectangle RC in the center of the face with respect to it. , Use it to determine. If the position of the leftmost pixel or the rightmost pixel is really used as the leftmost or rightmost pixel, it is likely to be affected by noise, so it is better to perform an appropriate selection process.
For example, using the nth data, or using the average up to the nth (excluding some from the maximum). For more accurate determination, projection between the upper and lower ends on the X axis may be taken to find the width of the projection peak.

After the position of the central portion of the face is determined, the subsequent processing is performed on the inside of the rectangular area defined as the central portion of the face in the image at one of the two time points. The image at one point may be either one of the two images used in the difference processing, but in general, at a later point in time, the face is slightly larger, so it is better to use this. .

Edge detection is performed on the original image of the central portion (step 1102). For edge detection, an appropriate differential process such as Sobel operator is used, and the result is binarized to take out a part having a large value (step 110).
3). The extracted pixel is treated as a pixel having a value of 1 in the following description. The other part has a value of 0.

Next, as shown in FIG. 11A, the image of the target portion is divided into small areas (an appropriate number of divisions is set in advance), and the number of 1 pixel contained in each small area. Is counted (step 1104). This number of pixels will be referred to as edge density.

Then, it is checked whether or not there is a portion having a large edge density in the lower part of the target area (a predetermined portion such as the portion A in FIG. 11B). If a portion with a high edge density is solidified at the bottom, it is determined that the mouth is visible and the process proceeds to the next step (step 1105YE).
S). If there is no such portion, it is determined that the face is not clearly visible (NO in step 1105), and the process ends (step 1106).

When the position of the mouth is almost found, FIG.
As indicated by B in (b), the edge density is added for one horizontal row of the small region from slightly above, and the sum of one row is obtained (step 1107). In other words, the projection of the edge density on the vertical axis is obtained. As a result, the distribution shown in FIG. 11B is obtained. It is examined from the bottom to find the distribution that creates the first peak (C in the figure) (step 110).
8). If there is no peak, it is determined that the face is not well seen and the process proceeds to step 1106 (step 1108N
O).

If there is a peak (step 1108YE)
S), the density (brightness) of the original image is checked for that portion (step 1109). As shown in FIG. 12, the method of checking is to divide an image vertically and determine the average density of the pixels in each area. Then, the result shown in the lower part of FIG. 12 is obtained.
When the center position of the mouth portion obtained in 105 is called the center of this area, it is checked whether or not the density at the center and the left and right end portions is small and there is a dense portion between them. If this is the case (YES in step 1109), it is determined that the face is finally visible and the process ends (step 11).
10). If no density peaks are detected on both sides, it is determined that the face is not visible (NO in step 1109). As described above, the judgment part can be realized by processing the image.

In the above embodiment, it is necessary to detect 2
I used the difference between the images at the time points.
Similar processing may be performed by obtaining a face portion or contour by binarization or edge detection. This may be used when the background is a simple condition. If there are many objects in the background and they are complicated, the method of the embodiment is suitable. Alternatively, face detection uses the difference as in the embodiment, and when the position of the face is determined, it is limited to the peripheral portion and binarization or edge detection is performed on the image at one time point, and then the face portion is detected. May be determined, and a determination process of whether the face is visible may be performed. In this method, even if the face moves considerably between the two time points, the position of the face is determined again from the image at the one time point, so that the central portion of the face can be detected accurately.

If it is desired to construct a reliable device by using an image at a temporary point, a strobe light is used as the illumination 18, and when the automatic door sensor 4 detects a person, the strobe is emitted to input the image. Good. As the position of lighting,
You may want to make sure your face is well lit near the camera.

Further, in the above-mentioned embodiment, it is explained that only one person can come at a time. When a plurality of people are lined up, the projection may have a plurality of mountains or a wide mountain in the face detection part. In the former case, for each mountain, face detection is performed thereafter, and it is determined whether the faces are visible, and when it is determined that the faces of all people can be detected and the faces are also visible, the door is opened. You can do it like this. In order to deal with the latter case, when a wide mountain is detected, as shown in FIG.
As shown in, a horizontal line is considered in the middle of the image, and the projection to the horizontal line is considered for the portion above it. Find the upper edge of the object, and first examine the horizontal line L1 slightly below it,
Detects mountains as wide as a human head. If this detects multiple peaks and their positions extend over almost the entire width of the initially obtained wide peak, then face detection is possible. If a mountain is detected only in a part corresponding to a part of the first mountain, the horizon is lowered to L2 and checked. Thereafter, similar processing is continued if necessary.

Alternatively, as shown in FIG. 14, the image is examined from above, and when a point 1 is found, the Y coordinate of that point is obtained as shown in the figure. The face portion can be detected by obtaining the portions (P1 and P2 in the figure) that show the maximum values that are considered to be the size of the head, as shown in FIG. 14, from the point sequence of the Y coordinate thus obtained.

In order to avoid such complicated processing,
The environment for entering the store should come in one by one. For example, as shown in FIG. 15, if the poles are set up in front of the door at intervals just enough for one person to pass through, one person will naturally enter the store.

Although the request is made by voice in the embodiment, this may be made by a moving image or other display. Moreover, you may use them together arbitrarily. In addition to the request after it was judged that the face could not be seen well, a mark or the like (in front of the camera was attached to the face or the face part was removed by the camera if voice or a notice was posted in advance). (Easy) You may give a guide to see.

When a request is displayed using a moving image, a flat display such as a liquid crystal or LED may be embedded in the door as shown in FIG. 16 (a). Then, attach a camera around the display. If you do this, the camera will face you naturally. Also in the case of voice guidance, it is advisable to attach a speaker to a door near the camera as shown in FIG. 16A, or to make a stereo sound so that the sound can be heard from the camera.

Further, as shown in FIG. 16 (b), a mascot doll may be placed near the door so that it can give a voice (for example, from the mouth) to guide. In this case, it is advisable to attach a camera to the face, such as the eyes or nose of the doll. Although revealing the camera increases the crime deterrent effect, such a method is suitable when it is not desirable to emphasize the presence of the camera.

In the above description, the apparatus has been described for a person who enters from the outside to the inside. However, a person who goes into the outside from the inside may be operated in the same manner as an ordinary automatic door. However, when entering and leaving the room with the same door, when someone tries to exit and the door opens, people can enter from outside without being checked. The fundamental way to prevent this is to have separate entrance and exit doors. Even in the same case, if it is difficult to enter only one person at a time, as shown in FIG. 15, it can be seen that someone is going out, and the person trying to enter must avoid it, so the problem can be mitigated. .

In the case of complete crime prevention, there is a problem of a person who enters the door immediately after the door is opened by an innocent person. To completely prevent this, a revolving door like that shown in FIG. 17 may be provided in front of the door so that only one person can stand in front of the door. As a mechanism of this part, as shown in FIG. 18, a sensor (an infrared sensor or the like) for detecting the passage of a person is placed in front of and behind the circuit breaker, and when one sensor detects that a person comes in front of the circuit breaker, the circuit is blocked. An automated system may be used so that when the machine is opened and a sensor behind the circuit breaker detects a person, the circuit breaker is closed immediately. These methods can also be used for the above-mentioned purpose of preventing people from getting side by side. Also, if a pole as shown in Fig. 15 is provided,
If the entrance is narrow, it will be difficult to enter immediately, so
Not perfect but effective. In the embodiment, the automatic door is used for the description, but the opening and closing may be performed manually instead of automatically, and only the opening and closing of the lock may be performed based on the determination result.

Further, without using the automatic door sensor, the difference processing is always performed from the television camera image, and when an object is detected, the process proceeds from the subsequent face detection to the determination process of whether the face is visible, and the invention is performed by the image processing. Can also be carried out.

Further, the present invention can be applied to various applications other than the use of opening and closing the door. For example, the image can be recorded only when it is determined that the face can be seen, and it can be used for the purpose of creating a list of images of visitors. It can also be used to detect and record when the face is visible with a security camera in the store.

[0043]

As described above in detail with reference to the embodiments, according to the present invention, it is possible to realize a surveillance camera capable of reliably recording the face of an attendant and having a great crime prevention function and a crime suppressing effect.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of an embodiment of the present invention.

FIG. 2 is a flowchart of the overall processing according to the embodiment of this invention.

FIG. 3 is a diagram showing a mounting example of a camera.

FIG. 4 is a diagram showing an example of a hand switch.

FIG. 5 is a flowchart of face detection processing.

FIG. 6 is a diagram showing an input image and a difference absolute value result.

FIG. 7 is a diagram showing a method of obtaining a projection on the X axis and a width.

FIG. 8 is a diagram showing a model of a width change pattern.

FIG. 9 is a flowchart of a process of determining whether a face is visible.

FIG. 10 is an explanatory diagram of determination near the center of a face.

FIG. 11 is an explanatory diagram of a method of dividing into small areas and checking the positions of the mouth and eyes.

FIG. 12: Description of a method for investigating the presence of eyes.

FIG. 13 is a diagram for explaining a case where a plurality of people come side by side.

FIG. 14 is a diagram for explaining a case where a plurality of people come side by side.

FIG. 15 is a diagram showing a method for entering each person one by one.

FIG. 16 is a diagram showing a request display on a flat display and a method of incorporating a camera into a mascot such as a doll.

FIG. 17 is a diagram showing a method for preventing a person from continuously entering.

FIG. 18 is a diagram showing a method for preventing a person from continuously entering.

[Explanation of symbols]

 1 TV camera 3 Microprocessor circuit 5 Door switch 7 Image recorder

Claims (2)

[Claims] 1. An image inputting means for inputting a person image, a judging means for judging whether or not a face is seen in the person image inputted from the image inputting means, and a face for seeing by the judging means. And a recording means for recording the image when it is determined that the surveillance camera device has the following features. 2. A surveillance camera device, characterized in that when the determination means according to claim 1 determines that the face is not visible, it is determined to be abnormal and the abnormality is notified.