JPS61116489A - Picture monitoring method - Google Patents

Abstract

PURPOSE:To clear an invador by a threat and to issue appropriate alarm depending upon the situation by issuing alarm when any change is found in a monitoring picture as the result of comparing the reference screen with the monitoring picture and by compressing the picture where the change was found and transmitting it in narrow band mode. CONSTITUTION:The cumulatively added value calculated by the projective operation circuit 16 corresponding to the reference screen made of primarily extracted reference data D1i and the current screen is compared with decision reference value, and if the cumulatively added value exceeds the decision reference value, the decision is made that the change between the previous screen and the current screen is not normal, and on the contrary if the cumulative ly added value does not exceed the decision reference value, the decision is made that the change is normal. The picture data within this change region is sent to encoder circuit 21 via data compression circuit 20 to compress and encode the picture data, and the compressed and encoded picture data is then transmitted in narrow band made through telephone line L via modem 23 and NCU24 under the control of data compression circuit 22. When any change occurs in the alarm region set, sound is issued for the threat, thereby being able to clear the invador and being able to detect the movement of critical state quickly since the alarm level is discriminated with the position of the changing object thereafter.

Description

[Detailed Description of the Invention] [Technical Field] The present invention detects changes in surveillance images using a TV camera.
The present invention relates to an image monitoring system that issues an alarm and transmits image information via a telephone line using narrowband transmission.

[Background Art] The present inventors have already proposed a still image transmission method that narrowband transmits the data of the changed part of the image when the image changes. However, since abnormalities are confirmed by humans using the transmitted data, human confirmation is required each time there is a change due to solar radiation or a change due to a cause other than an abnormality. Moreover, there was a problem in that the range that could be controlled by humans was limited.

On the other hand, image feature extraction, line drawing creation, and recognition technologies have been developed for robots and robots that can replace the human eye in fields such as pattern recognition and computer vision, and applying this to security systems is an issue for the future. , it is becoming possible to put it into practical use due to lower costs due to technological advances (1).

Image monitoring devices that are already in practical use for security use detect changing objects based on the difference in brightness and darkness within the monitored area.
There are systems that detect an abnormality when the size of this changing object exceeds a certain value, and systems that take images based on the operation of other sensors, but these are being researched for use in robots (production equipment) as mentioned above. In order to apply advanced technology to image monitoring equipment, it is necessary to have abnormality detection, recognition, and judgment functions that are different from those in production equipment.

[Objective of the Invention 1 The present invention has been made in view of the above-mentioned points, and its purpose is to be able to drive away intruders by intimidation, and also to be able to issue an accurate warning depending on the subsequent situation. Provides an image monitoring system [Disclosure of the Invention] The present invention compares a reference screen and a monitoring image, issues an alarm when there is a change in the monitoring image, and compresses the information of the image with the change to create a narrowband This will be explained below with reference to embodiment figures.

A plurality of monitoring TV cameras 11 to 1n, such as ITVs, are switched by a camera changeover switch 2 and send respective video signals to a synchronization separation circuit 3. The synchronization separation circuit 3 separates the received video signal into a horizontal/vertical synchronization signal and an image signal, converts the image signal from analog to digital using an A/D converter 4, and stores it in a 7-frame memory 5 with, for example, 256 x 256 pixels. It is digitized and written in 8 pixels (256 gradations). The write address is set by the address counter 6 which outputs write address data based on the synchronization signal from the synchronization separation circuit 3.

- The large change detection circuit section 7 is for determining the degree of change and the coordinates of the change area by cumulatively adding the absolute values of the difference signals between frames and projecting them onto the respective coordinate axes X and Y. The extraction circuit 8 extracts image data (16×16 pixels) of lattice points with 166 pixels from the 256×256 pixel image data written in the 7-frame memory 5 as primary sampling data. In this case, the original 16 X i
This corresponds to a case where there is no need to detect changes smaller than 6 pixels. Note that if 32×32 pixels are sampled every 8 pixels as the primary sampled data, it is possible to detect a pattern larger than 8×8 pixels. Now, the monitoring area setting circuit 9 registers primary sampled data corresponding to the background screen of the monitoring area set by the condition setting controller 10, which will be described later, in the storage circuit 11 as primary sampled reference data, and also makes a judgment to detect the presence or absence of a change. This is for creating reference values. The primary sampling reference data, which is the reference screen data, is composed of a plurality of data D11 to Den showing normal changes depending on the time and the illuminance of the monitoring location.

Primary sampling reference data DI registered in the memory circuit 11
l to Dln are the current detected illuminance of the illumination meter 12 and the clock 13
The data is selected in response to an instruction from the primary reference data selection circuit 14 based on the current time of the data, and is read out to the absolute value converting circuit 15. That is, of the primary sampling reference data corresponding to the current time, the primary sampling reference data corresponding to the current illuminance is selected and extracted. The absolute value converting circuit 15 extracts the primary sampling data Da of the current screen extracted from the 7-frame memory 5 by the primary determination data extraction circuit 8, the primary sampling data DI of the previous screen that serves as a reference, and the selected primary sampling data. The difference is converted into an absolute value by removing the sign from each difference with the reference data D11 or by squaring the difference.

Each absolute value data from the absolute value converting circuit 15 is subjected to projection calculation in the projection calculation circuit 16, and the projection calculation circuit 16 cumulatively adds the absolute value of the difference of each pixel for each line parallel to each coordinate axis. As shown in Figure 2, the sum of the absolute value data of the differences on each line parallel to the Y axis (or the sum divided by the number of pixels in one line) is projected onto the X axis. The sum of data on each line parallel to the X-axis is projected onto the axis. - The large change determination circuit 17 compares the cumulative sum calculated by the projection calculation circuit 16 corresponding to the previous screen and the current screen with a reference value, and if the cumulative sum exceeds the standard value, it determines that there is a change. Then, the cumulative sum value calculated by the projection calculation circuit 16 corresponding to the reference screen based on the primary sampling reference data D11 and the current screen is compared with the judgment reference value, and if the cumulative sum value exceeds the judgment reference value, it is determined as described above. The change between the previous screen and the current screen is determined to be abnormal, and conversely, if the accumulated value does not exceed the determination reference value, the change between the previous screen and the current screen is determined to be normal. Next, when it is determined that the change is not normal, the primary change area determination circuit 18 determines the change area. In other words, by detecting the minimum and maximum values of X and Y corresponding to the line exceeding the judgment reference value,
, ≦X≦X2 and Y, ≦Y≦Y2. The image data within this changing area is sent to an encoding circuit 21 via a data compression circuit 20 shown in FIG. The data compression circuit 20 and the encoding circuit 21 compress and encode the image data, and under the control of the transmission controller 22 shown in FIG. 4, the modem 23 and the NCU 2
Narrowband transmission is carried out by the telephone line via 4. The data compression circuit 20 and the encoding circuit 21 compress and encode the image data of the entire screen from the frame memory 5 when no primary change is detected, and compress the image data within the change area when a primary change is detected as described above. For example, as an actual example, a variable sample density prediction code shown in FIG. 5 is constructed by combining a variable sample density compression circuit 25, a variable sample density expansion circuit 26, and a previous value prediction circuit 27. A variable sample density compression circuit 25 and a variable sample density expansion circuit 26 are installed in a feedback loop outside the previous value prediction circuit 27.
is inserted into the feedback loop inside the previous value prediction circuit 27.
A line buffer 28 for each line is provided, and previous value prediction is performed using this line buffer 28 and sample values from the next line.

Now, when a large change is detected, the quadratic change detection circuit section 19 shown in FIG. converted. At this time, the reference data Dc for creating line drawing data is set using a spatial filter operator. In other words, the spatial filter operator has a window with a grid size of 3 x 3 or 5 x 5,
For example, in the case of 3×3, it is expressed as shown in Figure 6, and with this operator, the pixel at coordinates (i, j) is weighted in the window of Figure 6, which includes surrounding pixels as shown in Figure 7. be the value given. Then, the density value of the pixel (i, j) is set to V(
Lj), V(i, j) after the spatial filter operation is the sum of the original pixel and the window multiplied for each pixel.

For example, AXV (i-1, j-1) + BXV (i, j-
1) + E X V (i+j) 10...I X
It can be expressed as V(i+ 1 , j+ 1 ). The values of this A-I window will be various values depending on the creation of the line drawing,
This value will be included in the line drawing creation reference data Dc.

FIG. 8 shows an example of a differential operator, where (,) is a four-way Laplacian, and FIG. 8(b) is an eight-way Laplacian. Also, FIG. 9 shows the Robinson operator, 8
The edge strength and direction can be determined by applying the operators of the templates and using the maximum output value.

By the way, in order to ensure line detection at the stage of line drawing creation, if a binarized image of brightness and darkness is created from a differentiated edge and contour emphasized image, the lines will be discontinuous, so they are shaped by the line drawing creation circuit 30.

Edges and contours are important sources of information for understanding the shape of objects, and simple differences in brightness and darkness are easily affected by external light such as the weather, car headlights, and neon signs. Taking advantage of this feature, the present invention enables more accurate monitoring by specifying important lines in a line drawing and detecting when these lines are hidden by something. We are making it possible to do this. That is, line drawing data of the background of the monitoring area set in advance by the monitoring area setting circuit 32 in the storage circuit 31 is used as secondary reference data in accordance with the time and illuminance. ~D2
In addition to registering n, a judgment reference value for change judgment is registered, and - similarly to large change detection, the difference between the line drawing data D2a of the current screen and the line drawing data D2b of the previous screen is calculated by the difference extraction circuit 34. , the above secondary reference data D21 to D
2n, similarly to the primary change detection, the difference between the secondary reference data D2i selected by the secondary reference data selection circuit 33 based on the illuminance and time data from the illumination meter 12 and the clock 13 and the line drawing data on the current screen is detected. Furthermore, it is determined whether the difference between the line drawing data Da of the current screen and the line drawing data Db of the previous screen is abnormal by comparing the difference between the secondary reference data D21 and the line drawing data of the current screen with the judgment reference value. If the difference value determined by the next change determination circuit 35 exceeds the determination reference value, it is determined to be abnormal, and the line drawing cutout circuit 36 cuts out the portion of the line drawing data that has changed. Extraction and determination of surface differences,
The extraction method is to convert the difference used for primary change detection into an absolute value,
It is better to use the same method as projection calculation and area determination,
The specific operation will be omitted. Now, even if the image data that is compressed and encoded and sent out by the above-mentioned primary change detection is compressed, it cannot be sent as fast as line drawing data, so if a change is detected in the line drawing data while the image data is being sent, it will be cut out. The line drawing data is compressed in the line drawing data compression circuit 37, and then in the encoding circuit 21, the coordinates of the end points and the direction of the connected point sequence are calculated from 1 to 1 per link using a known chain code of Free+nan, which is different from the encoding of image data. 3
The modem 23, N
It is transferred to the telephone line via the CU 24.

In this case, if there is a change even while the image data of the changing area is being transmitted, the state of the change (change in position) can be notified moment by moment using line drawing data, and the receiver side can more accurately determine abnormal conditions based on the way the area moves. be. If alarm processing is performed at this time, it is possible to issue a reliable alarm that eliminates the influence of external light.

For reference, image data of 256 x 256 pixels and 256 gradations is compressed using variable sample density predictive coding 1 using a general subscriber telephone line, with a compression ratio of 1/8 to 1/10 in terms of information amount, and a code amount of 1 bit/1. When sending as pixels, 1200b
It takes about 1 minute at a transmission speed of ps, and about 15 seconds at a transmission speed of 4800 bps.

The image quality obtained at this time is inferior to the 4-picture (DPCM) method, but is at a sufficiently practical level as a surveillance image.

Furthermore, it goes without saying that the transmission time can be reduced from the transmission time of all image data by sending only the images in the change area determined by primary change detection.For example, if the change area is 20%, the transmission time is 115%. becomes.

In addition, the secondary reference data D2. ~D2n and the line drawing data on the current screen are determined based on the level of change, but there are important monitoring areas and line drawings on the screen, less important monitoring areas and line drawings, and ignored values. Judgment criteria are divided into areas such as good monitoring areas and line drawings, and if there is a change in an important monitoring area or line drawing, the alarm level will be raised, and if there is a change in a less important monitoring area or line drawing, an intermediate alert will be issued. level, and if there is a change in the monitoring area or line drawing that can be ignored, this may be ignored and the secondary change detection process may not be performed.

Incidentally, the occurrence of an abnormality can be determined by detecting the primary changes and secondary changes shown in the circuits of FIG. 1 and FIG.
This is the pattern matching circuit section 38 shown in FIG. In other words, the pattern matching/recovery section 38 registers various patterns of abnormal conditions as reference screen data in advance by the three monitoring pattern setting circuits under the control of the condition setting controller 10, and uses the registered patterns and the current screen image data. It is designed to determine an abnormal state by comparing the

Here, the feature extraction circuit 40 is for importing line drawing data and extracting feature points such as end points, bending points, and intersections of the line drawing, as well as closed areas surrounded by lines. Then, the extracted features are labeled and patterned into a hierarchical list. At this time, the hierarchical list is ordered by the inclusion relationship of the features, and the structure analysis circuit 41 performs the relationships so that the detailed features are placed in the lower order. Of course, structural analysis when numerically comparing the above symbol processing involves normalizing and aligning the cut out line drawing data, and then extracting the outline and skeleton line by distance calculation. Good too. Here, abnormal state pattern data D31 to Dsn having characteristic data of changing objects such as people, flames and smoke generated when a fire occurs, and animals are created in advance by the pattern creation circuit 42 and stored by the monitoring pattern setting circuit 39. It is registered in the circuit 43. At this time, each of the pattern data D31 to D3n has judgment criteria such as the shape, size, position, and moving direction of the changing object, so that it is possible to judge whether it is in an abnormal state or a normal state based on the judgment criteria. It is.

In addition, these elements such as shape, size, position, and direction of movement are divided into categories so that it is possible to identify which category a change in an object in a surveillance image falls into. When invading into an area, the patterns are divided according to the transition of the abnormal state at the time of intrusion by categorizing them according to where they are invading (the type of person entering is treated as the same category). Then, when a secondary change is detected and the extracted line drawing data is taken into the pattern matching circuit section 38, pattern data is created after structural analysis. The model fixing circuit 44 compares the difference between the pattern data D, b of the previous screen and the pattern data D31 of the current screen, and also verifies the changing object in the comparison between the registered pattern data D3i and the pattern data D3a of the current screen, Furthermore, the movement determination circuit 45 determines the size, position, and moving direction of the changing object for each registered pattern data D3i to detect movement of the changing object.

The threat processing circuit unit 46 is a circuit that determines the alarm level depending on the area where the movement of the object or abnormal change has occurred, and issues an alarm according to the level. If this occurs in a predetermined vigilance area, the vigilance level determination circuit 47 compares the change area with vigilance level determination data LD, which is divided into levels for each vigilance area and stored in advance, and determines that the change area is at the predetermined vigilance level. If the object falls under the warning area, a sound threat means 50 such as a siren, a bell, or a speaker is activated to intimidate the object, and in response to the threat, the changing object disappears or there is no change in the position of the changing area. The system selects a warning level such as lowest, intermediate, or highest depending on the situation, such as moving to a warning area with a higher warning level, and creates warning data corresponding to the warning level. Of course, the initial threat can also be expected to drive away intruders. Alarm data creation circuit 4
8 creates alarm data according to the judgment information of the alert level judgment circuit 47, and sends the alarm data to the modem 23 and NCU 24.
This is to transmit the image data and line drawing data to the telephone line via the telephone line.

In addition, the four alarm information display sections display the current alert level using a light emitting display means or the like according to the above alarm data, and
A message display setting circuit 52 for displaying the content of alarm data in characters on a monitor CRT 56 provided on a control panel 58 is set in accordance with the alarm level.

Now, the control panel 58 contains line drawing data, image data,
and video RAM for text data for message display.
51 is provided, and the data on the video RAM 51 is D/
The data can be D/A converted by the A converter 55 and displayed on the monitor CRT 561, and can also be used to set a monitoring area, to set a warning area corresponding to the warning level, to weight the importance of line drawings, or to detect abnormal conditions. Light pen 5 for pattern settings etc.
7 (or a mouse), instructions can be given on the screen displayed on the monitor CRT 56, and these data are sent to the monitoring area setting circuit 9, 32 and the monitoring pattern under the control of the condition setting controller 10. Setting circuit 39
etc., and are registered. The line drawing display circuit 53 is a circuit for causing the monitor CRT 56 to display a line drawing when setting conditions regarding the line drawing such as weighting the importance of the line drawing described above.

Now, the image data, line drawing data, and alarm data sent via the telephone line are sent to the receiving controller 6 via the NCU 62 and modem 63 of a predetermined receiver as shown in FIG.
In the receiver, these data are decoded by a decoding circuit 61, and then subjected to signal processing to reproduce images and line drawings on a monitor television 59, and also to emit an alarm sound from a speaker 60. .

To further explain the configuration and operation of the receiver in detail, the alarm notification circuit 67 in FIG. The alarm character display circuit 67 is for creating character data for displaying an alarm message based on the alarm data, and is for displaying the message on the monitor television 59 by being imposed on the image and line drawing signals described later. It is. The change area setting circuit 66 is the decoding circuit 61
The compressed data decompression circuit 68 is for detecting a change area (X , , Y , ) (X 2 . Y 2) from the image data and line drawing data extracted through the image data and line drawing data.
It is for decompressing and decoding line drawing data, and the decompressed and decoded data is written into the frame memory 69. The contents of the frame memory 69 are read by the address counter 70 and input to the D/A converter 73 to become an image signal, and this image signal is synthesized into a synchronous signal together with the synchronous signal from the synchronous signal generation circuit 71 and the above character data. The signals are combined in the circuit 72 and displayed on the monitor television 59.

FIG. 11 shows an example of the compressed data decompression circuit 68. This example is a variable sample density predictive decoding circuit configured by combining a variable density decoding circuit and a previous value predictive decoding circuit. A line buffer 776 is provided in the loop of the prediction circuit 75 for restoring the sample from the data expanded by the expansion circuit 74 according to the method.

By the way, in the above embodiment, the image data of the changing area or the cut out line drawing data is transmitted, but since the line drawing data also requires a certain amount of information, it is necessary to display it on the receiver side quickly. There are limits to what you can do. Therefore, the registered pattern data D31 to D are stored in the memory circuit 43.
3n as well as a symbol corresponding to the registered pattern data D31-D3n, and if the type, position, and size of the change pattern are similar to the previously registered pattern and belong to the same category, the change pattern is registered. Before transmitting the image data and line drawing data of the area, the symbol data corresponding to the registered pattern data D is transferred to the movement determination circuit 45.
The data is transmitted over the telephone line via the modem 23 and NCU 24, and the receiver side is provided with a memory circuit 78 as shown inside the broken line frame in FIG. is registered pattern data D 31~
Model pattern data is registered in advance with the same content as Den, and when the above-mentioned symbol data is transmitted, the corresponding model pattern data is read out and written to the frame memory 69 via the model pattern data display setting circuit 77. By displaying the contents of 69 on the monitor television 59, the general information can be quickly conveyed to the receiver side by transmitting a small amount of information in the form of symbol data.

In addition, as mentioned above - as a method for selecting the primary sampling reference data when detecting a large change, for example, if the change in the image every minute is less than a certain set value, the primary sampling reference data serving as the background is updated, and the primary sampling reference data that is the background is updated and A slow change in the value may not be detected as abnormal. In this case, the primary sampling standard data, which is the standard screen data to be updated, may be updated by taking a moving average so that the previous data is also affected at a certain rate, and It is also possible to perform averaging such as passing through a low-pass filter to remove minute changes.

Furthermore, it is of course possible to use methods other than the above-mentioned methods for creating contours, line drawings, and structural analysis.

[Effects of the Invention] The present invention is an image monitoring system that issues an alarm when a change occurs in a monitoring image, and when a change occurs in a preset warning area on the monitoring image, a threatening sound such as a bell is generated. The intruder can be chased away by threat, and the threat can encourage the intruder to move, making it possible to falsify the confirmation of the intruder.Moreover, it is important because the level of the alarm is determined based on the position of the object that changes after that. This has the effect of quickly detecting the transition of a situation.

[Brief explanation of the drawing]

FIG. 1 is a circuit configuration diagram corresponding to a major change detection according to an embodiment of the present invention, FIG. 2 is an explanatory diagram of the operation of the same change detection, and
The figure is a circuit configuration diagram corresponding to the quadratic change detection same as above, FIG. 4 is a circuit configuration diagram corresponding to the same pattern matching, FIG. 6 to 9 are explanatory diagrams of the line drawing creation method, FIG. 10 is a circuit configuration diagram of the receiver side of the same, FIG. 11 is a specific circuit configuration diagram of the compressed data decompression circuit of the same, and 7 A primary change detection circuit section, 47 is a warning level determination circuit, 48 is a warning data creation circuit, 50 is a threatening means, and LD is warning level determination data. Agent Patent Attorney Ishi 1) Ai Figure 7, Figure 9, MOM + Figure 10, Telephone Circulation L Written Amendment to Procedures (Voluntary) February 12, 1985 W Ia 8. (to) 1. Indication of the case: 1982 Patent Application No. 2372 + 2. 2. Name of the invention Image monitoring method 3. Relationship with the case of the person making the amendment Patent applicant Location 1048 Bold Kadoma, Kadoma City, Osaka Name (583) Matsushita Electric Works Co., Ltd. Representative Iku Kobayashi 4, Agent 5 , the date of the amendment order, issue 8, the content of the amendment as per the attached sheet, the application number for correction document No. 1, Japanese Patent Application No. 59-237212, and the word "details" in line 8 on page 5 of the specification of the present application is corrected to "change". 2. Same as above #19, line 4, "1impose" is corrected to "superimpose." 3. Correct Figure 5 of the drawings as shown in the attached sheet.

Claims (2)

[Claims] (1) In an image surveillance system that issues an alarm when a change occurs in the surveillance image, when a change occurs in a preset warning area on the surveillance image, a threatening sound such as a bell is generated, and the following changes An image monitoring method characterized by determining the alarm level based on the location of the image. (2) The lowest alarm level is when the changing object disappears after the threat, the intermediate warning level is when the changing object's position does not change, and the highest alert is when the changing object moves to a warning area with a higher alert level. 2. The image monitoring system according to claim 1, wherein the image monitoring system has a level of .