JPH10105839A - Monitoring image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce erroneous warning, that is, erroneously giving warning, and to normally warn with respect to a normal warning object, by continuing recognition processing for a fixed time so as to observe the varying area of the warning object, confirming that the area is not rapidly increased or reduced, and warning based on the result. SOLUTION: A recognition processing part 9A executes a varying area integration processing integrating varying areas supposed to belong to the same moving object as the result of binarization data to collect to by one varying area. Next, a feature value comparing part processing which judges the integrated varying area is a warning object or not is executed. Then, a varying area with a feature coinciding with a prescribed condition is regarded as the warning object. However, the judging result is dealt with as 'temporary' result and not reported yet. The varying area of the warning object is additionally tracked and observed for a fixed time to see the rapid change of its area. By receiving the result, whether the temporary finding of the warning object is right or not is judged and only when it is right, the report is transferred to a warning processing part 10.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a monitoring image processing apparatus installed in a center for remotely monitoring, for example, an unmanned electric station.

[0002]

2. Description of the Related Art The configuration of a conventional monitoring image processing apparatus will be described with reference to FIG. FIG. 24 is a block diagram showing a schematic configuration of a conventional monitoring image processing apparatus.

In FIG. 24, reference numeral 1 denotes a video input unit for receiving a video to be processed by the monitoring image processing apparatus, 2 denotes an A / D converter for quantizing a video input from the video input unit 1, and 3 denotes an A / D. A current image storage unit that receives an output from the conversion unit 2 and stores the current video data, and a background 4 that stores an output from the A / D conversion unit 2 and performs a corresponding operation to store a background image for comparison. The image storage units 5 and 5 are background update units that calculate and update the data of the background image storage unit 4 according to a certain rule. Generally, update with simple rules. For example, the average of the past five current images is used as the background image.

[0006] In FIG. 1, reference numeral 6 denotes a current image storage unit 3.
The background difference unit 7 compares the two images of the background image storage unit 4 and calculates the difference between them. Reference numeral 7 denotes a threshold value calculation unit that calculates a threshold value for binarizing the data of the background difference unit 6. In general, an optimum threshold value is obtained based on background difference data. Further, reference numeral 8 denotes a binarization unit for binarizing the difference data with the threshold value calculated by the threshold value calculation unit 7, and 9 receives the binary data output from the binarization unit 8, and outputs a target alert. Recognition processing units 10 and 10 that determine whether an object is present are alarm processing units that receive an instruction of “discovery target discovery” from the recognition processing unit 9 and issue an alarm.

FIG. 25 is a flowchart showing in detail the operation of the recognition processing section of the conventional monitoring image processing apparatus.

In step 11, the recognition processing section 9
Looking at the binarized data, a change area integration process is performed in which the change areas that are considered to belong to the same moving object are integrated into one change area.

[0007] Steps 12 to 15 are feature amount comparison portions for judging whether or not the change area integrated in step 11 is an alert target. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target. Here, the following determination is made as an example.

That is, in step 12, it is checked whether or not the change area exists continuously for a predetermined time. In step 13,
It is checked whether or not the change area has a predetermined area. Step 1
At 4, it is checked whether or not the change area has predetermined vertical and horizontal dimensions. In step 15, it is checked whether or not the change area has a predetermined moving speed.

[0009] In response to the above result, in step 16, it is determined whether a report target is found or there is no target. A change area having a feature meeting the above conditions is regarded as a report subject, and a report to the report processing unit 10 is made. Pass.

Next, a typical example of the operation will be described. In FIG. 24, a video signal externally given to the device is received by a video input unit 1 and sent to an A / D conversion unit 2. The A / D converter 2 converts the received video signal into digital data for subsequent processing. It is common to convert to 8-bit multi-value data per pixel. The current image storage unit 3 stores the input current image data and prepares for subsequent processing. In the background image storage unit 4,
The background image created based on the input current image data is accumulated and prepared for later processing. The background update unit 5 periodically updates the background image stored in the background image storage unit 4 according to a certain rule. This is generally calculated based on the current image that is periodically sampled.
In many cases, the average value of the current image of the scene is used as a background image, or the image is a video of a frame that is a predetermined time earlier.

The background difference unit 6 compares the outputs of the current image storage unit 3 and the background image storage unit 4, finds a difference for each pixel, and outputs the difference data as a frame image.
The difference data generally represents a portion of the video where the luminance has changed. Usually, a moving object such as a person or a vehicle is mainly used. However, in some rare cases, a change in luminance that is not a target of alarm, such as the appearance of a shadow of a building due to “shadow of sunlight”, may be exhibited.

The threshold value calculating section 7 calculates a threshold value for binarizing the difference data sent from the background difference section 6. Normally, background data is multi-valued data, which is inappropriate for recognition processing as it is. In general, binarization and conversion into binary data are also performed in order to make subsequent determinations at high speed.
However, there is a case where the difference data is directly recognized and processed without performing the binarization depending on the device, so this is a reference. The binarizing unit 8 converts the difference data into two using the threshold value.
Value. As a result, in the subsequent processing, the change area becomes binary data.

The change region will be described with reference to FIG. 26 as an example. FIG. 26 is an image of a normal main gate. Now, a person is moving in the image. In this case, the position of the moving person becomes the change area as it is. FIG. 27 is an image in which the superimposed areas in which the change areas are dyed black are overlapped. The recognition processing unit 9 receives the binarized change region data from the binarization unit 8, and the change region is a target (for example, generally corresponds to a person or a vehicle. Is determined.

The recognition processing section 9 performs recognition in the following procedure. However, this is only an example. Step 11
In the change area integration process, the change areas are considered, and the change areas that are considered to belong to the same moving object are integrated into one change area. In steps 12 to 15, step 11
It is determined whether or not the change area integrated in the above is an alarm target.
Here, it is compared whether or not the change region matches a predetermined feature amount. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following judgment is made as an example. In step 12, it is checked whether or not the change area exists continuously for a fixed time. In step 13, it is checked whether or not the change area has a predetermined area. In step 14, it is checked whether or not the change area has a predetermined vertical and horizontal dimension. In step 15, it is checked whether or not the change area has a predetermined moving speed.

The "area", "length and width" and "moving speed" are generally referred to by calculating an average during a certain time in step 12. However, conversely, even if a completely different object that is not a target to be emitted appears here (for example, the appearance of a shadow of a building due to the shadow of sunlight), if the feature amount condition is satisfied here, the device will generate an alert. Resulting in. Such an erroneous notification by a non-reporting target is hereinafter referred to as unnecessary detection. Regardless of whether it is unnecessary detection or normal detection, if it is determined in step 16 that the notification target is found, it is determined that the notification processing unit 1
0, where the alarm processing is performed. Usually, it is a process of sending a specific signal to a higher-level PC or the like or sounding a buzzer of the device. If the conditions are not met,
It is assumed that there was no alarm target. In this case, nothing is reported to the next process. Note that the operations of the recognition processing unit 9 and the alert processing unit 10 are generally processed by S / W.

[0017]

In the conventional monitoring image processing apparatus as described above, since the structure is as described above, the change area extracted by the background difference unit 6 is temporarily set as a non-reporting target (for example, , The appearance of building shadows due to the shade of sunlight,
Etc.), a warning is issued if the conditions in steps 11 to 15 are met. In particular, when performing monitoring work outdoors, in many cases, in addition to the appearance of shadows of buildings due to the shading of sunlight, it is often caused by fluctuations, glitters, and water marks on the water surface. In other words, there is a problem in that when the above non-reporting target meets the conditions, a report is issued.

The present invention has been made in order to solve the above-described problems. For example, such as "appearing gradually without moving, After the discovery of the alarm target, the alarm processing unit does not immediately execute the alarm processing for the `` relatively large change area that gradually disappears without moving, '' By observing the change area of the alert target, confirming that there is no sudden increase or decrease in area, and issuing an alert based on the result, an alert is erroneously issued for the above change area It is an object of the present invention to provide a monitoring image processing apparatus capable of reducing the number of "false alarms" and simultaneously issuing normal alarms to a normal alarm target such as "person".

[0019]

According to the present invention, there is provided a monitoring image processing apparatus comprising: a changing area extracting means for extracting a changing area based on a comparison operation between current image data and comparison background image data; And a recognition processing means for recognizing the area as an alert target when the area satisfies a predetermined condition and when the area of the change area does not change for a predetermined time for a predetermined time.

In the monitoring image processing apparatus according to the present invention, the change area extracting means calculates a background difference as a change area based on a comparison operation between the current image data and the comparison background image data. A threshold value calculating unit that calculates a threshold value for binarizing the background difference data based on the background difference data; and a binarizing unit that binarizes the background difference data with the threshold value. Things.

Further, the surveillance image processing apparatus according to the present invention further comprises: A / D conversion means for quantizing the input image;
Current image storage means for storing the current image data from the output of the A / D conversion means, background image storage means for creating and storing the background image data for comparison from the output of the A / D conversion means, A background updating means for calculating and updating the data of the background image accumulating means according to a predetermined rule, and an alarm processing means for notifying when the extracted change area is recognized as an alarm target, is notified.

Further, in the monitoring image processing apparatus according to the present invention, the change area extracting means may be configured to perform a change between the frames based on a comparison operation between the current image data and the comparison background image data which is the current image data one frame before. First background subtraction means for calculating a background difference as a change area; second background difference means for calculating a background difference as a change area based on a comparison operation between current image data and comparison background image data; Threshold calculating means for calculating first and second thresholds for binarizing the inter-frame background difference data and the background difference data based on the inter-frame background difference data and the background difference data; And binarizing means for binarizing the background difference data with the first and second threshold values and outputting first and second binary data, wherein the recognition processing means comprises: A first recognition processing means for recognizing a temporary alert target when the extracted change area matches a predetermined condition based on the first binary data, and the extraction processing based on the second binary data; And a second recognition processing means for recognizing as a notification target when the changed area meets a predetermined condition and is recognized as the temporary notification target by the first recognition processing means. It is.

Further, the surveillance image processing apparatus according to the present invention further comprises an A / D conversion means for quantizing the input video,
Current image storage means for storing the current image data from the output of the A / D conversion means, and comparison background image data which is current image data one frame before the output of the A / D conversion means; First background image storage means for storing;
A second background image storage unit that creates and stores the background image data for comparison from the output of the A / D conversion unit; and a second unit that calculates and updates the data of the first background image storage unit according to a predetermined rule. (1) a background updating means, (2) a second background updating means for calculating and updating data of the second background image accumulating means according to a predetermined rule, and, when the extracted change area is recognized as an alarm target, the fact is notified. Notification processing means for notifying.

Further, in the monitoring image processing apparatus according to the present invention, the change area extracting means may be configured to perform an inter-frame change based on a comparison operation between current image data and comparison background image data which is current image data one frame before. First background subtraction means for calculating a background difference as a change area; second background difference means for calculating a background difference as a change area based on a comparison operation between current image data and comparison background image data; Threshold calculating means for calculating first and second thresholds for binarizing the inter-frame background difference data and the background difference data based on the inter-frame background difference data and the background difference data; And binarizing means for binarizing the background difference data with the first and second threshold values and outputting first and second binary data, wherein the recognition processing means comprises: If the extracted change area matches a predetermined condition based on the second binary data, the change area is recognized as a tentative alert target, and the area of the extracted change area is determined based on the first binary data. If there is no predetermined change for a certain period of time, the temporary notification target is recognized as a notification target.

Further, the surveillance image processing apparatus according to the present invention further comprises A / D conversion means for quantizing the input video,
Current image storage means for storing the current image data from the output of the A / D conversion means, and comparison background image data which is current image data one frame before the output of the A / D conversion means; First background image storage means for storing;
A second background image storage unit that creates and stores the background image data for comparison from the output of the A / D conversion unit; and a second unit that calculates and updates the data of the first background image storage unit according to a predetermined rule. (1) a background updating means, (2) a second background updating means for calculating and updating data of the second background image accumulating means according to a predetermined rule, and, when the extracted change area is recognized as an alarm target, the fact is notified. Notification processing means for notifying.

Further, the surveillance image processing apparatus according to the present invention further comprises A / D conversion means for quantizing the input video,
Current image storage means for storing the current image data from the output of the A / D conversion means, background image storage means for creating and storing the background image data for comparison from the output of the A / D conversion means, A background updating means for calculating and updating data of the background image accumulating means according to a predetermined rule; and an alarm processing means for notifying when the extracted change area is recognized as an alarm target, the recognition processing means comprising: When the extracted change area meets a predetermined condition, the change area is recognized as a tentative alarm target, and the background updating means stores the background image data for comparison as the current image data of one frame before in the background image storage means. And when the area of the extracted change region has not been changed for a predetermined period of time based on the inter-frame background difference data, the provisional notification target is recognized as a notification target. A.

In the monitoring image processing apparatus according to the present invention, there is provided a changing area extracting means for extracting a changing area based on a comparison operation between the current image data and the background image data for comparison; When the condition is met, and when the threshold value for binarizing the change area has not been changed for a predetermined period of time, there is provided recognition processing means for recognizing the threshold as an alarm target.

In the monitoring image processing apparatus according to the present invention, the change area extracting means calculates a background difference as a change area based on a comparison operation between the current image data and the background image data for comparison. A threshold value calculating unit that calculates a threshold value for binarizing the background difference data based on the background difference data; and a binarizing unit that binarizes the background difference data with the threshold value. Things.

Further, the surveillance image processing apparatus according to the present invention further comprises A / D conversion means for quantizing the input image,
Current image storage means for storing the current image data from the output of the A / D conversion means, background image storage means for creating and storing the background image data for comparison from the output of the A / D conversion means, A background updating means for calculating and updating data of the background image accumulating means according to a predetermined rule; and an alarm processing means for notifying when the extracted change area is recognized as an alarm target, the recognition processing means comprising: When the extracted change area meets a predetermined condition, the change area is recognized as a tentative alarm target, and the background updating means stores the background image data for comparison as the current image data of one frame before in the background image storage means. Is stored, and when the threshold value for binarizing the extracted change area based on the inter-frame background difference data has not been changed for a predetermined period of time, the provisional notification target is set to the notification target. It is intended to recognized.

[0030]

BEST MODE FOR CARRYING OUT THE INVENTION

Embodiment 1 FIG. The configuration of the first embodiment of the present invention will be described with reference to FIG. FIG. 1 is a block diagram showing a schematic configuration of Embodiment 1 of the present invention. In the drawings, the same reference numerals indicate the same or corresponding parts.

In FIG. 1, reference numeral 1 denotes a video input unit for receiving a video to be processed by the monitoring image processing apparatus, 2 denotes an A / D converter for quantizing a video input from the video input unit 1, and 3 denotes an A / D converter.
The current image storage unit 4 receives the output from the / D conversion unit 2 and stores the current video data. The current image storage unit 4 receives the output from the A / D conversion unit 2 and performs a corresponding operation to store the background image for comparison. The background image storage units 5 and 5 are background update units that calculate and update data of the background image storage unit 4 according to a certain rule. Generally, update with simple rules. For example, the average of the past five current images is used as the background image.

In the same figure, reference numeral 6 denotes the current image storage unit 3.
The background difference unit 7 compares the two images of the background image storage unit 4 and calculates the difference between them. Reference numeral 7 denotes a threshold value calculation unit that calculates a threshold value for binarizing the data of the background difference unit 6. In general, an optimum threshold value is obtained based on background difference data. Reference numeral 8 denotes a binarizing unit for binarizing the difference data with the threshold value calculated by the threshold value calculating unit 7. 9A receives the binary data output from the binarization unit 8 and determines whether or not the target alert target exists in the binary data. The reference numeral 9A denotes the "discovery target discovery" from the recognition processing unit 9. Is a notification processing unit that receives and issues an instruction.

FIG. 2 is a flowchart showing the recognition processing section 9A of FIG. 1 according to the first embodiment in more detail. In step 21, a change area integration process is performed in which the change data that is considered to belong to the same moving object is integrated into one change area by looking at the binarized data. Next, steps 22 to
At 25, a feature amount comparing unit process is performed to determine whether or not the change area integrated at step 21 is an alarm target. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target. Here, for example, the following determinations in steps 22 to 25 are made.

In step 22, it is checked whether or not the change area exists continuously for a fixed time. In step 23, it is checked whether or not the change area has a predetermined area. In step 24, it is checked whether or not the change area has a predetermined vertical and horizontal dimension. In step 25, it is checked whether or not the change area has a predetermined moving speed.

In response to the above result, in step 26, it is determined whether the alarm target is found or not. However, here, the judgment result is treated as “temporary” and is not yet reported.
In step 217, the alert target change area is tracked and observed for a further fixed time. Therefore, we see a sudden change in the area. In response to the result, in step 28, it is determined whether or not the tentative notification target discovery is correct, and the report is passed to the notification processing unit 10 of FIG. 1 only when it is correct.

FIG. 3 is an example of an image in which a "person" and a "water mark" of a puddle are shown in the image. FIG. 4 is a state immediately after FIG. 3 and is an image in which water marks are reduced in a puddle.

Next, a typical example of the operation will be described. The components other than the recognition processing unit 9A are the same as those in the conventional example, and will not be described. FIG. 2 is a content diagram of the “recognition process” of the monitoring image processing apparatus according to the first embodiment.

In the recognition process, recognition is performed according to the following procedure. However, this is only an example. In the change area integration process in step 21, the change areas are considered, and the change areas that are considered to belong to the same moving object are integrated and combined into one change area. Next, in steps 22 to 25, it is determined whether or not the change area integrated in step 21 is an alarm target. Here, it is compared whether or not the change region matches a predetermined feature amount. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following judgment is made as an example. In step 22, it is checked whether or not the change area exists continuously for a fixed time. In step 23, it is checked whether or not the change area has a predetermined area. In step 24, it is checked whether or not the change area has a predetermined vertical and horizontal dimension. In step 25, it is checked whether or not the change area has a predetermined moving speed. The “area”, “length and width”, and “moving speed” are generally referred to by calculating an average during a predetermined time in step 22.

Next, even if it is determined in step 26 that the notification target is found, it is treated as "temporary" and the notification processing unit 1
It cannot be conveyed to 0. In step 27, the recognition process is further continued. In this case, the tracking is further continued on the changed area found as the tentative alarm target in step 26. Then, it is continuously observed for a certain period of time (for example, 0.5 seconds) whether or not the area is rapidly reduced or increased. If no abnormality is found within a predetermined time, step 2
At 8, it is determined that the report target is formal, and the report is sent to the report processing unit 10. The alert processing unit 10 performs an alert process. Usually, it is a process of sending a specific signal to a higher-level PC or the like or sounding a buzzer of the device. If a sudden change in the area is found in step 27, it is determined that the "temporary notification target discovery" was incorrect, and no "formal notification target discovery" report is made. The recognition processing unit 9A and the alarm processing unit 10 are generally processed by S / W.

The operation of the system according to the first embodiment will be described below. Now, let us consider a case where, in the image of FIG. 3, a “person” that is a target for regular notification and a “water mark” that is a target for non-reporting have occurred. In addition, it is assumed that both of them satisfy various conditions in the recognition processing of FIG. Also,
The “constant time” referred to in step 27 is 0.5 seconds,
"Abrupt change in area" is defined as less than half or more than double. Further, when the device finds a change area, it displays a circumscribed rectangle and informs the user of the fact.

FIG. 3 shows “puddle” and “person” in the video. The device finds the change area and displays a circumscribed rectangle. The subject of the alarm is “person”. Therefore, issuing a notification by a person is "normal notification", and issuing by a person other than a person is "unnecessary detection". Here, as determined by the above assumption, both the “person” and the “water pattern” are determined to be alarmed because the area, the vertical and horizontal dimensions, and the speed are within the predetermined ranges. Since this device is still "Temporary alarm target discovery",
There is no report. If the alarm is issued, the portion due to the water mark becomes “unnecessary detection”.

After that, the apparatus changes the “person” change area,
Track and observe the changing areas of the "water pattern". FIG.
Is an image 0.5 seconds after FIG. The apparatus was tracking and observing the “temporary alarm target”, “person” and “water mark”. Here, the area is checked again. The “person” has an area approximately equal to that of the “temporary alert target discovery”. Therefore, here, it is determined that there is no abnormality, and it is determined that “formal notification of the target of discovery” has been made.

However, the “water mark” has an area of only about 1/3 as compared with the area at the time of “provisional notification target discovery”. This is because the water marks change shape and disappear rapidly. Therefore, it is determined that there is an abnormality here,
It is not determined to be "formal alert target discovery".

As described above, the "person" is regarded as a legitimate report target discovery, and the "water mark" is not. This allows
Unnecessary detection due to “water pattern” does not occur. As described above, according to this processing, it is possible to prevent the unnecessary detection of the “water mark” without hindering the alert of the “person”.

As described above, according to the first embodiment,
It does not impair the ability to fire legitimate objects such as "people", such as the appearance of building shadows due to the shadows of sunlight and the appearance of building shadows due to puddle water marks. It is possible to realize a function of reducing unnecessary detection for “a relatively large change area that gradually appears without moving or gradually disappears without moving”.

In other words, the alert is not immediately issued when the alert target is found, but the recognition process is continued thereafter, and the change area is observed for a certain period of time to confirm that there is no sudden increase or decrease in the area. By issuing a notification after waiting for the result, it is possible to reduce "false alarms" that are erroneously issued to sudden changes in brightness such as "appearance of shadows due to sunshine" and "person" Has a function of issuing a normal notification for a regular notification target as described above.

Embodiment 2 The configuration of the second embodiment of the present invention will be described with reference to FIG. FIG. 5 is a block diagram showing a schematic configuration of the second embodiment of the present invention.

In FIG. 5, reference numeral 1 denotes a video input unit for receiving a video to be processed by the monitoring image processing apparatus, 2 denotes an A / D converter for quantizing a video input from the video input unit 1, and 3 denotes an A / D converter.
The current image storage section 4a receiving the output from the / D conversion section 2 and storing the current video data stores the background image for comparison, which is created by receiving the output from the A / D conversion section 2 and performing a corresponding operation. A background image storage unit, in particular, records the current image one frame before. 4b is a background image storage unit for storing a background image for comparison, which is created by receiving the output from the A / D conversion unit 2 and performing a corresponding operation, and in particular, recording an image different from the background image storage unit 4a. do.

In FIG. 5, reference numeral 5a denotes a background update unit for updating data of the background image storage unit 4a, and 5b denotes a background update unit for calculating and updating data of the background image storage unit 4b according to a certain rule. Generally, update with simple rules. For example, the average of the past five current images is used as the background image.

In the figure, reference numeral 6a denotes a background difference unit for comparing two images of the current image storage unit 3 and the background image storage unit 4a and calculating a difference between them.
Since the current image before the frame is recorded, the output is inter-frame difference data. Reference numeral 6b denotes a background difference unit that compares two images of the current image storage unit 3 and the background image storage unit 4b and calculates a difference between the two images.

Further, in FIG.
a threshold value calculating unit for calculating a threshold value for binarizing the data (difference data a and b) of the background difference unit 6b, and generally calculates an optimum threshold value based on the background difference data. is there. Numeral 8 denotes a binarizing unit for binarizing the difference data with the threshold value calculated by the threshold value calculating unit 7. Since the difference data a and b are binarized, the binarized data to be output is also binarized. Type (binary data a and b). 9a is a recognition processing unit that receives the binary data a output from the binarization unit 8 and determines whether or not the target alert target is present. 9b is a binary processing unit that outputs the binary data output from the binarization unit 8. A recognition processing unit that receives the data b and determines whether or not the target alert target exists;
At the same time, the recognition result of the recognition processing unit 9a is also integrated and determined.
Reference numeral 10 denotes a notification processing unit that receives an instruction of “discovery target discovery” from the recognition processing unit 9b and issues a notification.

FIG. 6 is a flowchart showing the recognition processing section 9b of FIG. 5 according to the second embodiment in more detail. In addition, regarding the recognition processing unit 9a, Step 3 of the recognition processing unit 9b is performed.
Except for the absence of 7, 38, the same is omitted. The input data is usually binary data b.

In step 31, the binarized data is viewed,
Change areas that are considered to belong to the same moving object
A change area integration process for combining the two change areas is performed.
Next, in steps 32 to 35, a feature amount comparison process is performed to determine whether or not the change area integrated in step 31 is an alert target. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following steps 32 to 35 are determined as examples. In step 32, it is checked whether the change area exists continuously for a certain period of time. In step 33, it is checked whether or not the change area has a predetermined area. In step 34, it is checked whether or not the change area has a predetermined vertical and horizontal dimension.
In step 35, it is checked whether or not the change area has a predetermined moving speed. In response to the above result, in step 36, it is determined whether the alarm target is found or not. However, here, the judgment result is treated as “temporary” and is not yet reported.

In step 37, the alert target change area is further tracked and observed. Here, hereinafter, the recognition processing unit 9a
Refer to the recognition result of. The recognition processing unit 9b also determines whether or not the same change area as the target of the “provisional notification target discovery” is also “provisional notification target discovery”. In response to the result, in step 38, it is determined whether or not the tentative notification target discovery is correct, and the report is passed to the notification processing unit 10 only when it is correct.

FIG. 7 is an example of an image in which a "person" and a "water mark" of a puddle are shown in the image. FIG. 8 is a state immediately after the above-described figure, and is an image in which the water mark of the puddle is slightly reduced. FIG. 9 shows a result of binarizing the above-described diagram by obtaining difference data by the method of the background difference unit 6b. FIG.
Is a result of binarizing the above-described figure by obtaining difference data by the method of the background difference unit 6a.

Next, a typical example of the operation will be described. In FIG. 5, a video signal externally supplied to the device is received by a video input unit 1 and sent to an A / D conversion unit 2. The A / D converter 2 converts the received video signal into digital data for subsequent processing. It is common to convert to 8-bit multi-value data per pixel. The current image storage unit 3 stores the input current image data and prepares for subsequent processing.

The background image storage unit 4a stores the input current image data of the previous frame and prepares for the subsequent processing.
On the other hand, the background image storage unit 4b stores a background image created based on the input current image data and prepares for the subsequent processing.

The background update unit 5a periodically updates the background image stored in the background image storage unit 4a according to a certain rule in preparation for a later inter-frame difference. This generally replaces the current image sampled every frame.
On the other hand, the background update unit 5b periodically updates the background image stored in the background image storage unit 4b according to a certain rule. This is generally performed on the basis of the current image that is periodically sampled. For example, a method of using the average value of the current image of the past 100 scenes as the background image, a method of setting the average value as a background image for a predetermined time (for example, 5 seconds) ) In many cases.
Here, it is assumed that update is performed to replace the image with the image of the frame 5 seconds before.

The background difference section 6b compares the outputs of the current image storage section 3 and the background image storage section 4b, finds a difference for each pixel, and outputs the difference data as a frame image. The difference data generally represents a portion of the video where the luminance has changed. Usually, a moving object such as a person or a vehicle is mainly used. However, in some rare cases, a change in luminance that is not a target of alarm, such as the appearance of a shadow of a building due to “shadow of sunlight”, may be exhibited.

On the other hand, the background subtraction section 6 a
Then, the output of the background image storage unit 4a is compared, a difference is obtained for each pixel, and the difference data as the difference is output as a frame image. Here, since the difference between the current image and the current image one frame before is calculated, the difference is necessarily an inter-frame difference. The difference data based on the inter-frame difference generally indicates a portion where the luminance changes between one frame. Usually, a moving object such as a person or a vehicle moving at high speed is mainly used. Due to its nature, a clear difference does not appear with a time difference of about one frame, so that a person moving at low speed cannot extract difference data well. However, relatively large changes that gradually appear without moving or gradually disappear without moving, such as "appearance of shadows of buildings due to the shadows of sunlight" and "water marks of puddles" Similarly, the difference data cannot be extracted from a change in luminance such as “region”.

The threshold value calculation section 7 calculates a threshold value for binarizing the difference data (difference data a, b) sent from the background difference section 6a and the background difference section 6b. Normally, background data is multi-valued data, which is inappropriate for recognition processing as it is. In general, binarization and conversion into binary data are also performed in order to make subsequent determinations at high speed. However,
This is for reference because some devices directly recognize the difference data without performing the binarization.

The binarizing section 8 binarizes the difference data a and b with the threshold value. As a result, in the subsequent processing, the change area becomes binary data (binary data a and b).

The change area will be described with reference to FIG. 26 as an example. FIG. 26 is an image of a normal main gate. Now, a person is moving in the image. In this case, the position of the moving person becomes the change area as it is. FIG. 27 is an image in which the superimposed areas in which the change areas are dyed black are overlapped.

The recognition processing unit 9a receives the binarized data a from the binarizing unit 8, and the change area indicated by the binarized data a (for example, generally corresponds to a person or a vehicle. Is determined. Also,
The recognition processing unit 9b receives the binarized data b from the binarizing unit 8, and the change area indicated by the binarized data b is a target (for example, generally corresponds to a person or a vehicle. Hereinafter, it is referred to as an alarm target). .) Is determined.

FIG. 6 is a diagram showing the contents of the recognition processing section 9b of the monitoring image processing apparatus according to the second embodiment. The contents of the recognition processing unit 9a do not include steps 37 and 38 of the contents of the recognition processing unit 9b.

The recognition processing section 9b performs recognition in the following procedure. However, this is only an example. Normally, the recognition processing unit 9b targets binary data b. In the change area integration process in step 31, the change areas are observed, and the change areas that are considered to belong to the same moving object are integrated and combined into one change area. Next, in steps 32 to 35, it is determined whether or not the change area integrated in step 31 is an alarm target. Here, it is compared whether or not the change region matches a predetermined feature amount. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following judgment is made as an example. In step 32, it is checked whether the change area exists continuously for a certain period of time. In step 33, it is checked whether or not the change area has a predetermined area. In step 34, it is checked whether or not the change area has a predetermined vertical and horizontal dimension. In step 35, it is checked whether or not the change area has a predetermined moving speed. The “area”, “vertical and horizontal dimensions”, and “moving speed” are generally referred to by calculating an average during a predetermined time in step 32.

Even if it is determined in step 36 that the notification target is found, it is treated as "temporary" and is not yet transmitted to the notification processing unit 10. In step 37, the alert target change area is tracked and observed for a further fixed time. In step 36, the tracking is further continued on the changed area found as the tentative alarm target. Here, the recognition result of the recognition processing unit 9a is referred to hereafter. The recognition processing unit 9a also determines whether the same change area that is the target of the “temporary notification target discovery” is set as the “temporary notification target discovery”.

The recognition processing unit 9a refers to the inter-frame difference data as described above, and determines the appearance of the shadow of the building due to the "appearance of the shadow of the building due to the shade of sunlight" or the "appearance of the shadow of the puddle". A difference in luminance such as “a relatively large change area that gradually appears without movement or gradually disappears without movement” cannot similarly extract difference data.

Then, it is determined whether the recognition processing unit 9b has found the same change area as the "temporary report target discovery" as "provisional report target discovery".
Due to the feature of the previous inter-frame difference, if "provisional alert target discovery" is based on "appearance of building shadows due to the shadows of sunlight" or "water puddles in puddles", a very large area cannot be extracted here. . Therefore, "provisional alarm target discovery"
Does not. From the result, the device can determine the abnormality.

Upon receiving the result, in step 38, it is determined whether or not the tentative notification target discovery is correct, and the report is passed to the notification processing unit 10 only when it is correct. If no abnormality is found, it is determined in step 38 that a formal report target has been found, and the report is sent to the report processing unit 10. The alarm processing section 10 performs an alarm processing. Usually, it is a process of sending a specific signal to a higher-level PC or the like or sounding a buzzer of the device. If an abnormality is found in step 37, it is determined that the "provisional alert target discovery" was incorrect,
No "formal alert target found" report is issued. The recognition processing units 9a and 9b and the alert processing unit 10 are generally processed by S / W.

Hereinafter, how the method according to the second embodiment operates will be described. Now, let us consider a case where, in the image of FIG. 7, a “person”, which is a target for normal notification, and a “water mark”, which is a target for non-reporting, respectively occur. In addition, it is assumed that both of them satisfy various conditions in the recognition processing of FIG. 6 and are individually notified. Also,
The “constant time” referred to in step 37 is 0.5 seconds,
"Abrupt change in area" is defined as less than half or more than double. Further, when the device finds a change area, it displays a circumscribed rectangle and informs the user of the fact.

FIG. 7 shows “puddle” and “person” in the video.
Is shown. The device finds the change area and displays a circumscribed rectangle. The subject of the alarm is "person"
It is. Therefore, issuing a notification by a person is "normal notification", and issuing by a person other than a person is "unnecessary detection". here,
As determined by the above assumption, both the “person” and the “water mark” are determined to be alarmed because the area, vertical and horizontal dimensions, and speed are within the predetermined ranges. In the present apparatus, the report is still "provisional report target discovery", and therefore no report is issued. If the alarm is issued, the portion due to the water mark becomes “unnecessary detection”.

Thereafter, the apparatus sets a change area of “person”,
Track and observe the changing areas of the "water pattern". FIG.
Is an image 0.5 seconds after FIG. The ripples are shrinking, but not yet small enough.

FIG. 9 shows the result of background subtraction by the background subtraction unit 6b from the current image of FIG. on the other hand,
FIG. 10 shows the result of background subtraction performed by the background subtraction unit 6a on the current image of FIG.

In FIG. 9, the watermark is a background image b (here, 5
In order to calculate the difference between the current image and the image of the frame a second before, the watermark and the person are cut out as they are. On the other hand, in FIG. 10, the watermark is the difference between the background image a (here, the image of the immediately preceding frame) and the current image. Only the part that has shrunk to "and the part of the person that has moved in the time of one frame are cut out.

The apparatus performs the recognition processing of the two kinds of background update methods in parallel. Here, two results will be seen again.
As a result of the difference between frames in FIG. 10, “person” has an area slightly smaller than that in FIG. However, since it does not reach half, it is determined here as "provisional alert target discovery". However, as a result of the difference between frames, the “water pattern” has an area of only about １ ／ as compared with FIG. Such a result is obtained since the water mark disappears only to that extent during one frame. Therefore, here, "provisional alert target discovery"
Is not determined. As described above, the “person” is determined to be a tentative alarm target discovery by both the recognition processing units 9a and 9b, and the “water mark” is not. Instead, the changing area of "water crest"
It is not considered a formal alert. As described above, according to this processing, it is possible to prevent the unnecessary detection of the “water mark” without hindering the alert of the “person”.

As described above, according to the second embodiment,
It does not impair the ability to fire legitimate objects such as "people", such as the appearance of building shadows due to the shadows of sunlight and the appearance of building shadows due to puddle water marks. It is possible to realize a function of accurately reducing unnecessary detection regarding “a relatively large change area that gradually appears without moving or gradually disappears without moving” more than the effect described in the first embodiment.

That is, a plurality of background subtraction means (of which one
One is the difference between frames), and the background difference is always multiple (of which 1
First, a frame-to-frame difference) is performed, and a notification is not issued immediately when the alarming target is found. Then, recognition processing based on the frame-to-frame difference is repeated, and it is confirmed that there is no large area difference between the two results. By issuing a report after waiting for the result, for example, a "false report" may be issued by mistake for a sudden change in luminance such as "appearance of a shadow due to sunshine".
Is reduced more reliably, and at the same time, it is possible to issue a normal notification for a regular notification target such as "person".

Embodiment 3 The configuration of the third embodiment of the present invention will be described with reference to FIG. FIG.
FIG. 7 is a block diagram showing a schematic configuration of Embodiment 3 of the present invention.

In FIG. 11, reference numeral 1 denotes a video input unit for receiving a video to be processed by the monitoring image processing apparatus, 2 denotes an A / D converter for quantizing a video input from the video input unit 1, and 3 denotes an A / D. A current image storage unit that receives an output from the conversion unit 2 and stores current video data.

In the figure, reference numeral 4a denotes a background image storage unit for storing a background image for comparison, which is created by receiving the output from the A / D conversion unit 2 and performing a corresponding operation. Records the previous current image. 4b also A / D
Created by receiving the output from the conversion unit 2 and performing a corresponding operation.
A background image storage unit that stores a background image for comparison. Here, an image different from that of the background image storage unit 4a is particularly recorded.

In the figure, reference numeral 5a denotes a background update unit for updating data of the background image storage unit 4a, and 5b denotes a background update unit for calculating and updating data of the background image storage unit 4b according to a certain rule. Update with simple rules. For example, the average of the past five current images is used as the background image.

In the figure, reference numeral 6a denotes a background difference unit for comparing two images of the current image storage unit 3 and the background image storage unit 4a and calculating a difference between the two images. In order to record the previous current image, this output is inter-frame difference data, which may be referred to as difference data a. On the other hand, reference numeral 6b denotes a background difference unit that compares the two images of the current image storage unit 3 and the background image storage unit 4b and calculates the difference between them. In particular, the background difference unit 6b is created here to distinguish it from the background difference unit 6a. The obtained difference data is temporarily referred to as non-interframe difference data. Note that it may be referred to as difference data b.

Further, in the figure, reference numeral 7 denotes a background subtraction unit 6
a and data of the background difference unit 6b (difference data a, b)
This is a threshold value calculation unit that calculates a threshold value for binarizing, and generally calculates an optimum threshold value based on background difference data. Numeral 8 denotes a binarizing unit for binarizing the difference data with the threshold value calculated by the threshold value calculating unit 7. Since the difference data a and b are binarized, the binarized data to be output is also binarized. Kind. (Hereinafter, referred to as binarized data a and b)
9B is the binary data a output from the binarization unit 8,
b, and is a recognition processing unit that determines whether or not the target alert target exists. Usually, the recognition processing unit performs the determination process using the binary data b. Details will be described later. 10 is a recognition processing unit 9
This is a notification processing unit that receives and issues an instruction of “discovery target discovery” from B.

FIG. 12 is a flowchart showing the recognition processing section 9B of FIG. 11 according to the third embodiment in more detail. The input data is usually binary data b. And
In particular, when there is an instruction, it is binary data a. This instruction will be described later.

In step 41, the binarized data is viewed,
Change areas that are considered to belong to the same moving object
This is a change area integration process for combining the change areas into two change areas. Next, Steps 42 to 45 are a feature amount comparison part that determines whether or not the change area integrated in Step 41 is an alarm target. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following steps 42 to 45 are determined as examples. In step 42, it is checked whether or not the change area exists continuously for a fixed time. In step 43, it is checked whether or not the change area has a predetermined area. In step 44, it is checked whether or not the change area has predetermined vertical and horizontal dimensions.
In step 45, it is checked whether or not the change area has a predetermined moving speed. In response to the above result, in step 46, it is determined whether the alarm target is found or not. However, here, the judgment result is treated as “temporary” and is not yet reported.

In step 47, the report target change area, which is the target of the "temporary report target discovery", is tracked and observed for a certain period of time. First, in step 41,
Is switched from the binary data b to the binary data a. Then, in step 47, the area is checked. The area determined based on the binary data a is compared with the area determined based on the binary data b, and a rapid change in the area is observed. Receiving the result, in step 48, it is determined whether or not the tentative notification target discovery is correct, and the report is passed to the notification processing unit 10 only when it is correct.

FIG. 13 is an example of an image in which a “person” and a “water mark” of a puddle are shown in the image. FIG. 14 is a state immediately after the above figure, and is an image in which the water mark of the puddle is slightly reduced. FIG. 15 shows the result of binarizing the above-described figure by obtaining difference data by the method of the background difference unit 6b.
FIG. 16 shows the result of binarizing and obtaining difference data from the above-described diagram by the method of the background difference unit 6a.

Next, a typical example of the operation will be described. In FIG. 11, a video signal externally supplied to the device is received by a video input unit 1 and sent to an A / D conversion unit 2. The A / D converter 2 converts the received video signal into digital data for subsequent processing. It is common to convert to 8-bit multi-value data per pixel.
The current image storage unit 3 stores the input current image data and prepares for subsequent processing.

The background image storage section 4a stores the input current image data of the previous frame and prepares for the subsequent processing.
On the other hand, the background image storage unit 4b stores a background image created based on the input current image data and prepares for the subsequent processing.

The background update unit 5a periodically updates the background image stored in the background image storage unit 4a according to a certain rule in preparation for a later inter-frame difference. This generally replaces the current image sampled every frame.
On the other hand, the background update unit 5b periodically updates the background image stored in the background image storage unit 4b according to a certain rule. This is generally performed on the basis of the current image that is periodically sampled. For example, a method of using the average value of the current image of the past 100 scenes as the background image, a method of setting the average value as a background image for a predetermined time (for example, 5 seconds) ) In many cases.
Here, it is assumed that the update is performed to replace the image with the image of the frame 5 seconds before.

The background difference section 6b compares the outputs of the current image storage section 3 and the background image storage section 4b, finds a difference for each pixel, and outputs the difference data as a frame image. The difference data generally represents a portion of the video where the luminance has changed. Usually, a moving object such as a person or a vehicle is mainly used. However, in some rare cases, a change in luminance that is not a target of alarm, such as the appearance of a shadow of a building due to “shadow of sunlight”, may be exhibited.

On the other hand, the background difference section 6a
Then, the output of the background image storage unit 4a is compared, a difference is obtained for each pixel, and the difference data as the difference is output as a frame image. Here, since the difference between the current image and the current image one frame before is calculated, the difference is necessarily an inter-frame difference. The difference data based on the inter-frame difference generally indicates a portion where the luminance changes between one frame. Usually, a moving object such as a person or a vehicle moving at high speed is mainly used. Due to its nature, a clear difference does not appear with a time difference of about one frame, so that a person moving at low speed cannot extract difference data well. However, relatively large changes that gradually appear without moving or gradually disappear without moving, such as "appearance of shadows of buildings due to the shadows of sunlight" and "water marks of puddles" Similarly, the difference data cannot be extracted from a change in luminance such as “region”.

The threshold value calculating section 7 calculates a threshold value for binarizing the difference data (difference data a, b) sent from the background difference section 6a and the background difference section 6b. Normally, background data is multi-valued data, which is inappropriate for recognition processing as it is. In general, binarization and conversion into binary data are also performed in order to make subsequent determinations at high speed. However,
This is for reference because some devices directly recognize the difference data without performing the binarization.

The binarizing section 8 binarizes the difference data a and b using the threshold value. As a result, in the subsequent processing, the change area becomes binary data (binary data a and b).

The change region will be described with reference to FIG. 26 as an example. FIG. 26 is an image of a normal main gate. Now, a person is moving in the image. In this case, the position of the moving person becomes the change area as it is. FIG. 27 is an image in which the superimposed areas in which the change areas are dyed black are overlapped.

The recognition processing unit 9B receives the binarized data a and b from the binarization unit 8, and the change area indicated by the data is a target (for example, generally corresponds to a person or a vehicle. Is determined.

FIG. 12 is a content diagram of the recognition processing section 9B of the monitoring image processing apparatus according to the third embodiment. The recognition processing section 9B performs recognition in the following procedure. However, this is only an example. Usually, in this recognition processing, the binary data b is targeted. The binary data a may be referred to as necessary, but at a specific time as described later.

In the changing area integration process in step 41, the changing areas are considered, and the changing areas that are considered to belong to the same moving object are integrated into one changing area. Next, in steps 42 to 45, the change area integrated in step 41 is
It is determined whether or not the notification is to be issued. Here, it is compared whether or not the change region matches a predetermined feature amount. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following judgment is made as an example. In step 42, it is checked whether or not the change area exists continuously for a fixed time. In step 43, it is checked whether or not the change area has a predetermined area. In step 44, it is checked whether or not the change area has predetermined vertical and horizontal dimensions. In step 45, it is checked whether or not the change area has a predetermined moving speed. The “area”, “length and width”, and “moving speed” are generally referred to by calculating an average during a certain time in step 42. Even if it is determined in step 46 that the notification target is found, it is "temporary"
And is not yet transmitted to the alert processing unit 10.

In step 47, the alert target change area is tracked and observed for a further fixed time. In step 46, the tracking is further continued on the changed area found as the provisional notification target. Prior to the tracking, first, the binary data input to the change area integration processing in the previous step 41 is converted into binary data b.
To binary data a.

As described above, this is inter-frame difference data, and does not involve movement, such as the appearance of a building shadow due to the shading of sunlight or the appearance of a building shadow due to a puddle in a puddle. In the same manner, difference data such as "a relatively large change area which gradually appears or gradually disappears without moving" cannot be used to extract difference data. Then, the changed area that has just been “provisionally made a report target discovery” is tracked again by the changed area based on the binary data a, and the area obtained based on the binary data b is compared with the binary data a Compared to the area that was calculated based on, we see a sudden change in the area.

According to the feature of the difference between the frames described above, if the “provisional alert target discovery” is based on “appearance of a shadow of a building due to shading of sunlight” or “watermark of a puddle”, a very large area is used here. Cannot be extracted. Therefore, the apparatus can determine the abnormality from the sudden change in the area.

In response to the result, in step 48, it is determined whether or not the tentative notification target discovery is correct, and the report is passed to the notification processing unit 10 only when it is correct. If no abnormality is found within a predetermined period of time, it is determined in step 48 that a formal report target has been found, and the report is sent to the report processing unit 10. The alarm processing section 10 performs an alarm processing. Usually, it is a process of sending a specific signal to a higher-level PC or the like or sounding a buzzer of the device. If a sudden change in the area is found in step 47, it is determined that the "provisional notification target discovery" was incorrect, and no "formal notification target discovery" report is made. The recognition processing unit 9B and the alert processing unit 10 are generally processed by S / W.

Hereinafter, how the method according to the third embodiment operates will be described. Now, let us consider a case where a “person” that is a target for normal notification and a “water mark” that is a target for non-reporting have occurred in the image of FIG. In addition, it is assumed that both of them satisfy various conditions in the recognition processing of FIG. 12 and are individually issued. Also, the “constant time” in step 47 is defined as 0.5 seconds, and the “abrupt change in area” is defined as half or less or twice or more. Further, when the device finds a change area, it displays a circumscribed rectangle and informs the user of the fact.

FIG. 13 shows “puddle” and “person” in the video. The device finds the change area and displays a circumscribed rectangle. The subject of the alarm is “person”. Therefore, issuing a notification by a person is "normal notification", and issuing by a person other than a person is "unnecessary detection". Here, as determined by the above assumption, both the “person” and the “water pattern” are determined to be alarmed because the area, the vertical and horizontal dimensions, and the speed are within the predetermined ranges. With this device, "Provisional alert target discovery"
Therefore, no alarm is issued. If the alarm is issued, the portion due to the water mark becomes “unnecessary detection”.

[0111] After that, the apparatus sets a change area of "person",
Track and observe the changing areas of the "water pattern". FIG.
Is an image 0.5 seconds after FIG. The ripples are shrinking, but not yet small enough.

FIG. 15 shows the result of the background subtraction performed by the background subtraction unit 6b on the current image of FIG. On the other hand, FIG. 16 shows a background difference unit 6a for the current image of FIG.
Shows the result of the background subtraction.

In FIG. 15, since the watermark is the difference between the background image b (here, the image of the frame 5 seconds before) and the current image, the watermark and the person are cut out as they are. on the other hand,
In FIG. 16, since the watermark is a difference between the background image a (here, the image of the immediately preceding frame) and the current image, the difference is inevitably an inter-frame difference, and the watermark is shrunk during one frame. Only the part that has been moved and the part of the person that has moved in the time of one frame are cut out.

The apparatus sets the background updating method to the background updating unit 5b.
Is changed to the inter-frame difference,
I was pursuing the "person" and "water crest", but here I look at the area again. As a result of the inter-frame difference, "person"
It has an area that is slightly smaller than the “provisional alert target discovery”. However, since it does not reach half, it is determined that there is no abnormality here, and it is determined that "formal notification target discovery". However, as a result of the difference between frames, “water mark” has an area of only about １ ／ as compared with the area at the time of “provisional notification target discovery”. Such a result is obtained since the water mark disappears only to that extent during one frame. Therefore, here, it is determined that there is an abnormality, and it is not determined that “formal alert target discovery”.

As described above, the "person" is regarded as a legitimate report target discovery, and the "watermark" is not. This allows
Unnecessary detection due to “water pattern” does not occur. As described above, according to this processing, it is possible to prevent the unnecessary detection of the “water mark” without hindering the alert of the “person”.

As described above, according to the third embodiment,
It does not impair the ability to fire legitimate objects such as "people", such as the appearance of building shadows due to the shadows of sunlight and the appearance of building shadows due to puddle water marks. It is possible to realize a function of accurately reducing unnecessary detection regarding “a relatively large change area that gradually appears without moving or gradually disappears without moving” more than the effect described in the first embodiment. In addition, the configuration of the device for that purpose is simpler than in the second embodiment.

That is, a plurality of background subtraction means (one of them)
One is the difference between frames), the alert is not immediately issued when the alert target is found, the recognition process is continued by changing the background difference means to the inter-frame difference and the change area is observed for a certain period of time.
By confirming that there is no sudden increase or decrease in area, and issuing an alert after the result, without performing recognition processing twice,
The second embodiment has the function of obtaining the same effect as the second embodiment without taking unnecessary time.

Embodiment 4 The configuration of Embodiment 4 of the present invention will be described with reference to FIG. FIG.
FIG. 14 is a block diagram showing a schematic configuration of a fourth embodiment of the present invention.

In FIG. 17, reference numeral 1 denotes a video input unit for receiving a video to be processed by the monitoring image processing apparatus, 2 denotes an A / D converter for quantizing a video input from the video input unit 2, and 3 denotes an A / D. The current image storage unit 4 receives the output from the conversion unit 2 and stores the current video data. The background 4 stores the background image for comparison, which is created by receiving the output from the A / D conversion unit 2 and performing a corresponding operation. The image storage unit 5 is a background update unit that calculates and updates the data of the background image storage unit 4 according to a certain rule, and generally updates the data according to a simple rule. For example, the average of the past five current images is used as the background image. Further, the background update unit 5A can switch the background update method according to designation of a recognition processing unit described later. Reference numeral 6 denotes a background difference unit that compares the two images of the current image storage unit 3 and the background image storage unit 4 and calculates the difference between them. 7 calculates a threshold value for binarizing the data of the background difference unit 6. It is a threshold value calculation unit, and generally calculates an optimum threshold value based on background difference data. 8
Is a binarization unit for binarizing the difference data with the threshold value calculated by the threshold value calculation unit 7, 9C receives the binary data output from the binarization unit 8, and the target alert target exists there. Recognition processing units 10 and 10 that make a determination as to whether or not to perform the notification are a notification processing unit that receives and issues an instruction of “discovery target discovery” from the recognition processing unit 9C.

FIG. 18 is a flowchart showing the recognition processing section 9C of FIG. 17 according to the fourth embodiment in more detail.

Step 51 is a change area integration process in which the change areas considered to belong to the same moving object are integrated into one change area by looking at the binarized data. next,
Steps 52 to 55 are feature amount comparison portions for determining whether or not the change area integrated in step 51 is an alarm target. Here, the change area having the feature that matches the predetermined condition is
It is considered as a target for alarm.

Here, the following steps 52 to 55 are determined as examples. In step 52, it is checked whether or not the change area exists continuously for a fixed time. In step 53, it is checked whether or not the change area has a predetermined area. In step 54, it is checked whether or not the change area has a predetermined vertical and horizontal dimension.
In step 55, it is checked whether or not the change area has a predetermined moving speed. In response to the above result, in step 56, it is determined whether the alarm target is found or not. However, here, the judgment result is treated as “temporary” and is not yet reported.

In step 57, the alert target change area is tracked and observed for a further fixed time. In this case, first, the background update unit 5A is instructed to make the image to be taken into the background image storage unit 4 the current image. Then, in step 57, the area is checked. Compare the area determined after the switching with the area determined before the switching, and observe a rapid change in the area. In response to the result, in step 58, it is determined whether or not the tentative notification target discovery is correct, and the report is passed to the notification processing unit 10 only when it is correct.

FIG. 19 is an example of an image in which a “person” and a “water mark” of a puddle are shown in the image. FIG. 20 is a state immediately after the above figure, and is an image in which the water mark of the puddle is slightly reduced. FIG. 21 shows a result obtained by switching the background updating unit 5A to obtain the difference data and binarizing the above image so that the image captured in the background image storage unit 4 becomes the current image.

Next, a typical example of the operation will be described. In FIG. 17, a video signal externally supplied to the device is received by a video input unit 1 and sent to an A / D conversion unit 2. The A / D converter 2 converts the received video signal into digital data for subsequent processing. It is common to convert to 8-bit multi-value data per pixel.
The current image storage unit 3 stores the input current image data and prepares for subsequent processing. The background image storage unit 4 stores a background image created based on the input current image data, and prepares for the subsequent processing.

The background updating section 5A periodically updates the background image stored in the background image storing section 4 according to a certain rule. This is generally performed on the basis of the current image that is periodically sampled. For example, a method of using the average value of the current image of the past 100 scenes as the background image, a method of setting the average value as a background image for a predetermined time (for example, 5 seconds) ) In many cases. Here, it is assumed that update is performed to replace the average value of the current image of the past 100 scenes. However, here, there is a function of switching the background update method to another method according to an instruction of the recognition processing unit 9C. The background updating method that can be switched at that time is, for example, that the current image is directly taken into the background image storage unit 4.

The background difference unit 6 compares the outputs of the current image storage unit 3 and the background image storage unit 4, finds a difference for each pixel, and outputs the difference data as a frame image.
The difference data generally represents a portion of the video where the luminance has changed. Usually, a moving object such as a person or a vehicle is mainly used. However, in some rare cases, a change in luminance that is not a target of alarm, such as the appearance of a shadow of a building due to “shadow of sunlight”, may be exhibited.

However, when the background updating unit 5A changes the background updating method to the method of “fetching the current image as it is into the background image storage unit 4”, and when the original image is restored, the current image is instantaneously added to the current image.
The difference between the current image before the frame is calculated, and the inter-frame difference is inevitably obtained only at this moment. After that, even if the background updating method is restored by the background updating unit 5A, it takes another 100 scenes to return to the background image similar to the original background image. Past 10
The background difference result during the update is temporarily close to the inter-frame difference (because the update that replaces the average value of the current image of the 0 scene is performed). (They are gradually turning to normal background subtraction results.)

The difference data based on the inter-frame difference generally indicates a portion where the luminance changes between one frame. Usually, a moving object such as a person or a vehicle moving at high speed is mainly used.
Due to its nature, a clear difference does not appear with a time difference of about one frame, so that a person moving at low speed cannot extract difference data well. However, relatively large changes that gradually appear without moving or gradually disappear without moving, such as "appearance of shadows of buildings due to the shadows of sunlight" and "water marks of puddles" Similarly, the difference data cannot be extracted from a change in luminance such as “region”.

The threshold value calculating section 7 calculates a threshold value for binarizing the difference data sent from the background difference section 6. Normally, background data is multi-valued data, which is inappropriate for recognition processing as it is. In general, binarization and conversion into binary data are also performed in order to make subsequent determinations at high speed.
However, there is a case where the difference data is directly recognized and processed without performing the binarization depending on the device, so this is a reference.

The binarizing section 8 binarizes the difference data using the threshold value. As a result, in the subsequent processing, the change area becomes binary data.

The change area will be described with reference to FIG. 26 as an example. FIG. 26 is an image of a normal main gate. Now, a person is moving in the image. In this case, the position of the moving person becomes the change area as it is. FIG. 27 is an image in which the superimposed areas in which the change areas are dyed black are overlapped.

The recognition processing section 9C receives the binarized data from the binarizing section 8, and the change area indicated by the binarized data is a target (for example, generally corresponds to a person or a vehicle. Is determined.

FIG. 18 is a content diagram of the recognition processing section 9C of the monitoring image processing apparatus according to the fourth embodiment. The recognition processing unit 9C performs recognition in the following procedure. However, this is only an example.

In the changing area integration process in step 51, the changing areas which are considered to belong to the same moving object are integrated by integrating the changing areas into one changing area. Next, in steps 52 to 55), it is determined whether or not the change area integrated in step 51 is an alarm target. Here, it is compared whether or not the change region matches a predetermined feature amount. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following judgment is made as an example. In step 52, it is checked whether or not the change area exists continuously for a fixed time. In step 53, it is checked whether or not the change area has a predetermined area. In step 54, it is checked whether or not the change area has a predetermined vertical and horizontal dimension. In step 55, it is checked whether or not the change area has a predetermined moving speed. The “area”, “length and width”, and “moving speed” are generally referred to by calculating an average during a predetermined time in step 52. Even if it is determined in step 56 that the notification target is found, it is "temporary"
And is not yet transmitted to the alert processing unit 10.

In step 57, the alert target change area is tracked and observed for a further fixed time. Prior to the tracking, first, the background updating unit 5A is instructed to make the image captured in the background image storage unit 4 the current image. As it is, the current image is compared by the background difference unit 6, and no difference is generated. Immediately, it instructs the preceding background updating unit 5A to return the image captured in the background image storage unit 4 to the original method.

Thereafter, the background image of the background image storage unit 4 returns to the original data after a lapse of 100 scenes (the current background update method performs update by replacing the average value of the current image of the past 100 scenes). However, during that time, the background update is incomplete.

Immediately after the switching, the difference between the current image and the current image one frame before is instantaneously obtained, and inevitably the inter-frame difference only at this instant. However, thereafter, the background image gradually changes to a normal background image, and thus gradually returns to normal background difference data.

As described above, in the inter-frame difference data,
"Slowly appearing without moving, or gradually disappearing without moving, such as the appearance of building shadows due to the shadows of sunlight" or "watermarks of puddles" Similarly, the difference data cannot be extracted from a change in luminance such as “a relatively large change area”.

Then, the changed area that has just been “provisional alarm target discovery” is tracked again by the changed area due to the inter-frame difference, and the resulting areas are compared to see a rapid change in the area. According to the feature of the difference between the previous frames, "Temporary alarm target discovery" is replaced by "Appearance of building shadows due to the shade of sunlight."
In the case of "water puddle of a puddle", a very large area cannot be extracted here. Therefore, the apparatus can determine the abnormality from the sudden change in the area.

In response to the result, in step 58, it is determined whether or not the tentative notification target discovery is correct, and the report is passed to the notification processing unit 10 only when it is correct. If no abnormality is found within a predetermined period of time, it is determined in step 58 that a formal report target has been found, and the report is sent to the report processing unit 10. The alarm processing section 10 performs an alarm processing. Usually, it is a process of sending a specific signal to a higher-level PC or the like or sounding a buzzer of the device. If an abrupt change in area is found in step 57, it is determined that the "temporary alert target discovery" was incorrect, and no "formal alert target discovery" report is made. Note that the recognition processing unit 9C and the alert processing unit 10 are generally processed by S / W.

The following describes how the method according to the fourth embodiment works. Now, let us consider a case where a “person” that is a regular notification target and a “water mark” that is a non-reporting target have occurred in the image of FIG. It is also assumed that both of them satisfy various conditions in the recognition processing of FIG. Further, the “constant time” in step 57 is defined as 0.5 seconds, and the “abrupt change in area” is defined as half or less or twice or more. Further, when the device finds a change area, it displays a circumscribed rectangle and informs the user of the fact.

FIG. 19 shows “puddle” and “person” in the video. The device finds the change area and displays a circumscribed rectangle. The subject of the alarm is “person”. Therefore, issuing a notification by a person is "normal notification", and issuing by a person other than a person is "unnecessary detection". Here, as determined by the above assumption, both the “person” and the “water pattern” are determined to be alarmed because the area, the vertical and horizontal dimensions, and the speed are within the predetermined ranges. With this device, "Provisional alert target discovery"
Therefore, no alarm is issued. If the alarm is issued, the portion due to the water mark becomes “unnecessary detection”.

After that, the apparatus changes the “person” change area,
Track and observe the changing areas of the "water pattern". FIG.
Is an image 0.5 seconds after FIG. The ripples are shrinking, but not yet small enough.

FIG. 21 shows the result of the background subtraction of the current image of FIG. 20 immediately after the switching of the background updating method. There, the watermark is a background image (here, 1
The difference between the current image and the current image is necessarily obtained, so the difference between the frames is inevitably an inter-frame difference, and the “water-shrinking part of one frame” and the person “moving to the time of one frame” Only the "done part" is cut out.

The apparatus changes the background updating method from the conventional background updating method to the difference between frames, and tracks and observes "person" and "water mark" which are "temporary alarm targets". Look at the area again. "People" as a result of the difference between frames
Has an area slightly smaller than that at the time of “provisional notification target discovery”. However, since it does not reach half, it is determined that there is no abnormality here, and it is determined that "formal notification target discovery".

However, as a result of the inter-frame difference, the “water mark” has an area of only about １ ／ as compared with the area at the time of “provisional notification target discovery”. Such a result is obtained since the water mark disappears only to that extent during one frame. Therefore, here, it is determined that there is an abnormality, and it is not determined that “formal alert target discovery”.

As described above, the “person” is regarded as a legitimate report target discovery, and the “water mark” is not. This allows
Unnecessary detection due to “water pattern” does not occur. As described above, according to this processing, it is possible to prevent the unnecessary detection of the “water mark” without hindering the alert of the “person”.

As described above, according to the fourth embodiment,
It does not impair the ability to fire legitimate objects such as "people", such as the appearance of building shadows due to the shadows of sunlight and the appearance of building shadows due to puddle water marks. It is possible to realize a function of accurately reducing unnecessary detection regarding “a relatively large change area that gradually appears without moving or gradually disappears without moving” more than the effect described in the first embodiment. In addition, the configuration of the device for that purpose is simpler than in the second and third embodiments.

In other words, instead of "notifying immediately when an alerting target is found, and then changing the background subtraction means to an inter-frame difference and continuing recognition processing to observe a change area for a certain period of time,""Do not immediately issue an alert when the alert is found, then change the background update method to" Replace and update with the current image "only once, and then return the background update means to continue recognition processing for a certain period of time. By observing the change area, it is confirmed that there is no sudden increase or decrease in the area, and by issuing a report after waiting for the result, the apparatus scale is not increased because there are no multiple background difference means. It has a function capable of obtaining the same effect as the third embodiment.

Embodiment 5 FIG. The configuration of Embodiment 5 of the present invention will be described with reference to FIG. FIG.
FIG. 14 is a block diagram showing a schematic configuration of Embodiment 5 of the present invention.

In FIG. 22, reference numeral 1 denotes a video input unit for receiving a video to be processed by the monitoring image processing apparatus, 2 denotes an A / D converter for quantizing a video input from the video input unit 1, and 3 denotes an A / D converter. A current image storage unit that receives an output from the conversion unit 2 and stores the current video data, and a background 4 that stores an output from the A / D conversion unit 2 and performs a corresponding operation to store a background image for comparison. The image storage unit 5A is a background update unit that calculates and updates data of the background image storage unit 4 according to a certain rule, and generally updates the data according to a simple rule. For example, the average of the past five current images is used as the background image. Also, the background update unit 5A is configured by a recognition processing unit, which will be described later,
The background update method can be switched. One of them is the “inter-frame difference method”. The other background updating method is referred to as “non-frame difference method” for convenience. 6 is the current image storage unit 3
A background difference unit that compares the two images of the background image storage unit 4 and calculates a difference between them; a threshold value calculation unit 7 that calculates a threshold value for binarizing the data of the background difference unit 6; In general, an optimum threshold value is obtained based on background difference data. Further, reference numeral 8 denotes a binarizing unit for binarizing the difference data with the threshold value calculated by the threshold value calculating unit 7, and 9D receives the binary data output from the binarizing unit 8, and outputs a target alert there. A recognition processing unit 10 for determining whether an object exists exists, and a recognition processing unit 9
This is an alarm processing unit that receives an instruction of “discovery target discovery” from D and issues an alarm.

FIG. 23 is a flowchart showing the recognition processing section 9D of FIG. 22 according to the fifth embodiment in more detail.

Step 61 is a change area integration processing part for integrating the change areas considered to belong to the same moving object by looking at the binarized data and combining them into one change area. Next, Steps 62 to 65 are a feature amount comparison portion that determines whether or not the change area integrated in Step 61 is a notification target. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following steps 62 to 65 are determined as examples. In step 62, it is checked whether the change area exists continuously for a certain period of time. In step 63, it is checked whether or not the change area has a predetermined area. In step 64, it is checked whether or not the change area has predetermined vertical and horizontal dimensions.
In step 65, it is determined whether or not the change area has a predetermined moving speed. In response to the above result, in step 66, it is determined whether the alarm target is found or not. However, here, the judgment result is treated as “temporary” and is not yet reported.

In step 67, the alert target change area is tracked and observed for a further fixed time. What is actually observed is the difference between the threshold values calculated by the threshold value calculation unit 7 in the “non-frame difference method” and the “inter-frame difference method”. Receiving the result, in step 68, it is determined whether or not the tentative notification target discovery is correct, and the report is passed to the notification processing unit 10 only when it is correct.

Next, a typical example of the operation will be described. In FIG. 22, a video signal externally supplied to the device is received by a video input unit 1 and sent to an A / D conversion unit 2. The A / D converter 2 converts the received video signal into digital data for subsequent processing. It is common to convert to 8-bit multi-value data per pixel.
The current image storage unit 3 stores the input current image data and prepares for subsequent processing. The background image storage unit 4 stores a background image created based on the input current image data, and prepares for the subsequent processing.

The background updating unit 5A periodically updates the background image stored in the background image storage unit 4 according to a certain rule. This is generally calculated based on the current image that is periodically sampled. For example, the average value of the current image of the past 100 scenes may be used as a background image, or a method may be used for a certain time before (for example, 5 seconds). ) In many cases. Here, it is assumed that update is performed to replace the average value of the current image of the past 100 scenes.

However, here, there is a function of switching the background update method to another method according to an instruction of the recognition processing section 9D. The background updating method that can be switched at that time is, for example, that the current image is directly taken into the background image storage unit 4.

The background difference unit 6 compares the outputs of the current image storage unit 3 and the background image storage unit 4, finds a difference for each pixel, and outputs the difference data as a frame image.
The difference data generally represents a portion of the video where the luminance has changed. Usually, a moving object such as a person or a vehicle is mainly used. However, in some rare cases, a change in luminance that is not a target of alarm, such as the appearance of a shadow of a building due to “shadow of sunlight”, may be exhibited.

However, when the background updating unit 5A changes the background updating method to the “capture the current image as it is into the background image storage unit 4” and restores it, the current
The difference between the current image before the frame is calculated, and the inter-frame difference is inevitably obtained only at this moment. After that, even if the background updating method is restored by the background updating unit 5A, it takes another 100 scenes to return to the background image similar to the original background image. Past 10
(Because it is assumed that updating is performed to replace the average value of the current image of scene 0), the background difference result during that period is also temporarily close to the inter-frame difference. (They are gradually turning to normal background subtraction results.)

The difference data based on the inter-frame difference generally indicates a portion where the luminance changes between one frame. Usually, a moving object such as a person or a vehicle moving at high speed is mainly used.
Due to its nature, a clear difference does not appear with a time difference of about one frame, so that a person moving at low speed cannot extract difference data well. However, relatively large changes that gradually appear without moving or gradually disappear without moving, such as "appearance of shadows of buildings due to the shadows of sunlight" and "water marks of puddles" Similarly, the difference data cannot be extracted from a change in luminance such as “region”.

The threshold value calculating section 7 calculates a threshold value for binarizing the difference data sent from the background difference section 6. Normally, background data is multi-valued data, which is inappropriate for recognition processing as it is. In general, binarization and conversion into binary data are also performed in order to make subsequent determinations at high speed.
However, there is a case where the difference data is directly recognized and processed without performing the binarization depending on the device, so this is a reference.

When comparing the inter-frame difference data with the non-frame difference data, the value of the inter-frame difference data that produces a difference between the approximated image data is generally smaller than that of the non-frame difference data. Therefore, the calculated threshold value is lower than the other threshold value.

The binarizing section 8 binarizes the difference data using the threshold value. As a result, in the subsequent processing, the change area becomes binary data.

The change region will be described with reference to FIG. 26 as an example. FIG. 26 is an image of a normal main gate. Now, a person is moving in the image. In this case, the position of the moving person becomes the change area as it is. FIG. 27 is an image in which the superimposed areas in which the change areas are dyed black are overlapped.

The recognition processing unit 9D receives the binarized data from the binarizing unit 8, and the change area indicated by the binarized data is the target (for example, generally corresponds to a person or a vehicle. Is determined.

FIG. 23 is a content diagram of the recognition processing section 9D of the monitoring image processing apparatus according to the fifth embodiment. The recognition processing unit 9D performs recognition in the following procedure. However, this is only an example.

In the changing area integration process in step 61, the changing areas are considered, and the changing areas that are considered to belong to the same moving object are integrated into one changing area. Next, in steps 62 to 65, the change area integrated in step 61 is
It is determined whether or not the notification is to be issued. Here, it is compared whether or not the change region matches a predetermined feature amount. Here, a change area having a feature that meets a predetermined condition is regarded as an alarm target.

Here, the following judgment is made as an example. In step 62, it is checked whether the change area exists continuously for a certain period of time. In step 63, it is checked whether or not the change area has a predetermined area. In step 64, it is checked whether or not the change area has predetermined vertical and horizontal dimensions. In step 65, it is determined whether or not the change area has a predetermined moving speed. The “area”, “length and width”, and “moving speed” are generally referred to by calculating an average during a certain time in step 62. Even if it is determined in step 66 that the notification target is found, it is treated as "temporary" and is not yet transmitted to the notification processing unit 10.

In step 67, the alert target change area is tracked and observed for a certain period of time. Prior to the tracking, first, the background updating unit 5A is instructed to make the image captured in the background image storage unit 4 the current image. As it is, the current image is compared by the background difference unit 6, and no difference is generated. Immediately, it instructs the preceding background updating unit 5A to return the image captured in the background image storage unit 4 to the original method.

Thereafter, the background image of the background image storage unit 4 returns to the original data after a lapse of 100 scenes (the current background updating method performs updating by replacing the current image with the average value of the current image of the past 100 scenes). However, during that time, the background update is incomplete.

Immediately after the switching, the difference between the current image and the current image one frame before is instantaneously obtained, and inevitably the inter-frame difference only at this instant. However, thereafter, the background image gradually changes to a normal background image, and thus gradually returns to normal background difference data (non-frame difference data).

As described above, in the inter-frame difference data,
"Slowly appearing without moving, or gradually disappearing without moving, such as the appearance of building shadows due to the shadows of sunlight" or "watermarks of puddles" Similarly, the difference data cannot be extracted from a change in luminance such as “a relatively large change area”.

At this time, the calculation result (that is, the threshold value) in the threshold value calculation unit 7 is lower than usual as a result of trying to cut out a change region that is difficult to cut out even a little.
Then, the threshold value is compared with the threshold value at the time of “provisional alert target discovery”, and a change in the value is observed. If the value changes abruptly, it can be determined as abnormal, otherwise it can be determined as normal.

Receiving the result, in step 68, the report of the discovery of the report target is passed to the report processing unit 10 only in the normal state. The alarm processing section 10 performs an alarm processing. Normally,
This is a process of sending a specific signal to an upper-level PC or the like or sounding a buzzer of the apparatus. If an abrupt change in the value is found in step 67, it is determined that the "provisional report target discovery" was incorrect, and no "formal report target discovery" report is made. Therefore, the device does not issue an alert. Note that the recognition processing section 9D and the alert processing section 10 are generally processed by S / W.

As described above, such as “appearance of building shadow due to shading of sunlight” and “appearance of building shadow due to puddle water pattern”, “appears gradually without moving, A relatively large change area that gradually disappears without being accompanied is not issued and does not become unnecessary detection. Conversely, differently from the above, a change area such as "person" is accurately notified.

As described above, according to the fifth embodiment,
It does not impair the ability to fire legitimate objects such as "people", such as the appearance of building shadows due to the shadows of sunlight and the appearance of building shadows due to puddle water marks. It is possible to realize a function of accurately reducing unnecessary detection regarding “a relatively large change area that gradually appears without moving or gradually disappears without moving” more than the effects described in the first to fourth embodiments. .

In other words, instead of "confirming that there is no sudden increase or decrease of the area", "confirming that there is no sudden change in the threshold value of binarization", By confirming that there is no appearance or disappearance, and issuing a notification after waiting for the result, the work of confirming whether or not there is a false report can be completed earlier than in the third and fourth embodiments. It has a function capable of obtaining the same effects as in Embodiments 3 and 4.

[0181]

According to the monitoring image processing apparatus of the present invention,
As described above, the changing area extracting means for extracting the changing area based on the comparison operation between the current image data and the background image data for comparison, the changing area extracting means when the extracted changing area meets a predetermined condition, and And a recognition processing means for recognizing as an alarm target when there is no predetermined change in the area of the image for a certain period of time. Relatively large changes that gradually appear without moving or gradually disappear without moving, such as the appearance of building shadows due to shading or the appearance of building shadows due to puddle water marks It is possible to reduce the unnecessary detection regarding the “region”.

Further, in the monitoring image processing apparatus according to the present invention, as described above, the change area extracting means sets the background difference as a change area based on a comparison operation between the current image data and the background image data for comparison. Background difference means for calculating, threshold value calculating means for calculating a threshold value for binarizing the background difference data based on the background difference data,
Since there is provided a binarizing means for binarizing the background difference data with the threshold value, it does not hinder the alarming ability for a normal alarming target such as "person", and " There is no need for "a relatively large change area that gradually appears without moving or disappears without moving," such as the appearance of shadows of buildings due to "appearance of shadows" and "watermarks of puddles" There is an effect that detection can be reduced.

Further, as described above, the surveillance image processing apparatus according to the present invention further comprises an A / D conversion means for quantizing an input image and the current image data from the output of the A / D conversion means. A current image storage means for storing,
Background image storage means for creating and storing the background image data for comparison from the output of the D conversion means, background update means for calculating and updating data of the background image storage means according to a predetermined rule, and the extracted change area And a notification processing means for notifying the user of the fact that the
It does not impair the ability to fire legitimate objects such as "people", such as the appearance of building shadows due to the shadows of sunlight and the appearance of building shadows due to puddle water marks. This has the effect of reducing unnecessary detection of “a relatively large change area that gradually appears without moving or gradually disappears without moving”.

Further, in the monitoring image processing apparatus according to the present invention, as described above, the change area extracting means performs the comparison operation between the current image data and the background image data for comparison which is the current image data one frame before. First background difference means for calculating the inter-frame background difference as a change area based on the second background difference means for calculating the background difference as a change area based on a comparison operation between the current image data and the background image data for comparison. Means, a threshold value calculating means for calculating first and second threshold values for binarizing the inter-frame background difference data and the background difference data based on the inter-frame background difference data and the background difference data; Binarizing the inter-frame background difference data and the background difference data with the first and second threshold values and outputting first and second binary data;
Value recognizing means, wherein the recognition processing means comprises the first 2
First recognition processing means for recognizing as a tentative alarm target when the extracted change area matches a predetermined condition based on the value data, and determining the extracted change area based on the second binary data And the second recognition processing means for recognizing as a notification target when the first recognition processing means recognizes the temporary notification target. Such as the appearance of shadows of buildings due to the shadows of sunlight and the appearance of shadows of buildings due to water puddles in puddles. Unnecessary detection relating to “a relatively large change area that gradually appears without moving or gradually disappears without moving” can be more accurately reduced.

Further, as described above, the surveillance image processing apparatus according to the present invention further comprises an A / D conversion means for quantizing an input image and the current image data from the output of the A / D conversion means. A current image storage means for storing,
First background image storage means for creating and storing the comparison background image data which is the current image data one frame before the output of the D conversion means, and the comparison background image based on the output of the A / D conversion means A second background image storage unit for creating and storing data; a first background update unit for calculating and updating data of the first background image storage unit according to a predetermined rule;
A second background updating means for calculating and updating the data of the second background image accumulating means according to a predetermined rule; and a notifying processing means for notifying when the extracted change area is recognized as a notification target. It does not impair the ability to fire regular objects such as "people", and the appearance of building shadows due to "shading of the sun due to the shadows of sunlight" and "watermarks of puddles" As described above, it is possible to more accurately reduce unnecessary detection regarding “a relatively large change area that gradually appears without moving or gradually disappears without moving”.

Further, as described above, in the monitoring image processing apparatus according to the present invention, the changing area extracting means performs the comparison operation between the current image data and the background image data for comparison which is the current image data one frame before. First background difference means for calculating the inter-frame background difference as a change area based on the second background difference means for calculating the background difference as a change area based on a comparison operation between the current image data and the background image data for comparison. Means, a threshold value calculating means for calculating first and second threshold values for binarizing the inter-frame background difference data and the background difference data based on the inter-frame background difference data and the background difference data; Binarizing the inter-frame background difference data and the background difference data with the first and second threshold values and outputting first and second binary data;
Value recognition means, and the recognition processing means includes a second
When the extracted change area meets a predetermined condition based on the value data, it is recognized as a provisional alert target, and based on the first binary data, the area of the extracted change area changes for a predetermined time for a predetermined time. When there is no alarm, the temporary alarm target is recognized as the alarm target, so that the alarming ability for the normal alarm target such as "person" is not hindered. More accurate detection of unnecessary detections for "relatively large change areas that gradually appear without moving or gradually disappear without moving", such as the appearance of building shadows due to "" Therefore, there is an effect that the configuration of the apparatus can be simplified.

Further, as described above, the surveillance image processing apparatus according to the present invention further comprises an A / D conversion means for quantizing an input video, and the current image data from the output of the A / D conversion means. A current image storage means for storing,
First background image storage means for creating and storing the comparison background image data which is the current image data one frame before the output of the D conversion means, and the comparison background image based on the output of the A / D conversion means A second background image storage unit for creating and storing data; a first background update unit for calculating and updating data of the first background image storage unit according to a predetermined rule;
A second background updating means for calculating and updating the data of the second background image accumulating means according to a predetermined rule; and a notifying processing means for notifying when the extracted change area is recognized as a notification target. It does not impair the ability to fire regular objects such as "people", and the appearance of building shadows due to "shading of the sun due to the shadows of sunlight" and "watermarks of puddles" It is possible to more accurately reduce unnecessary detection for "a relatively large change area which gradually appears without moving or gradually disappears without moving", and the apparatus for that purpose. There is an effect that the configuration can be simplified.

Further, as described above, the surveillance image processing apparatus according to the present invention further comprises an A / D conversion means for quantizing an input image, and the current image data from the output of the A / D conversion means. A current image storage means for storing,
Background image storage means for creating and storing the background image data for comparison from the output of the D conversion means, background update means for calculating and updating data of the background image storage means according to a predetermined rule, and the extracted change area When the recognized change area matches a predetermined condition, the recognition processing means recognizes the temporary change target when the extracted change area matches a predetermined condition, and The background updating means causes the background image storage means to store comparison background image data which is current image data of one frame before,
When the area of the extracted change area has not been changed for a predetermined period of time based on the inter-frame background difference data, the provisional alarm target is recognized as the alarm target, so that the temporary alarm target such as "person" is recognized. Does not impair the alerting ability, such as "Appearance of building shadows due to sunlight shading" or "Appearance of building shadows due to puddle water marks" Unnecessary detection relating to "a relatively large change region that gradually disappears without accompanying" can be more accurately reduced, and the configuration of the device therefor can be simplified.

As described above, the surveillance image processing apparatus according to the present invention further includes a change area extracting means for extracting a change area based on a comparison operation between the current image data and the background image data for comparison. When the change area meets a predetermined condition, and when the threshold value for binarizing the change area has not been changed for a predetermined period of time for a predetermined time, there is provided recognition processing means for recognizing the threshold as an alarm target. It does not impair the ability to fire regular objects such as "persons" and does not impede the appearance of building shadows due to the shadows of sunlight or the appearance of building shadows due to puddle water marks. Unnecessary detection relating to “a relatively large change region that gradually appears without moving or gradually disappears without moving” can be more accurately reduced.

Further, in the monitoring image processing apparatus according to the present invention, as described above, the change area extracting means sets the background difference as a change area based on a comparison operation between the current image data and the background image data for comparison. Background difference means for calculating, threshold value calculating means for calculating a threshold value for binarizing the background difference data based on the background difference data,
Since there is provided a binarizing means for binarizing the background difference data with the threshold value, it does not hinder the alarming ability for a normal alarming target such as "person", and " There is no need for "a relatively large change area that gradually appears without moving or disappears without moving," such as the appearance of shadows of buildings due to "appearance of shadows" and "watermarks of puddles" There is an effect that detection can be more accurately reduced.

Further, as described above, the surveillance image processing apparatus according to the present invention further comprises an A / D conversion means for quantizing an input image and the current image data from the output of the A / D conversion means. A current image storage means for storing,
Background image storage means for creating and storing the background image data for comparison from the output of the D conversion means, background update means for calculating and updating data of the background image storage means according to a predetermined rule, and the extracted change area When the recognized change area matches a predetermined condition, the recognition processing means recognizes the temporary change target when the extracted change area matches a predetermined condition, and The background updating means causes the background image storage means to store comparison background image data which is current image data of one frame before,
When the threshold value for binarizing the extracted change region based on the inter-frame background difference data does not have a predetermined change for a certain period of time, the tentative alarm target is recognized as the alarm target, so that a "person" is recognized. It does not impede the ability to raise alarms for regular alarms, and does not involve movement, such as the appearance of shadows of buildings due to the shadows of sunlight or the appearance of shadows of buildings due to puddles of water. , The unnecessary detection relating to “a relatively large change area that gradually appears or disappears without accompanying movement” can be more accurately reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a first embodiment of the present invention.

FIG. 2 is a flowchart illustrating an operation of a recognition processing unit according to the first embodiment of the present invention.

FIG. 3 shows an image according to Embodiment 1 of the present invention;
It is an example of a video in which a "person" and a "water crest" of a puddle are shown.

FIG. 4 is an image according to Embodiment 1 of the present invention, in which a puddle has a small water mark;

FIG. 5 is a block diagram showing a configuration of a second embodiment of the present invention.

FIG. 6 is a flowchart illustrating an operation of a recognition processing unit according to the second embodiment of the present invention.

FIG. 7 is a diagram showing an example of an image according to the second embodiment of the present invention;
It is an example of a video in which a "person" and a "water crest" of a puddle are shown.

FIG. 8 is an image according to Embodiment 2 of the present invention, in which a water mark of a puddle is slightly reduced.

FIG. 9 is a diagram showing a result obtained by obtaining difference data by a method of one background difference unit and binarizing the data according to the second embodiment of the present invention;

FIG. 10 is a diagram illustrating a result obtained by obtaining difference data by the method of the other background difference unit and binarizing the data according to the second embodiment of the present invention.

FIG. 11 is a block diagram showing a configuration of a third embodiment of the present invention.

FIG. 12 is a flowchart illustrating an operation of a recognition processing unit according to the third embodiment of the present invention.

FIG. 13 is a video example according to Embodiment 3 of the present invention, in which a “person” and a “water mark” of a puddle are shown in the video.

FIG. 14 is an image according to Embodiment 3 of the present invention, in which a water mark of a puddle is slightly reduced.

FIG. 15 is a diagram showing a result obtained by obtaining difference data by a method of one background difference unit and binarizing the data according to Embodiment 3 of the present invention;

FIG. 16 is a diagram illustrating a result obtained by obtaining difference data by the method of the other background difference unit and binarizing the data according to the third embodiment of the present invention.

FIG. 17 is a block diagram showing a configuration of a fourth embodiment of the present invention.

FIG. 18 is a flowchart illustrating an operation of a recognition processing unit according to the fourth embodiment of the present invention.

FIG. 19 is a video example according to Embodiment 4 of the present invention, in which the video includes a “person” and a “water mark” of a puddle.

FIG. 20 is an image according to Embodiment 4 of the present invention, in which a puddle of a puddle is slightly reduced.

FIG. 21 is a diagram illustrating a result obtained by switching a background update unit to obtain difference data and binarizing the data so that an image to be taken into a background image storage unit is set to a current image according to Embodiment 4 of the present invention;

FIG. 22 is a block diagram showing a configuration of a fifth embodiment of the present invention.

FIG. 23 is a flowchart showing an operation of the recognition processing unit according to the fifth embodiment of the present invention.

FIG. 24 is a block diagram illustrating a configuration of a conventional monitoring image processing apparatus.

FIG. 25 is a flowchart showing an operation of a recognition processing unit of the conventional monitoring image processing apparatus.

FIG. 26 is an image of a normal front gate according to a conventional monitoring image processing apparatus.

FIG. 27 is a video showing a superimposition in which a change area is dyed black, according to a conventional monitoring image processing apparatus.

[Explanation of symbols]

1 video input unit, 2 A / D conversion unit, 3 current image storage unit, 4, 4a, 4b background image storage unit, 5, 5a, 5b,
5A background update section, 6, 6a, 6b background difference section, 7
Threshold value calculation unit, 8 binarization unit, 9A, 9a, 9b, 9
B, 9C, 9D recognition processing unit, 10 alarm processing unit.

Claims (11)

[Claims] A changing region extracting means for extracting a changing region based on a comparison operation between the current image data and the background image data for comparison, wherein the extracted changing region matches a predetermined condition;
And a recognition processing means for recognizing as an alarm target when the area of the change area has not changed for a predetermined time. 2. The image processing apparatus according to claim 1, wherein the change area extracting means calculates a background difference as a change area based on a comparison operation between current image data and comparison background image data, and the background difference based on the background difference data. 2. The monitoring system according to claim 1, further comprising: a threshold value calculating unit that calculates a threshold value for binarizing data; and a binarizing unit that binarizes the background difference data with the threshold value. 3. Image processing device. 3. An A / D converter for quantizing an input video, a current image storage for storing the current image data from an output of the A / D converter, and an output of the A / D converter. A background image accumulating unit that creates and stores the comparison background image data from; a background updating unit that calculates and updates data of the background image accumulating unit according to a predetermined rule; 3. The monitoring image processing apparatus according to claim 2, further comprising: a notification processing unit that notifies the user of the recognition. 4. The method according to claim 1, wherein the changing area extracting means calculates a background difference between frames as a changing area based on a comparison operation between the current image data and the background image data for comparison which is current image data of one frame before. Background subtraction means, second background difference means for calculating the background difference as a change area based on a comparison operation between the current image data and the background image data for comparison, and the background difference data between frames and the background difference data Threshold value calculating means for calculating first and second threshold values for binarizing the inter-frame background difference data and the background difference data; and calculating the first and second threshold values of the inter-frame background difference data and the background difference data. Binarizing means for binarizing the data with a threshold value and outputting first and second binary data, wherein the recognition processing means performs the extraction based on the first binary data. A first recognition processing unit for recognizing a temporary alert target when the change area matches a predetermined condition; and a case where the extracted change area matches the predetermined condition based on the second binary data. 2. The monitoring image according to claim 1, further comprising second recognition processing means for recognizing the temporary notification target as the notification target when the first recognition processing unit recognizes the temporary notification target. Processing equipment. 5. An A / D converter for quantizing an input video, a current image storage for storing the current image data from an output of the A / D converter, and an output of the A / D converter. First background image storage means for creating and storing the background image data for comparison which is the current image data one frame before, and creating the background image data for comparison from the output of the A / D conversion means A second background image storage unit for storing; a first background update unit for calculating and updating data of the first background image storage unit according to a predetermined rule; 5. The apparatus according to claim 4, further comprising: second background update means for calculating and updating according to the following rule: and alarm processing means for notifying when the extracted change area is recognized as an alarm target. Monitoring image processing device. 6. The method according to claim 1, wherein the change area extracting means calculates a background difference between frames as a change area based on a comparison operation between the current image data and the background image data for comparison which is the current image data of one frame before. Background subtraction means, second background difference means for calculating the background difference as a change area based on a comparison operation between the current image data and the background image data for comparison, and the background difference data between frames and the background difference data Threshold value calculating means for calculating first and second threshold values for binarizing the inter-frame background difference data and the background difference data; and calculating the first and second threshold values of the inter-frame background difference data and the background difference data. Binarizing means for binarizing the data with a threshold value and outputting first and second binary data, wherein the recognition processing means performs the extraction based on the second binary data. If the change area meets a predetermined condition, the change area is recognized as a tentative alert target, and if the area of the extracted change area does not change for a predetermined time based on the first binary data, the tentative alarm is issued. 2. The monitoring image processing apparatus according to claim 1, wherein the notification target is recognized as a notification target. 7. An A / D converter for quantizing an input video, a current image storage for storing the current image data from an output of the A / D converter, and an output of the A / D converter. First background image storage means for creating and storing the background image data for comparison which is the current image data one frame before, and creating the background image data for comparison from the output of the A / D conversion means A second background image storage unit for storing; a first background update unit for calculating and updating data of the first background image storage unit according to a predetermined rule; 7. The apparatus according to claim 6, further comprising: second background update means for calculating and updating according to the following rule: and alarm processing means for notifying when the extracted change area is recognized as an alarm target. Monitoring image processing device. 8. An A / D converter for quantizing an input video, a current image storage for storing the current image data from an output of the A / D converter, and an output of the A / D converter. A background image accumulating unit that creates and stores the comparison background image data from; a background updating unit that calculates and updates data of the background image accumulating unit according to a predetermined rule; And a report processing means for notifying the recognition when the recognition is made, wherein the recognition processing means recognizes the extracted change area as a provisional report target when the extracted change area meets a predetermined condition, and the background update means The background image storage means stores comparison background image data which is current image data of one frame before, and the area of the extracted change region is determined based on the inter-frame background difference data for a predetermined time. Of the monitoring image processing apparatus according to claim 2, wherein the recognizing the temporary calling report object as alarm object in the absence. 9. A changing area extracting means for extracting a changing area based on a comparison operation between current image data and background image data for comparison, wherein the extracted changing area matches a predetermined condition,
And a recognition processing means for recognizing a threshold value for binarizing the change area as a target for notification when there is no predetermined change for a predetermined time. 10. The change area extracting means calculates a background difference as a change area based on a comparison operation between current image data and comparison background image data, and the background difference data based on the background difference data. 10. The monitoring system according to claim 9, further comprising: a threshold value calculating unit that calculates a threshold value for binarizing data; and a binarizing unit that binarizes the background difference data with the threshold value. Image processing device. 11. An A / D for quantizing an input image.
Conversion means; current image storage means for storing the current image data from the output of the A / D conversion means; background image storage for creating and storing the background image data for comparison from the output of the A / D conversion means Means, a background updating means for calculating and updating data of the background image accumulating means according to a predetermined rule, and a report processing means for notifying when the extracted change area is recognized as a report target, and The recognition processing means recognizes the extracted change area as a provisional alert target when the extracted change area meets a predetermined condition, and compares the current image data of the current image data of the previous frame to the background image storage means by the background updating means. When the threshold value for binarizing the extracted change area based on the inter-frame background difference data has not been changed for a predetermined period of time, the temporary generation is performed. Monitoring the image processing apparatus according to claim 10, wherein recognizing the target as alarm object.