JPH11275581A - Abnormal image detecting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To always detect the abnormality of a compressed digital image with an appropriate sensitivity by deciding it as abnormal when the ratio of the specific value of mean deviation of code quantity per coded unit and the variance exceed preliminarily set threshold. SOLUTION: An image compressing part 1 encodes digital image data in a frame unit by a JPEG system. An abnormality detecting part 2 stores JPEG code quantity per frame, calculates the average and the variance from values for the past fixed period, decides it as abnormal when the ratio of the means deviation square of the latest code quantity and the variance exceeds preliminarily set threshold and notifies it to an image code storing part 3. Thus, even when code quantity per coded unit includes trend fluctuation and random fluctuation, abnormal detection sensitivity can be appropriately changed in accordance with the fluctuations. When the part 3 receives an abnormal notice, it performs attribute change of image codes existing in a temporary storage area into a storage area, notifies the abnormality externally and can transfer the stored image codes in response to an external request.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an abnormal image detecting apparatus for continuously inspecting a change in an image and detecting the change in the image as abnormal, and more particularly to an abnormal image detecting apparatus based on a code amount output when an image is compressed. The present invention relates to an abnormal image detection device that detects an error.

[0002]

2. Description of the Related Art Conventionally, a remote monitoring system as shown in FIG. 6 is known as Japanese Patent Publication No. 9-205640. In this device, an image input unit 31 captures an image and supplies a digital image to an image compression unit 32. The image compression unit 32 continuously compresses the input digital image. The image compression unit 32 supplies the image code to the data transmission unit 35 and supplies the code amount to the code amount comparison unit 33. Code amount comparison unit 33
Compares the code amount with the reference code amount stored in the reference code amount storage unit 34, and determines an abnormality if the difference is equal to or larger than a preset threshold value. That is, the normal state of the image is set as a state in which the code amount falls within a range of a certain threshold value with respect to the reference code amount, and when the code amount is excessively increased or decreased, it is determined to be abnormal. Here, as the reference code amount, an image photographed in a normal state is used as a standard image, and the code amount is used as a fixed value, or the reference code amount is used along with time using the code amount of the image a fixed time ago. It can be used while being updated one after another, and statistical information can be used as the threshold.

[0003]

However, in the conventional example, when the code amount of the standard image is used as the reference code amount, an abnormality is detected when the image code amount in a normal state has a trend that gradually changes. There is a problem that the detection sensitivity shifts, and the sensitivity becomes sensitive on one side of the reference code amount, while the sensitivity becomes insensitive on the other side. In order to deal with this problem, in the conventional example, the code amount of the image a predetermined time before could be used as the reference code amount, but in this case, it is possible to cope with a time-dependent slowly changing fluctuation due to a trend. Is affected by random fluctuations included in the image code amount. In other words, since the code amount of the image before a certain time includes a random fluctuation component, there is a problem that the sensitivity of abnormality detection deviates from a desirable state when a sudden large fluctuation occurs. As described above, in the conventional example, there is a problem that the detection sensitivity is affected by the fluctuation when the code amount of the captured image has the trend fluctuation and the random fluctuation. Due to this problem, the conventional apparatus has to perform an operation test with an actually photographed image, and adjust the sensitivity of abnormality detection by adjusting a threshold value, for example.
It takes time and effort to install, and because the standard of adjustment is not clear, the sensitivity of the adjustment may vary due to human factors, resulting in the responsibility of the person who made the adjustment to make sure that the adjustment was appropriate. There were problems such as getting lost. Furthermore, if the characteristics of the fluctuation indicated by the image code amount change, it is still necessary to adjust the threshold value to correct this, but with a conventional device, this fluctuation can be automatically corrected. There was a problem with the use of.

[0004]

According to an embodiment of the present invention, there is provided an image compression unit that encodes a digital image using a fixed number of frames as an encoding unit, and an image compression unit configured to output the image. An abnormal image detecting apparatus comprising: an abnormality detecting unit that detects an abnormality based on a code amount for each coding unit to be processed. The abnormality detection unit obtains a square and a variance of an average deviation of the code amount for each coding unit. An abnormal image detecting apparatus is characterized in that when the ratio of these two values exceeds a preset threshold value, it is determined that an abnormality has occurred. The image compression unit
In order to reduce the data amount of a digital image, intra-frame coding or inter-frame coding is performed. When the coding unit is one frame, all codes are intra-frame codes. As the intra-frame encoding method, a known still image compression method such as the JPEG method, or a known moving image encoding method such as MPEG or H.261 or H.263 intra-picture encoding method can be used. . When the coding unit of the image compression unit is composed of a plurality of frames, inter-frame coding can be used, and as the method, a known MPEG method,
The H.261 system, the H.263 system, and the like can be used.
The abnormality detection unit receives the code amount for each coding unit as input, obtains an average and a variance from the values for a certain time, and obtains the square of the average deviation of the latest code amount. An abnormality is determined when the ratio of the square of the average deviation to the variance exceeds a preset threshold. It is not necessary to perform division to find the ratio. To make this determination easily, find the product of the variance and a preset threshold, and determine the magnitude of the square of the mean deviation and the product of the variance multiplied by the threshold. Is determined, and when the square of the average deviation is larger, it is determined to be abnormal. Further, the abnormality detection unit performs a trend removal process from the input code amount for each coding unit, and performs abnormality determination on the processed code amount. When the abnormality detection unit detects an abnormality, the data communication unit starts data communication to a preset report destination and reports the abnormality. The remote control unit that performs data communication instructs start and stop of the abnormality detection operation by performing data communication with the abnormal image detection device. Further, when the abnormality detecting unit detects the abnormality, the data communication unit reports the abnormality to the remote control unit and transmits the stored image code.

[0005]

Next, an embodiment of an abnormal image detecting apparatus according to the present invention will be described with reference to the drawings. FIG. 1 shows a first embodiment of the present invention. The image compression unit 1 receives digital image data as input, and performs intra-frame encoding in a JPEG system on a frame basis. Digital image data is obtained from an external camera or video device, and is input to the image compression unit 1 in a format that can be processed as image data. The code amount output from the image compression unit 1 by JPEG encoding is supplied to the abnormality detection unit 2, and the JPEG code is supplied to the image code storage unit 3. The image code storage unit 3 receives the supplied JPEG
The memory is managed in two types, a storage area for temporarily storing codes and a storage area for storing and storing codes.
The storage area for temporarily storing is managed as a ring memory, and overwrites the old image code with the new image code when there is no free space. As a result, an image of a recent certain time is always stored.

[0006] When the JPEG code amount per frame is received from the image compression unit 1, the abnormality detection unit 2 stores the JPEG code amount per frame, and calculates the average and variance from the values for the past fixed time. Next, the product of the square of the average deviation of the latest code amount and the variance multiplied by a preset threshold is obtained. Next, the magnitude of the square of the average deviation and the product of the variance and the threshold value are compared, and if the square of the average deviation is larger, the abnormality detection unit 2 notifies the image code accumulation unit 3 of the abnormality. Upon receiving the notification of the abnormality from the abnormality detection unit 2, the image code storage unit 3 changes the management attribute from the area in the temporary storage area for temporarily storing the image code up to a predetermined time before the notification time to the area for storage and storage. Is changed, and the image code after a predetermined time from the notification time is stored as a storage area for storage so as not to be discarded. The time to accumulate at the time of abnormality notification is
An appropriate time may be selected in consideration of the memory capacity and the generation speed of the image code amount, and the time can be arbitrarily set between 1 and 20 seconds.

In order to reproduce the image code stored in the image code storage section 3, the data communication port provided in the image code storage section 3 and the data communication port of the personal computer are connected by a cable, and the image code is transferred. Images are reproduced using image reproduction software on a personal computer.

FIG. 2 is a flowchart for explaining the operation of the abnormality detecting section. In FIG. 2, 10 is the start, 21
Is over. Processing 11 is counter initialization, sum and 2
This is the process of initializing the variables of the sum of squares. At this point,
The average and variance of the code amount are not determined, and it is not possible to judge the abnormality detection for a while after the start of the process. Process 12 is a process of receiving the code amount, and the image compressing unit 1 notifies the code amount. Process 13 is a process for obtaining the sum and the sum of squares. The judgment process 14 is a process for judging whether or not the value of the counter is equal to or larger than a predetermined value N. If the counter is equal to or larger than N, the process 15 is performed. Process 15 is a process for obtaining an average and a variance from the sum and the sum of squares, multiplying the variance by a preset threshold, and resetting the counter to zero.

The average may be obtained by dividing the sum by N.
The sum of squares is divided by N to obtain a mean square, and the mean square may be subtracted from the mean square. N may be determined according to the characteristics of the frame rate and the code amount variation, but is set to a value such as 16 for example. The threshold value determines the sensitivity of abnormality detection and can be adjusted to an appropriate value empirically, but may be set between 1 and 4. This assumes that the variation of the code amount follows a normal distribution, sets a rough value based on the probability,
It may be adjusted later according to the actual operation. Judgment process 14
If the counter is less than the fixed value N in step 16, the counter is incremented by one in step 16.
Subsequent to the processing 15 and the processing 16, in a processing 17, the square of the average deviation is obtained.

This is a process for subtracting the average from the code amount notified in the process 12 and obtaining the square thereof. In the next determination process 18, it is determined whether the square of the average deviation is larger than the product of the variance and the threshold value. If the square of the average deviation is larger, it is determined to be abnormal. Alternatively, a condition that the product of the variance and the threshold value is not 0 may be added in the determination processing 18. In this case, the abnormality may not be notified to the initial operation Note in which the variance is 0. In the case of an abnormality, the abnormality is notified to the image code storage unit 3 and the like in a process 19. If it is not abnormal, the process proceeds to the determination process 20.

Next, a judgment process 20 checks the end flag. The end flag is for terminating the flow in response to an external request, and is normally set to continue processing without terminating. If the processing is completed, the process proceeds to 21. If the processing is not completed, the process returns to step 12 to receive the notification of the code amount. In this flow, the average and the variance are obtained from the past N code amounts, and these are constant while the N code amounts are notified thereafter, but it is not always necessary to perform such processing.
Every time the code amount is notified, a new average and a new variance may be calculated again.

FIG. 3 shows an abnormality detection unit 2 based on the generated code amount.
3 is a graph for explaining the operation of FIG. The horizontal axis is the number of the coding unit, and here is equal to the frame number. The vertical axis of the upper graph in FIG. 3 represents the code amount per coding unit, and here is the code amount of the JPEG code per frame. The code amount increases as the frame number changes,
A state in which the amount of code returns to the original level after an abnormality is indicated is shown. However, the code amount includes a slight random fluctuation component, and a level line 27 representing the average is shown in the graph.

The vertical axis of the lower graph in FIG. 3 is the square of the average deviation, and the horizontal line 28 is a level line indicating the value obtained by multiplying the variance by a threshold value. A point beyond the level line 28 is a point determined to be abnormal, and is indicated by an arrow below the horizontal axis. Since the magnitude of the variance is determined by the magnitude of the past change in the code amount, the level line 28 moves upward in an environment with large fluctuations, and the sensitivity of abnormality detection becomes low. Conversely, in an environment with small fluctuations, the level line 28 moves downward and the sensitivity of abnormality detection becomes more sensitive.

FIG. 4 is a graph for explaining the operation of the abnormality detecting section 2 when the code amount per frame has drift fluctuation and random fluctuation. As in FIG. 3, the horizontal axis of the graph represents the frame number, the vertical axis represents the code amount, the middle graph represents the code amount from which trend fluctuation has been removed, and the lower graph represents the square of the average deviation. The upper graph shows a case where an abnormality has occurred in a situation having a drift fluctuation such that the code amount gradually increases. By the process of removing the drift fluctuation, values scattered randomly around the horizontal axis are obtained in the middle graph.

When the average and the variance of the value subjected to the trend removal processing are obtained, the average value substantially overlaps the horizontal axis. When the square of the average deviation is obtained from this new value, FIG.
In the lower graph, the horizontal line 29 is a level line representing the product of the variance and the threshold value. A point where the square of the average deviation exceeds the level line 29 is a point to be determined as abnormal, and is indicated by an arrow below the horizontal axis. If the trend fluctuation is not removed, the variance of the code amount becomes large due to the influence of the trend fluctuation and the random fluctuation, the level line 29 moves upward, and the sensitivity of the abnormality detection becomes low. However, the variance of the value after detrending is due to only the random fluctuation component,
The variance becomes smaller and the sensitivity of abnormality detection becomes more sensitive.

FIG. 5 shows a second embodiment of the present invention. FIG.
In FIG. 1, in addition to FIG. 1, a data communication unit 4 for reporting an abnormality is provided. The data communication means includes a PHS (Personal Handy Phone System) telephone used in Japan, and is connected to the ISDN network 5 through a PHS line. On the other hand, the remote control device 6 is also connected to the ISDN network 5 and receives a telephone call for reporting an abnormality. The data communication means 4 and the remote control device 6 communicate with each other up to 29.2 using the PIAFS protocol.
Error-free data communication is possible at a communication speed of kbps.
When the abnormality detecting unit 2 detects an abnormality, the abnormality detecting unit 2 notifies the image code storage unit 3 of the abnormality, and further notifies the data communication unit 4 of the abnormality. Upon receiving the notification of the abnormality, the image code storage means 3 stores and stores the image codes before and after that time. When the data communication means 4 receives the notification of the abnormality, the data communication means 4 makes a PHS call to a preset report destination. The remote control device 6 monitors the telephone call of the call destination telephone, and when the telephone is connected according to the call, the connection of the PIAFS protocol is started. When the PIAFS connection is completed, data communication between the data communication means 4 and the remote control device 6 becomes possible, and data for notifying abnormality is transmitted.

Thereafter, the image code stored in the image storage unit and the image code stored for a predetermined time after the abnormality is detected are transmitted to the remote control terminal 6. Change all stored image codes to 2
It takes 10 minutes to several tens of minutes to transfer at a communication speed of 9.2 kbps. The remote control unit 6 plays a role of controlling the operation of the abnormality detection unit 2 in a normal state, in addition to receiving the notification of the abnormality. In other words, the remote control unit 6 sets a threshold for multiplying the variance, starts and ends abnormality monitoring, and sets a time when performing abnormality monitoring using a timer. The remote control unit 6 also sets the telephone number of the report destination to be called when the data communication unit 4 detects an abnormality, the compression ratio of the JPEG encoder of the image compression unit 1, the number of seconds of images stored in the image storage unit 3, and the like. Do from.

[0018]

According to the abnormal image detecting device of the present invention,
It is possible to monitor an input digital image and detect an image abnormality without using an expensive image processing device. According to the present invention, even when the code amount per image coding unit includes a trend fluctuation and a random fluctuation, the sensitivity of abnormality detection changes according to the magnitude of the fluctuation,
Anomalies are always detected with appropriate sensitivity. For this reason, it can be installed in a wide place including a place where the code amount fluctuates. Also, it is almost unnecessary to manually adjust the sensitivity of the abnormality detection. Further, according to the present invention, it is possible to accumulate the image codes before and after the time when the abnormality is detected in the image code storage unit, and record the image when the abnormality is detected. Further, when an abnormality is detected, the abnormality can be promptly reported to a remote place using the data communication means, and the stored image code can be transmitted.

[Brief description of the drawings]

FIG. 1 is a block diagram of an abnormal image detection device according to the present invention.

FIG. 2 is a flowchart illustrating processing of an abnormality detection unit.

FIG. 3 is a graph illustrating a detection operation of an abnormality detection unit.

FIG. 4 is a graph illustrating a detection operation of an abnormality detection unit when there is drift fluctuation.

FIG. 5 is a second embodiment of the abnormal image detection device according to the present invention.

FIG. 6 is a block diagram of a conventional abnormal image detection device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image compression part 2 Abnormality detection part 3 Image code storage part 4 Data communication means 5 PHS network and ISDN network 6 Remote control means 10 Start 11 Counter and variable initialization processing 12 Notification of code amount 13 Calculation of sum and square sum Processing 14 Counter comparison processing 15 Processing such as updating of average and variance 16 Counter increment processing 17 Processing to find the square of the average deviation 18 Comparison processing of the square of the average deviation 19 Processing to notify abnormality 20 Processing to check the end flag 21 End 27 Level line representing average of code amount 28 Level line representing product of code amount variance and threshold value 29 Level line representing product of code amount variance and threshold value 31 Image input unit 32 Image compression unit 33 Code amount comparison Unit 34 reference code amount storage unit 35 data transmission unit

Claims (6)

[Claims] An abnormal image detecting apparatus for detecting an abnormality by comparing an input image with a reference image, comprising: an image compression unit for encoding a fixed number of frames of a digital image as an encoding unit; An abnormality detection unit that detects an abnormality based on a code amount for each coding unit output by the unit,
The abnormality detection unit calculates the average deviation of the code amount for each coding unit by 2
An abnormal image detection apparatus that calculates a power and a variance, and determines that an abnormality occurs when a ratio of the two values exceeds a preset threshold. 2. An abnormal image detecting apparatus in which an abnormality is detected by comparing an input image with a reference image,
An image compression unit that encodes a fixed number of frames of a digital image as an encoding unit, an abnormality detection unit that detects an abnormality based on the code amount of each encoding unit output by the image compression unit, and an abnormality detection unit. The unit performs a trend removal process on the code amount for each coding unit, obtains the square and the variance of the average deviation of the code amount from which the trend has been removed, and the ratio of these two values exceeds a preset threshold. An abnormal image detection device that determines an abnormality in a case. 3. An image code accumulating section for accumulating an image code output from the image compressing section, wherein when the abnormality detecting section determines that the image is abnormal, the image codes before and after the determination are stored in the image code accumulating section. 2. The abnormal image detecting device according to claim 1, wherein the abnormal image is stored. 4. An abnormal image according to claim 1, further comprising data communication means for reporting an abnormality, wherein the abnormality is reported by the data communication means when the abnormality detecting section determines that the abnormality is abnormal. Detection device. 5. The apparatus according to claim 3, further comprising remote control means for performing data communication with the data communication means, wherein the remote control means instructs the abnormality detection section to start and stop an abnormality detection operation. Abnormal image detection device. 6. An image code accumulating section for accumulating an image code output from the image compressing section, and when the abnormality detecting section determines that an abnormality has occurred, the image codes before and after the determination are stored in the image code accumulating section. 6. The abnormal image detecting apparatus according to claim 5, wherein the data communication means stores the information and notifies the data control means of the abnormality, and further transmits the stored image codes before and after the determination to the remote control means by the data communication means. .