JPH0916254A - Plant abnormality detecting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically generate a mask in a short time without reference to the skillfulness of an operator, to detect abnormality occurring to an outline part, and remove noise except at the outline part. SOLUTION: The plant abnormality detecting device 1 converts a monitor image from a television camera 11 into a digital image by an A/D conversion part 2 and finds the difference image based on a reference image stored in a memory 7 for reference image storage by a difference processing part 3. A mask generation part 6 performs nonlinear luminance conversion for plural difference images found by the difference generation part 3 when an image state decision part 5 decides that there is no abnormality, and stores them in a memory 9 for mask generation and also adds them together to generate a gradation mask. This gradation mask is stored in a memory 10 for mask storage. A mask processing part 4 removes noise components by using the gradation mask stored in the memory 10 for mask storage and stores the result in a difference image storage memory 8. An image state decision part 5 decides whether or not abnormality occurs to the difference image stored in the memory 8.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a plant abnormality detecting device for automatically detecting abnormality in a plant.

[0002]

2. Description of the Related Art In recent years, various devices have been proposed for automatically detecting abnormality in a plant by image processing without the need for a person to visit the site or detect a video image from a surveillance television. Among these devices, when detecting abnormalities such as oil leaks, steam leaks, human invasion, smoke generation, etc., compare the image in which no abnormality has occurred and the image in which abnormality has occurred, and identify the different parts. There is a device that uses a detection method. An image in a state in which no abnormality has occurred is stored in advance as a reference image and compared with the image captured at the time of inspection. A plurality of reference images are prepared according to the state of the plant, or are automatically updated following changes in the state of the plant. The reference image and the image captured at the time of inspection have features of the image emphasized through a filter or a mask in order to make it easy to detect different portions, and unnecessary noise portions are reduced.

FIG. 12 shows an example of a plant abnormality detecting apparatus in the conventional system. In FIG. 12, 11 is a plurality of television cameras or infrared cameras, and the image signals thereof are selected by the camera selector 12 and the plant abnormality detection device 2 is selected.
Sent to 1. This plant abnormality detection device 21 has A
/ D conversion unit 22, difference processing unit 23, mask processing unit 24,
An image state determination unit 25, a reference image storage memory 26, a difference image storage memory 27, and a mask storage memory 28 are provided.

The plant abnormality detecting device 21 converts the image selected by the camera selector 12 into a digital image by the A / D converter 22 and stores it in the reference image storage memory 26. In the reference image storage memory 26, an image captured in advance by the television camera 11 as shown in FIG. 3 in a normal state in which no abnormality has occurred is stored as a reference image. The reference image previously stored in the reference image storage memory 26 and the current image at the same point as shown in FIG. 4 captured at the time of inspection are compared in the difference processing unit 23, and the difference image as shown in FIG. create. The difference image obtained by the difference processing unit 23 is stored in the difference image storage memory 27 via the mask processing unit 24. The image state determination unit 25 determines whether or not there is an abnormality in the state of the difference image stored in the difference image storage memory 27, and outputs an abnormality signal to the series control computer 13 when the abnormality is detected. The sequence control computer 13 has a plant abnormality detection device 21.
When an abnormal signal is sent from the device, an alarm is issued to notify the observer that an abnormality has occurred.

[0005]

As described above, in the conventional method, the image in the normal state in which no abnormality has occurred is stored as the reference image in the reference image storage memory 26, and is stored in the television camera 11. The abnormal portion is extracted from the difference image with the captured image and the abnormality is determined. At this time, the difference image contains the noise component 16 due to flicker of illumination, minute vibration, etc. as shown in FIG. As a means for removing this noise component, a method of masking a noise portion using a mask processing unit 24 and a mask storage memory 28 as shown in FIG. 12 is generally used.

Conventionally, as a method for producing this mask, FIG.
As shown in FIG. 0, a window-type mask creating method for inspecting only a portion where an abnormality is considered to occur in advance or a contour mask creating method as shown in FIG. Has been. In the masks shown in FIGS. 10 and 11, the reference numeral 18 indicates a portion for which abnormality determination is performed, and the reference numeral 19 indicates a portion for which abnormality determination is not performed. However, with the window-type mask shown in FIG. 10, the operator must create a mask for each inspection point, and the accuracy of the inspection is affected by the skill of the operator. However, there is a drawback that it is difficult to update the mask. Further, in the contour mask shown in FIG. 11, since the contour portion is masked, it is not possible to detect an abnormality that has occurred near the contour, and noises other than those near the contour cannot be removed.

The present invention has been made in view of the above circumstances, and a mask can be automatically produced in a short time and stable performance can be maintained regardless of the skill of an operator.
Moreover, it is an object of the present invention to provide a plant abnormality detection device capable of detecting an abnormality that has occurred in the contour portion and removing noise other than the contour portion.

[0008]

SUMMARY OF THE INVENTION The present invention provides a plant abnormality detection apparatus for capturing an image taken by a monitoring camera installed at an arbitrary monitoring place in a plant to detect the presence or absence of abnormality in the plant. A / D conversion means for converting the photographed image into a digital image, difference processing means for obtaining a difference image between the image captured via this means and a reference image in which no abnormality has occurred, and output from this means. Mask creating means for creating a gradation mask by performing non-linear luminance conversion and addition of a plurality of normal-time difference images, a mask storage memory for storing the gradation mask created by this means, and this memory Mask processing means for removing noise components from the difference image output from the difference processing means by using the gradation mask stored in Characterized by comprising a determination means for determining whether an abnormality has occurred based on that difference image.

(Operation) An image of the monitoring area of the plant is taken by the monitoring camera, and the taken image is
The digital images are sequentially converted by the A / D conversion means. A difference image between the digitized image to be monitored and the reference image is obtained by the difference processing means. And
In a state where no abnormality has occurred, nonlinear brightness conversion is performed on a plurality of difference images output from the difference processing means and addition processing is performed to create a gradation mask, which is stored in the mask storage memory. Then, the noise component is removed from the difference image output from the difference processing means by using the gradation mask stored in the mask storage memory. After that, the presence / absence of abnormality is determined by the determination means based on the difference image from which the noise component is removed.

By creating a gradation mask from an actual captured image as described above, the mask creation work by an operator is not required, and the mask is automatically created in a short time regardless of the skill of the operator. Can be created. Further, since the mask creation method does not depend on the contour portion, it is possible to detect an abnormality that has occurred in the contour portion and reliably remove noise other than the contour portion.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows an example of constructing an abnormality detection system using the plant abnormality detection device according to the present invention. Further, FIG. 2 is a flowchart showing the processing operation of the abnormality detection in FIG. In FIG. 1, reference numeral 11 denotes a plurality of monitoring cameras for monitoring a plant abnormality, for example, a television camera (or an infrared camera), the image output of which is selected by a camera selector 12 and sent to the plant abnormality detection apparatus 1.

The plant abnormality detection apparatus 1 includes an A / D conversion unit 2, a difference processing unit 3, a mask processing unit 4, an image state determination unit 5, a mask creation unit 6, a reference image storage memory 7, and a difference image storage memory. 8, a mask creation memory 9 and a mask storage memory 10. The reference image storage memory 7 stores in advance an image to be monitored in a state where no abnormality has occurred, for example, a reference image as shown in FIG. In FIG. 3, 14 is a device to be monitored.

A series control computer 13 is connected to the downstream side of the plant abnormality detection device 1 and issues an alarm when the plant abnormality detection device 1 outputs an abnormality signal. Next, the operation of the plant abnormality detection device configured as described above will be described with reference to the flowchart of FIG.

An image of the monitoring area of the plant is taken by the television camera 11. This TV camera 1
The images photographed by 1 are collected by the camera selector 12, one required image is selected and sent to the plant abnormality detection device 1 (step A1).

Now, assuming that an image including the abnormality occurrence portion 15 is sent from the camera selector 12 to the plant abnormality detecting apparatus 1 as shown in FIG.
It is converted into a digital image by the / D converter 2. The digitized monitoring target image is sent to the difference processing unit 3 and subjected to difference processing with the reference image shown in FIG. 3 stored in the reference image storage memory 7 (step A).
2). The difference image obtained by the difference processing unit 3 contains a noise component 16 in addition to the abnormality occurrence portion 17 as shown in FIG. 5, for example. The difference image including the noise component 16 is then removed by the mask processing unit 4 using the mask image stored in the mask storage memory 10 (step A3), as shown in FIG. The differential image of only the abnormal portion 17 is obtained and stored in the differential image storage memory 8. Then, the image state determination unit 5 processes the difference image stored in the difference image storage memory 8 and determines whether there is an abnormality (steps A4, A
5) If there is an abnormality, an abnormality signal is output to the series control computer 13 (step A6). The sequence control computer 13 gives an alarm when an abnormal signal is sent from the plant abnormality detection device 1, and informs the supervisor that an abnormality has occurred.

On the other hand, in the mask creating section 6, when the image state judging section 5 judges that there is no abnormality, the difference processing section 3 outputs an image containing a noise component 16 as shown in FIG. Non-linear luminance conversion is performed according to the non-linear graph (luminance conversion characteristic 20) shown in FIG. 9 (step A7) and stored in the mask creating memory 9. When a certain number of images are stored in the mask creating memory 9, the mask creating unit 6 adds all the stored images and creates a gradation mask as shown in FIG. Then, this gradation mask is transferred to the mask storage memory 10 at a predetermined timing, and the stored contents are updated (step A8). In the gradation mask shown in FIG. 8, reference numeral 18 denotes a portion that makes an abnormality determination, and 19 denotes a portion that does not make an abnormality determination.

The method of forming the gradation mask will be described in more detail. The difference processing unit 3 creates a difference image as shown in FIG. 7 from the basic image as shown in FIG. 3 stored in the reference image storage memory 7 and the captured image in a state where no abnormality has occurred at the same point. To do. At the same point,
Repeat this work multiple times to create multiple difference images,
It is stored in the mask forming memory 9. As shown in FIG. 7, these difference images are images that show the noise component 16 due to flicker of illumination, minute vibrations, etc. By adding all of these difference images, a mask with gradation can be obtained. Created. However, the difference image has A
The / D conversion error, noise peculiar to the video signal, and the like are also included, and these components are enhanced when added, and the required gradation mask cannot be obtained. However, paying attention to the fact that the A / D conversion error, the noise of the video signal, and the like are minute amounts compared with the noise caused by the flicker of the illumination and the minute vibration, and the difference image is nonlinear according to the nonlinear graph shown in FIG. Shape-to-luminance conversion is performed to remove a small amount of noise. A target gradation mask is created by adding the difference images from which this minute amount of noise has been removed.

By thus creating the gradation mask from the actual captured image, the mask creation work by the operator is not required, and the mask is automatically created in a short time regardless of the skill of the operator. can do. Further, since the mask creation method does not depend on the contour portion, it is possible to detect an abnormality that has occurred in the contour portion and reliably remove noise other than the contour portion.

[0019]

As described above in detail, according to the present invention, a plurality of difference images are created from a basic image and a captured image in a state where no abnormality has occurred at the same point, and the difference images are non-linear. Shape and brightness conversion is performed to remove a small amount of noise, and then a gradation mask is created by adding the noise, so a mask can be created automatically in a short time, and it is possible to relate to the skill of the operator. Thus, stable performance can be maintained, an abnormality occurring in the contour portion can be detected, and noise other than the contour portion can be removed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a plant abnormality detection apparatus according to an embodiment of the present invention.

FIG. 2 is a flowchart of abnormality detection according to the same embodiment.

FIG. 3 is a conceptual diagram showing an image serving as a criterion for determination.

FIG. 4 is a conceptual diagram showing an image when an abnormality occurs.

FIG. 5 is a conceptual diagram showing a difference image before noise removal when an abnormality occurs.

FIG. 6 is a conceptual diagram showing a difference image after noise removal when an abnormality occurs.

FIG. 7 is a conceptual diagram showing a difference image before noise removal when no abnormality has occurred.

FIG. 8 is a conceptual diagram showing a gradation mask in the present invention.

FIG. 9 is a conversion graph used when performing brightness conversion of a gradation mask.

FIG. 10 is a conceptual diagram showing a conventional window mask.

FIG. 11 is a conceptual diagram showing a conventional contour mask.

FIG. 12 is a block diagram showing the configuration of a plant abnormality detection device using a conventional method.

[Explanation of symbols]

 1 Plant Abnormality Detection Device 2 A / D Conversion Unit 3 Difference Processing Unit 4 Mask Processing Unit 5 Image State Determination Unit 6 Mask Creation Unit 7 Reference Image Storage Memory 8 Difference Image Storage Memory 9 Mask Creation Memory 10 Mask Storage Memory 11 TV Camera or Infrared Camera 12 Camera Selector 13 Series Control Calculator 14 Monitored Device 15 Abnormality Occurring Area 16 Noise Component 17 Abnormality Occurring Area 18 Abnormality Determining Area 19 Abnormality Determining Area 20 Luminance Conversion Characteristics

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G06T 1/00 G08B 21/00 E G08B 21/00 25/00 510M 25/00 510 G06F 15/62 380

Claims (1)

[Claims] 1. A plant abnormality detecting apparatus for capturing an image taken by a monitoring camera installed at an arbitrary monitoring place in a plant to detect the presence or absence of abnormality of the plant,
A / D conversion means for converting the captured image of the surveillance camera into a digital image, difference processing means for obtaining a difference image between the image captured via this means and a reference image in which no abnormality has occurred, and this means From the normal state, a plurality of differential images are subjected to non-linear luminance conversion and added together to create a gradation mask, and a mask storage memory for storing the gradation mask created by this means. And a mask processing means for removing a noise component from the difference image output from the difference processing means using the gradation mask stored in this memory, and whether or not an abnormality has occurred based on the difference image output from this means. A plant abnormality detection apparatus comprising: a determination unit for determining