JPH0822320A - Plant abnormality detecting device - Google Patents

Abstract

PURPOSE:To provide a device which excludes the influence of vibration and the noise of a device and localizes an abnormality occurrence place. CONSTITUTION:This device is equipped with a memory 6 for connective video storage which inputs video from a camera 8 and stores the connective video, a Fourier transform part 4 which inputs the connective video from the memory 6 and performs Fourier transform, and an image state decision part 5 which inputs a power spectrum image from the transform part 4 and decides whether or not there is plant abnormality and then misdetection due to the distortion of the video under the influence of vibration and noises of a device to be monitored and a photographing device can be excluded, so the abnormality detection precision is improved and if abnormality such as oil leak, steam leak, the trespass of a man, and smoke generation is caused, the position of a value appearing in the power spectrum image can be localized, so that the abnormality can be detected with high efficiency.

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 apparatus applied to automatically detect abnormality in a plant.

[0002]

2. Description of the Related Art In recent years, various devices have been proposed which can automatically detect abnormalities in a plant by image processing, without the need for an operator to visit the site or to watch a monitoring television.

To detect abnormalities such as oil leaks, steam leaks, human invasion and smoke generation, a method is used in which an image in which no abnormality has occurred is compared with an image in which abnormality has occurred and a different portion is detected. . The image in which no abnormality has occurred is stored in advance as a reference image, and is 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 by following changes in the state of the plant.

In the reference image and the image captured at the time of inspection, the characteristics of the image are emphasized through a filter or a mask in order to make it easy to detect different parts.
In order to reduce unnecessary noise components, several images are averaged.

An example of a conventional plant abnormality detecting device having such a function will be described with reference to FIG. The conventional device 01 shown in FIG. 8 inputs images output from a plurality of television cameras or infrared cameras 8 via a camera selector 9,
In addition, an operation signal is input from the operation panel 12 and an output signal is output to the series control computer 10 and the monitor 11, and the controller unit 02, the A / D conversion unit 03, the difference calculation unit 21, the image state determination unit 05, It is composed of a video memory 23 and a differential video memory 24.

In the above description, one of the images captured by the plurality of television cameras or infrared cameras 8 is selected by the camera selector 9 and sent to the plant abnormality detecting device 01.

The image sent from the camera selector 9 is sequentially converted into a digital signal by the A / D conversion unit 03 and stored in the image storage memory 23. The video stored in the video storage memory 23 is subtracted from the reference image by the difference calculator 21 to obtain a difference video storage memory 24.
Is stored.

For the video stored in the differential video storage memory 24, the image state determination unit 05 makes an abnormality determination based on the difference amount. The result of the presence / absence determination of the image state determination unit 05 is displayed on the monitor 11 and, if there is an abnormality, is output to the series control computer 10 to perform processing for the abnormality. The controller unit 02 is A /
The operations of the D conversion unit 03, the difference calculation unit 21, and the image state determination unit 05 are controlled and operated by the operation from the operation panel 12.

[0009]

In the conventional plant abnormality detecting apparatus, an image in a normal state where no abnormality as shown in FIG. 3 (a) is stored as described above is stored as a reference image. Every time, the difference processing is performed with the newly captured image at the same location, and it is determined from the difference whether or not there is an abnormality.

However, in the case of the conventional apparatus, a desirable image when an abnormality occurs is as shown in FIG. 9B, but the image is distorted due to the influence of the vibration or noise of the monitored device or the photographing device. As shown in FIG. 9A, the difference 31 due to the vibration appears as a difference in addition to the abnormal portion, and thus there is a high possibility that the abnormality is erroneously detected.

To remove this, a feature enhancement method using a filter or a mask and a noise component reduction method by averaging are used. However, due to the nature of the difference processing, it is difficult to eliminate the difference due to image distortion only by the difference processing. It was Therefore, in the case where the above-mentioned erroneous detection is made, it is structured to be eliminated in the post-processing.

Further, when an abnormality occurs, it is not possible to limit the occurrence location on the video, so the entire video is processed, and there is a problem that the processing efficiency of the video decreases.

In order to solve the above problems, the present invention is to realize an apparatus capable of reducing the influence of image distortion at the time of abnormality detection, reducing erroneous detection, and improving processing efficiency.

[0014]

A plant abnormality detecting apparatus of the present invention is a plant abnormality detecting apparatus for automatically detecting presence or absence of abnormality in a plant by inputting a video of an installation site from a camera installed in the plant, Connected video storage memory that inputs the video captured by the camera and stores the connected video, Fourier transform unit that inputs the connected video from the memory and performs Fourier transform, and power spectrum image input from the transform unit It is characterized by including an image state determination unit that determines the presence or absence of the image and outputs the determination result.

[0015]

In the above, the connected image storage memory stores a fixed amount of continuous images from the camera and stores the connected images. This connected image is input to the Fourier transform unit, which uses the Fourier transform to obtain a power spectrum image.

This power spectrum image has a characteristic that it has a maximum point with respect to a periodical change of the image, but in the connected image, the continuity of the image becomes periodic, and the periodicity corresponds to the number of the connected images. A power spectrum image in which such local maximum points appear can be obtained.

In the power spectrum image in the concatenated image in which the completely identical images are concatenated, complete periodicity can be obtained, so that no value occurs in the portions other than the periodic maximum points. Further, in the power spectrum image when completely different images are connected, the values are concentrated in the vicinity of the low frequency component because the periodicity is small.

When there is no abnormality in the plant, the concatenated images continuously captured and concatenated at the same location have extremely high periodicity, so that the power spectrum image has periodic maximum points.

In the case of a concatenated image containing an image in which a part of the image has changed due to an abnormality in the plant, the periodicity collapses, so that a value near the low frequency component other than the periodic maximum point is generated. Will appear. A plant abnormality can be detected by measuring the position and size of this value.

When there is no abnormality in the plant, but distortion of the image occurs due to the influence of vibration or noise of the monitored device or the photographing device, the connected image is different from the case where a part of the image changes. Differently, the periodicity of the image is not greatly deteriorated, and the values are concentrated near the maximum point. By using this principle, it is possible to eliminate the distortion of the image due to the influence of the vibration of the target device or the photographing device or the noise.

The location of the maximum value in the power spectrum image when there is no abnormality in the plant can be calculated in advance depending on the number of connected images and the connection method. Since the appearance position can be predicted, the image state determination unit that has input the power spectrum image output by the Fourier transform unit compares this with the predicted value, and thus it is possible to efficiently detect abnormality.

Further, even if the image distortion occurs due to the influence of the vibration or noise of the monitored device or the photographing device, the periodicity of the continuous image is high and the influence on the power spectrum image is small, so that the false detection is reduced.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A plant abnormality detecting apparatus according to an embodiment of the present invention will be described with reference to FIG. The device 1 of the present embodiment is
Similar to the conventional apparatus shown in FIG. 8, images output from a plurality of TV cameras or infrared cameras 8 are input through the camera selector 9, and operated by the operation panel 12, and the output signal thereof is controlled by the sequence control computer 10. And output to the monitor 11.

The device 1 of this embodiment shown in FIG.
Connected video storage memory 6 for inputting the digital signal of the video from the conversion unit 3, fast Fourier transform unit 4 for inputting the connected video from the memory 6, and power spectrum image storage for inputting the power spectrum image from the conversion unit 4. A memory 7, an image state determination unit 5 that inputs a power spectrum image from the memory 7 and outputs a determination result, and an A / D conversion unit 3 and a fast Fourier transform unit 4 that input an operation signal from the operation panel 12. The controller unit 2 that controls the image state determination unit 5 is provided.

FIG. 2 is a flow chart of abnormality detection performed by the apparatus of this embodiment, and the operation of the apparatus of this embodiment will be described below with reference to FIGS. 1 to 6. An image captured by the television camera or the infrared camera 8 and sent from the camera selector 9 as shown in FIG. 3A (an image of the steam pipe 13 provided with the joint portion 13a.
3b is its shaded portion) is successively converted into a digital signal by the A / D conversion portion 3, connected as shown in FIG. 4 and stored in the connected video storage memory 6.

This is because when the same image is arranged in the horizontal direction to form one image, an image having a strong periodic characteristic in the horizontal direction can be obtained. It should be noted that FIG. 4 shows horizontally read images for four frames at the same location that are continuously read.

The connected image stored in the connected image storage memory 6 is converted into a power spectrum image by the fast Fourier transform unit 4 which performs a Fourier transform for analyzing the frequency characteristic of the image, and the power spectrum image storage memory 7 is obtained. Memorized in.

FIG. 5 shows an example of a power spectrum image to be stored in the memory 7 for storing the power spectrum image. The power spectrum of the connected image obtained by horizontally connecting the images of four frames at the same position which are continuously read in the horizontal direction ( The darker the black, the higher the power spectrum value) is shown, and it is possible to know how many components have the same period. When four frames of images are connected in the horizontal direction, a vertical maximum appears in every 4 dots in the power spectrum image.

The image state determination unit 5 detects an abnormality such as oil leak, steam leak, human invasion or smoke generation in the plant on the basis of the power spectrum image stored in the power spectrum image storage memory 7, and detects the abnormality. It is determined whether there is.

In the power spectrum image in which completely the same video is connected, vertical line-shaped maximum values 15 of every 4 dots appear as shown in FIG. When image distortion occurs due to the influence of vibration or noise of the device or the photographing device, the maximum value 15 in the vertical line of every 4 dots and the vicinity of the maximum value 15 are low as shown in FIG. 6A. The value 16 appears.

As shown in FIG. 3B, a connected image including an image in which a part of the image has changed due to an abnormality in the plant such as oil leak, steam leak, human invasion or smoke generation. In this case, since the periodicity is lost, a low value 17 appears near the low frequency component as shown in FIG. 6 (b).

The location of the maximum value 15 in the power spectrum image when no abnormality occurs in the plant can be calculated and predicted in advance depending on the number of connected videos and the connection method. Therefore, it is possible to detect the abnormality of the plant by measuring the magnitude of the value other than the vertical maximum of every 4 dots near the low frequency component and comparing it with the preset value. .

The result of the presence / absence determination of the image state determination unit 5 is displayed on the monitor 11, and if there is an abnormality, it is input to the series control computer 10 and processing for the abnormality is performed.

In the present embodiment, the case where four video images are connected in one row in the horizontal direction has been described as the video connecting method, but the number of frames to be connected, the connecting direction, and the number of connected rows may be changed. it can. When the number of connected columns is increased, a power spectrum image having local maximum points 32 at equal intervals as shown in FIG. 7 is obtained.

[0035]

According to the plant abnormality detecting apparatus of the present invention, a connected image storage memory for inputting an image from a camera to store the connected image, a Fourier transform unit for inputting the connected image from the memory and performing a Fourier transform, and By providing the image state determination unit that inputs the power spectrum image from the conversion unit and determines whether there is a plant abnormality, it is possible to make a mistake when image distortion occurs due to the vibration or noise of the monitoring target device or the imaging device. Since detection can be eliminated, it is possible to improve the accuracy of anomaly detection, and also to determine the position of the value that appears on the power spectrum image when anomalies such as oil leaks, steam leaks, human intrusion, and smoke occur. Since it can be limited, it is possible to detect the occurrence of abnormality with high efficiency.

[Brief description of drawings]

FIG. 1 is an explanatory diagram 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 above embodiment.

3A and 3B are explanatory diagrams of an image captured by the camera according to the embodiment, FIG. 3A is an explanatory diagram of a normal image, and FIG.

FIG. 4 is an explanatory diagram of a combined image formed by a captured image according to the embodiment.

FIG. 5 is an explanatory diagram of a power spectrum image formed from a concatenated image when there is no change between captured images according to the embodiment.

6A and 6B are explanatory diagrams of a power spectrum image according to the above-described embodiment, where FIG. 6A is an explanatory diagram of a power spectrum image at the time of normal vibration and FIG.

FIG. 7 is an explanatory diagram of a power spectrum image when the number of connected columns is increased according to the above embodiment.

FIG. 8 is an explanatory diagram of a conventional abnormality detection device.

FIG. 9 is an explanatory diagram of an image related to the conventional device, (a)
3A is an explanatory diagram in the case where there is vibration, and FIG.

[Explanation of symbols]

 1 Plant Abnormality Detection Device 2 Controller Section 3 A / D Conversion Section 4 Fast Fourier Transform Section 5 Image State Judgment Section 6 Memory for Connected Video Storage 7 Memory for Power Spectrum Image Storage 8 Television Camera or Infrared Camera 9 Camera Selector 10 Series Control Computer 11 Monitor 12 Operation Panel 13 Steam Piping 14 Smoke 15 Vertical Line Maximum Value 16 Low Value Near Vertical Line Maximum Value 17 Low Value Near Low Frequency Components

Claims (1)

[Claims] 1. In a plant abnormality detection device for automatically detecting the presence or absence of abnormality in a plant by inputting the image of the installation site from a camera installed in the plant, the image captured by the camera is input and the connected image is stored. An image state in which a connected video storage memory, a Fourier transform unit that inputs a connected video from the memory and performs a Fourier transform, and a power spectrum image is input from the transform unit to determine whether there is a plant abnormality and output a determination result A plant abnormality detection device comprising a determination unit.