Monitoring device, monitoring system, and monitoring method

The monitoring device and system use image analysis and normalization processing to differentiate between normal flares and abnormalities, reducing false alarms by accurately identifying flare issues and distinguishing them from sunset or sunrise clouds.

JP7583112B2Active Publication Date: 2024-11-13ENEOS HLDG INC
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
JP2023093985
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2023-06-07
Publication Date
2024-11-13
Estimated Expiration
2039-03-26

AI Technical Summary

Technical Problem

Existing flare monitoring systems struggle with accurately distinguishing between normal flares and abnormalities, particularly when sunset or sunrise clouds are mistakenly detected as flares, leading to false alarms.

Method used

A monitoring device and system that includes image analysis to recognize abnormal flare regions, determines flare size and temporal changes, and employs normalization processing, shape symmetry, hue analysis, and temporal behavior to differentiate between flares and clouds, reducing false detections.

Benefits of technology

The system effectively reduces false alarms by accurately distinguishing between normal flares and abnormalities, enabling early detection of actual flare issues while minimizing confusion with sunset or sunrise clouds.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007583112000001
    Figure 0007583112000001
  • Figure 0007583112000002
    Figure 0007583112000002
  • Figure 0007583112000003
    Figure 0007583112000003
Patent Text Reader

Abstract

To provide a monitoring device which can precisely detect an abnormality in a flare discharged from a flare stack.SOLUTION: The monitoring device includes: a picture acquisition unit for acquiring a picture; a picture analysis unit for analyzing a picture and recognizing a predetermined abnormal flare region set on the basis of a burner part for discharging a flare as a reference; an abnormal flare determination unit for detecting an abnormal flare on the basis of the size of the flare in the abnormal flare region; and a false detection determination unit for determining the possibility of a false detection of an abnormal flare by analyzing a change with time of the abnormal flare on the basis of a plurality of pictures acquired in a predetermined length of time.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a monitoring device, a monitoring system, and a monitoring method. [Background technology]

[0002] At oil refineries, excess gas emitted during the refining process is incinerated in equipment called a flare stack. Flames (flares) and black smoke resulting from the combustion of excess gas are emitted from the tip of the burner section of the flare stack. To prevent accidents and to be environmentally friendly, flare stacks monitor the state of the flares and the generation of black smoke. Flares and other such phenomena are often monitored by staff who view footage of the area around the burner section filmed by security cameras.

[0003] For example, Patent Document 1 describes a black smoke detection system that detects black smoke emitted from flare stacks and chimneys in chemical plants, thermal power plants, and waste incineration facilities. The black smoke detection system described in Patent Document 1 uses images of emitted smoke captured continuously to calculate optical flow estimates for predetermined areas of two images. Then, a velocity field vector is calculated from the optical flow estimate, and an exhaust smoke area is extracted from the area where the velocity field exists. Furthermore, a brightness histogram is calculated for all pixels in the exhaust smoke area, and if there are a certain number or more pixels with dark brightness values, it is determined that black smoke exists in the exhaust smoke area. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2001-256475 A Summary of the Invention [Problem to be solved by the invention]

[0005] In the above-mentioned flare monitoring, it is required to detect flare abnormalities early and accurately. For example, when trying to detect a flare from an image of a chimney, there is a possibility that a sunset cloud captured near the chimney may be erroneously detected as a flare, but no technology has been provided to prevent such erroneous detection and detect abnormal flares accurately. In addition, Patent Document 1 does not disclose a method for detecting flare abnormalities.

[0006] SUMMARY OF THE PRESENT DISCLOSURE An object of the present invention is to provide a monitoring device, a monitoring system, and a monitoring method that can solve the above-mentioned problems. [Means for solving the problem]

[0007] According to one aspect of the present invention, a monitoring device includes an image acquisition unit that acquires an image, an image analysis unit that analyzes the image and recognizes a predetermined abnormal flare area that is set based on a burner unit that emits a flare, an abnormal flare determination unit that detects an abnormal flare based on a size of the flare that occupies the abnormal flare area, and an erroneous detection determination unit that determines the possibility of an erroneous detection of the abnormal flare by analyzing a temporal change in the abnormal flare based on a plurality of the images acquired during a predetermined period.

[0008] According to one aspect of the present invention, the erroneous detection determination unit determines the possibility of erroneous detection based on a change over time in the proportion of the flare in the multiple images.

[0009] According to one aspect of the present invention, the erroneous detection determination unit determines the possibility of erroneous detection based on a value after standardization processing, which is obtained by dividing a value indicating the fluctuation per unit time of the number of pixels corresponding to the flare in the multiple images by the number of pixels.

[0010] According to one aspect of the present invention, when the value after the normalization process is greater than a predetermined threshold, the erroneous detection determination section determines that the detection of the abnormal flare is not an erroneous detection.

[0011] According to one aspect of the present invention, when the value after the normalization process is equal to or smaller than a predetermined threshold, the erroneous detection determination unit determines that there is a possibility that the detection of the abnormal flare is an erroneous detection.

[0012] According to one aspect of the present invention, the erroneous detection determination unit compares the behavior of the flare contained in the multiple images with the behavior of clouds when the images contain clouds, and determines the possibility of erroneous detection.

[0013] According to one aspect of the present invention, the erroneous detection determination unit determines the possibility of erroneous detection based on a degree of symmetry of the shape of the flare included in the image in which an abnormal flare is detected.

[0014] According to one aspect of the present invention, the erroneous detection determination unit determines the possibility of erroneous detection based on the degree to which the hue of the flare contained in the image in which abnormal flare is detected deviates from a predetermined hue used as a reference.

[0015] According to one aspect of the present invention, the erroneous detection determination unit determines the possibility of erroneous detection based on a trained model constructed by learning the video when the abnormal flare was erroneously detected.

[0016] According to one aspect of the present invention, the image analysis unit calculates the number of pixels in the abnormal flare region where the flare is reflected, based on an RGB luminance value corresponding to the flare in the image and setting information indicating a range of the RGB luminance values, and the abnormal flare determination unit determines that a flare is abnormal when the number of pixels is equal to or greater than a predetermined threshold.

[0017] According to one aspect of the present invention, a surveillance system includes an imaging device that captures video of the periphery of a flare emission portion of a flare stack, and any of the surveillance devices described above that acquire the video.

[0018] According to one aspect of the present invention, a monitoring method includes the steps of acquiring an image, analyzing the image to recognize a predetermined abnormal flare region set based on a burner portion that emits a flare, detecting an abnormal flare based on a size of the flare occupying the abnormal flare region, and determining the possibility of erroneous detection of the abnormal flare by analyzing a temporal change in the abnormal flare based on a plurality of the images acquired during a predetermined period. Effect of the Invention

[0019] According to the present invention, it is possible to prevent erroneous detection of abnormal flare. [Brief description of the drawings]

[0020] [Figure 1] FIG. 1 is a diagram illustrating an example of a monitoring system according to an embodiment of the present invention. [Diagram 2] 1 is a functional block diagram illustrating an example of a monitoring device according to a first embodiment of the present invention. [Diagram 3] 3 is a diagram showing an example of an image analyzed by the monitoring device according to the first embodiment of the present invention; FIG. [Figure 4] FIG. 2 is a diagram showing an example of a monitoring screen of the monitoring system according to the first embodiment of the present invention. [Diagram 5] FIG. 4 is a first diagram illustrating the process of determining whether or not an erroneous detection has occurred based on a shape in the first embodiment of the present invention. [Figure 6] FIG. 11 is a second diagram illustrating the process of determining whether or not an erroneous detection has occurred based on a shape in the first embodiment of the present invention. [Figure 7] FIG. 4 is a first diagram illustrating the erroneous detection determination process based on hue in the first embodiment of the present invention. [Figure 8] FIG. 11 is a second diagram illustrating the erroneous detection determination process based on hue in the first embodiment of the present invention. [Figure 9] FIG. 11 is a third diagram illustrating the erroneous detection determination process based on hue in the first embodiment of the present invention. [Figure 10] 5A to 5C are diagrams illustrating a process of determining whether an error has been detected based on characteristics of change over time in the first embodiment of the present invention. [Figure 11] 5 is a flowchart showing an example of a false alarm determination process in the first embodiment of the present invention. [Figure 12] 5 is a flowchart showing an example of a monitoring process in the first embodiment of the present invention. [Figure 13] FIG. 11 is a functional block diagram illustrating an example of a monitoring device according to a second embodiment of the present invention. [Figure 14] FIG. 11 is a diagram illustrating a monitoring process according to a second embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] BEST MODE FOR CARRYING OUT THE PRESENT DISCLOSURE A flare stack monitoring system according to a first embodiment of the present invention will now be described with reference to Figs. (System Configuration) FIG. 1 is a diagram showing an example of a monitoring system according to an embodiment of the present invention. 1 shows the configuration of a surveillance system 1. As shown in FIG. 1, the surveillance system 1 includes a camera 10, a video acquisition device 15, a video distribution device 20, a surveillance monitor 25, and a surveillance device 30. Camera 10 captures images of the burner portion (the tip that dissipates the flare, the chimney) at the tip of the flare stack from which flares and smoke are emitted, and the surrounding area. FIG. 3 shows an example of an image captured by camera 10. The image captured by camera 10 will hereinafter be referred to as a flare image. Note that camera 10 is preferably an imaging device that captures video, but may also be one that captures still images at a predetermined time interval, for example.

[0022] Image capturing device 15 captures the flare image captured by camera 10 and outputs the flare image to image distribution device 20. Image capturing device 15 is generally connected directly to monitoring monitor 25, and a monitor looks at the flare image displayed on monitoring monitor 25 to check the area around and above the burner part of the flare stack, and visually monitors whether the flare is large, whether there has been a misfire, whether black smoke has been generated, etc. In monitoring system 1 of this embodiment, image capturing device 15 is connected to image distribution device 20.

[0023] The video distribution device 20 has a function of distributing video. The video distribution device 20 is connected to a monitoring monitor 25 and a monitoring device 30. The video distribution device 20 outputs the flare video acquired from the video acquisition device 15 to the monitoring monitor 25 and the monitoring device 30. The flare image is displayed on the monitoring monitor 25. The monitor monitors the flare image displayed on the monitoring monitor 25.

[0024] Monitoring device 30 analyzes the flare image and detects abnormalities in the size of the flare shown in the flare image, misfire of the flare, and generation of black smoke. When monitoring device 30 detects an abnormality in the flare, it displays the detection result on a display device connected to monitoring device 30 and notifies a monitor. Next, monitoring device 30 will be described in detail.

[0025] First Embodiment (Functions of monitoring devices) 2 is a functional block diagram showing an example of a monitoring device in the first embodiment of the present invention. The monitoring device 30 is a computer such as a PC (Personal Computer) or a server equipped with a calculation device such as a CPU (Central Processing Unit). As shown in the figure, the monitoring device 30 includes an image acquisition unit 31, a setting information acquisition unit 32, an image analysis unit 33, an abnormal flare determination unit 34, an erroneous detection determination unit 341, a misfire determination unit 35, a black smoke determination unit 36, a storage unit 37, and an output control unit 38.

[0026] The image acquiring unit 31 acquires the flare image. The image acquiring unit 31 outputs the flare image to the image analyzing unit 33. The setting information acquisition unit 32 analyzes the flare image and acquires setting information necessary for detecting flare abnormalities, etc. The setting information acquisition unit 32 records the acquired setting information in the storage unit 37.

[0027] The video analysis unit 33 analyzes the flare video. Here, FIG. 3 is referred to. FIG. 3 is a diagram showing an example of a video analyzed by the monitoring device in the first embodiment of the present invention. FIG. 3 shows an example of a flare video 50. Note that rectangular frames 51 to 54 are not the flare video 50 itself, but image information added to the flare video 50 by the output control unit 38. (1) For example, the video analysis unit 33 detects the burner part 56 shown in the flare video by pattern recognition based on the shape of the burner part. The video analysis unit 33 sets a predetermined position of the detected burner part 56 (for example, the center of the burner part 56) as a reference position in the flare video 50. For example, the video analysis unit 33 records in the storage unit 37 the coordinate information of the reference position when the lower left corner of the flare video 50 is set as the origin.

[0028] (2) Furthermore, the image analyzer 33 sets predetermined ranges based on the reference position in the flare image 50 as a normal flare region 51, an abnormal flare region 52, a black smoke detection region 53, and a background region 54. Normal flare region 51 defines a range that includes, for example, a part of burner portion 56 and in which the size of flare 55 can be considered normal. The abnormal flare region 52 is set to include, for example, a region farther from the reference position than the normal flare region 51. The abnormal flare region 52 is set to a position where a part of the flare 55 may reach when it expands. For example, the abnormal flare region 52 is set on the upper sky side outside the normal flare region 51, with a width wider than that of the normal flare region 51. When the flare 55 expands and a part of it reaches the abnormal flare region 52 and occupies a predetermined ratio, the monitoring device 30 determines that the flare 55 is large (abnormal). The abnormal flare region 52 may be set so as not to include the normal flare region 51 as shown in the figure, or may be set so as to include the normal flare region 51. In the following, unless otherwise specified, a case will be described as an example in which the abnormal flare region 52 is set so as not to include the normal flare region 51 as shown in the figure.

[0029] The black smoke detection region 53 is set to include a position farther away from the reference position than the abnormal flare region 52, for example. The black smoke detection region 53 is set to a position where black smoke may reach when black smoke occurs. For example, the black smoke detection region 53 is set on the sky side outside the abnormal flare region 52, with a width wider than the abnormal flare region 52. This is because black smoke may reach a wide range due to the influence of wind in the sky, and black smoke can be detected reliably. When the monitoring device 30 detects the presence of black smoke in the black smoke detection region 53, it determines that black smoke has occurred. Note that in order to detect the presence of black smoke, the video analysis unit 33 uses the brightness value of a pixel included in the background region 54 as a reference, and if the brightness value of a pixel present in the black smoke detection region 53 is a predetermined value lower than the reference brightness value (if it is darker), it determines that the pixel present in the black smoke detection region 53 indicates black smoke. Note that black smoke detection region 53 may be set so as not to include abnormal flare region 52 and normal flare region 51 as shown in the figure, or may be set so as to include abnormal flare region 52, or abnormal flare region 52 and normal flare region 51. In the following, unless otherwise specified, a case will be described as an example in which black smoke detection region 53 is set so as not to include abnormal flare region 52 and normal flare region 51 as shown in the figure.

[0030] The background region 54 is set for detecting black smoke as described above. The background region 54 is set in an area where flares and black smoke do not reach. For example, the background region 54 is set in a part of space that is sufficiently far from the burner section 56 and where no structures exist. The normal flare region 51, the abnormal flare region 52, the black smoke detection region 53, and the background region 54 can be set arbitrarily by the observer.

[0031] (3) Furthermore, video analysis unit 33 detects flare 55 from flare video 50 based on RGB luminance values ​​corresponding to the flare (256 gradation values ​​for each of red, green, and blue in the RGB color space). The RGB luminance values ​​corresponding to the flare and values ​​defining their ranges may be input by a monitor to monitoring device 30, and the setting information acquisition unit 32 may acquire the values. Alternatively, video analysis unit 33 may analyze pixels in which flare 55 is reflected among pixels around burner unit 56, and set the RGB luminance values ​​corresponding to the flare.

[0032] Abnormal flare determination unit 34 determines that flare 55 is abnormal (large) when the size of flare 55 occupying abnormal flare region 52 is equal to or greater than a predetermined ratio based on the analysis result of image analysis unit 33. More specifically, abnormal flare determination unit 34 determines that flare 55 is abnormal when the number of pixels having RGB luminance values ​​corresponding to the flare in abnormal flare region 52 is equal to or greater than a threshold value.

[0033] The erroneous detection determination unit 341 determines the possibility of erroneous detection of an abnormal flare based on the analysis result of the flare image in which the abnormal flare is detected. For example, the erroneous detection determination unit 341 determines the possibility of erroneous detection based on the shape (lateral symmetry) of the flare captured in the flare image. The erroneous detection determination unit 341 also determines the possibility of erroneous detection based on the degree of deviation between the hue of the flare captured in the flare image and a reference hue (45°). The erroneous detection determination unit 341 also determines the possibility of erroneous detection based on the characteristics of the change over time of the flare captured in the flare image. The processing by the erroneous detection determination unit 341 will be described later with reference to Figs. 5 to 11.

[0034] Misfire determination unit 35 determines that flare 55 has occurred as a misfire when the size of the flare in normal flare region 51 becomes equal to or smaller than a predetermined threshold based on the analysis result of image analysis unit 33. More specifically, misfire determination unit 35 determines that a misfire has occurred when the number of pixels having RGB luminance values ​​corresponding to the flare in normal flare region 51 becomes equal to or smaller than a threshold. Alternatively, misfire determination unit 35 determines that a misfire has occurred when the number of pixels having RGB luminance values ​​corresponding to the flare becomes equal to or smaller than a threshold continuously for a predetermined period of time based on the analysis result of image analysis unit 33 for flare image 50 captured by camera 10.

[0035] The black smoke determination unit 36 ​​determines that black smoke has occurred when the number of pixels having a luminance value corresponding to black smoke in the black smoke detection area 53 is equal to or greater than a threshold value based on the analysis result of the video analysis unit 33. As will be described later, 256 gradation luminance values ​​between black and white (black and white luminance values) are used to determine black smoke.

[0036] The storage unit 37 stores various information, such as coordinate information of the reference position of the burner unit 56, and setting information of RGB luminance values ​​and their ranges corresponding to flares.

[0037] The output control unit 38 generates an image of a user interface (for example, a monitoring screen 60 in FIG. 4) for displaying the results of determination by the abnormal flare determination unit 34, the erroneous detection determination unit 341, the misfire determination unit 35, and the black smoke determination unit 36, and for inputting setting information. The output control unit 38 outputs the user interface image to the display device of the monitoring device 30 and causes it to be displayed. An example of the monitoring screen is shown in FIG. 4.

[0038] (Example of monitoring device output) FIG. 4 is a diagram showing an example of a monitoring screen of the monitoring system in the first embodiment of the present invention. The output control unit 38 generates an image of a monitoring screen 60, an example of which is shown in Fig. 4. The monitoring screen 60 includes a result display field 61, flare luminance display fields 62a to 62b, setting information input fields 63a to 63f, a registration button 64, and a flare image 50. In the result display field 61, the judgment results by the abnormal flare judgment unit 34, the misfire judgment unit 35, and the black smoke judgment unit 36 ​​are displayed. When the abnormal flare judgment unit 34 does not detect a flare abnormality, the misfire judgment unit 35 does not detect a misfire, and the black smoke judgment unit 36 ​​does not detect black smoke, the output control unit 38 displays the text data "Normal" in the result display field 61. When the abnormal flare judgment unit 34 detects a flare abnormality, the output control unit 38 displays, for example, the text data "Flare abnormality". When the erroneous detection judgment unit 341 detects the possibility of erroneous detection of a flare abnormality, the output control unit 38 displays, for example, the text data "Flare abnormality (possibility of false alarm)". When the misfire judgment unit 35 detects a misfire, the output control unit 38 displays, for example, the text data "Misfire". When the black smoke judgment unit 36 ​​detects black smoke, the output control unit 38 displays, for example, the text data "Black smoke generated". Furthermore, when the abnormal flare determination unit 34 detects a flare abnormality and the black smoke determination unit 36 ​​detects black smoke, the output control unit 38 displays, for example, text data "Flare abnormality and black smoke occurrence".

[0039] The output control unit 38 displays, in the flare luminance display fields 62a-62b, RGB luminance values ​​(e.g., average values) of pixels in a portion of the flare image 50 that corresponds to the flare 55. The output control unit 38 displays, for example, red luminance values ​​in the flare luminance display field 62a and green luminance values ​​in the flare luminance display field 62b. The output control unit 38 displays an image in which the boundaries of a normal flare region 51, an abnormal flare region 52, a black smoke detection region 53, and a background region 54 are superimposed on a flare image 50 captured by the camera 10. The observer monitors the flare by looking at result display field 61, flare image 50, and flare brightness display fields 62a-62b, in addition to flare image 50 displayed on monitoring monitor 25. Note that flare image 50 displayed on monitoring monitor 25 does not display the boundaries indicating normal flare region 51, abnormal flare region 52, black smoke detection region 53, and background region 54.

[0040] The observer inputs a threshold value for abnormal flare determination unit 34 to determine that a flare is abnormal, that is, a threshold value for the ratio of the number of pixels that can be regarded as showing flare 55 to the total number of pixels in abnormal flare region 52, in setting information input field 63a. The observer inputs the threshold value at which the black smoke determination unit 36 ​​determines that black smoke has occurred, that is, the threshold value for the ratio of the number of pixels that can be considered to show black smoke to the total number of pixels in the black smoke detection area 53, into the setting information input field 63b.

[0041] The observer inputs threshold values ​​used by the video analysis unit 33 when detecting the flare 55 from the flare video 50 into the setting information input fields 63c to 63e. For example, the observer inputs a threshold value of the red (R) luminance value among the RGB luminance values ​​of the flare 55 into the setting information input field 63c, and inputs a threshold value of the green (G) luminance value among the RGB luminance values ​​of the flare 55 into the setting information input field 63d. The observer inputs a luminance width of the threshold values ​​input into the setting information input fields 63c and 63d into the setting information input field 63e. From these inputs, the video analysis unit 33 recognizes, as the flare 55, a pixel having a red luminance value whose allowable range is the value input into the setting information input field 63e centered on the value input into the setting information input field 63c, and a green luminance value whose allowable range is the value input into the setting information input field 63e centered on the value input into the setting information input field 63d.

[0042] The observer inputs into the setting information input field 63f a value that the video analysis unit 33 uses when detecting black smoke in the flare video 50. The video analysis unit 33 recognizes pixels (pixels after grayscaling) in the black smoke detection region 53 having a luminance value equal to or less than the value obtained by subtracting the value input into the setting information input field 63f from the luminance value after the image in the background region 54 is converted from an image in the RGB color space (RGB image) into a black and white image by grayscaling, as pixels reflecting black smoke.

[0043] When the observer presses the register button 64, the setting information acquisition unit 32 records the values ​​entered in the setting information input fields 63c to 63f in the storage unit 37. The video analysis unit 33 analyzes the flare video 50 based on each piece of setting information recorded in the storage unit 37.

[0044] Next, a process of confirming whether or not the detection is valid when abnormal flare determination unit 34 detects an abnormal flare in the analysis of flare image 50 will be described. (False positive determination) 1. False positive detection based on shape FIG. 5 is a first diagram illustrating the erroneous detection determination process based on shape in the first embodiment of the present invention. Image 5a is a flare image acquired by image acquisition unit 31. Image 5b is an image in which pixels having RGB luminance values ​​corresponding to the flare in image 5a are displayed in white and the rest are displayed in black. Images 5a and 5b are examples of abnormal flares. As can be seen from image 5b, the shape of the abnormal flare is approximately symmetrical with respect to burner portion 56. Most abnormal flares are observed to have symmetrical shapes as shown in FIG. 5.

[0045] FIG. 6 is a second diagram illustrating the erroneous detection determination process based on shape in the first embodiment of the present invention. Image 6a is a flare image acquired by image acquisition unit 31. Image 6b is an image in which pixels having RGB luminance values ​​corresponding to flare in image 6a are displayed in white, and the rest are displayed in black. Images 6a and 6b are examples of a case where a sunset cloud is erroneously detected as an abnormal flare. As can be seen from image 6b, the shape of a sunset cloud is often not symmetrical.

[0046] As illustrated in FIG. 6, if abnormal flare is detected based only on the number of pixels having RGB luminance values ​​in the abnormal flare region 52, it may be confused with sunset clouds or morning clouds photographed with similar RGB luminance values. Therefore, the erroneous detection determination unit 341 performs, for example, binarization processing on the flare image in which the abnormal flare is detected by the abnormal flare determination unit 34, extracts pixels having RGB luminance values ​​in a predetermined range corresponding to the flare, and determines the possibility of erroneous detection of the abnormal flare by focusing on the shape formed by the extracted pixels. More specifically, the erroneous detection determination unit 341 sets a vertical line to the ground based on an appropriate position of the burner part 56. The appropriate position of the burner part 56 is a position suitable for determining the left-right symmetry of the shape formed by the pixels determined to be an abnormal flare. For example, the erroneous detection determination unit 341 may determine the center of the burner part 56 as the appropriate position, or may determine the average value of the horizontal coordinate information of the leftmost and rightmost pixels among the pixels forming the flare as the appropriate position. The erroneous detection determination unit 341 divides the flare image into left and right of the vertical line to generate images A1 and A2, and then performs mirror image conversion on one of the images (for example, image A1) to generate image A1'. Next, the erroneous detection determination unit 341 calculates the similarity between image A1' and image A2 by a known method. If the similarity is equal to or greater than a predetermined threshold, the erroneous detection determination unit 341 determines that image A1 and image A2 are symmetrical, that is, the detection of the abnormal flare by the abnormal flare determination unit 34 is not an erroneous detection. If the similarity is less than the predetermined threshold, the erroneous detection determination unit 341 determines that image A1 and image A2 are not symmetrical, that is, sunset clouds or the like have been detected, and the detection of the abnormal flare may be an erroneous detection. Regarding the left-right similarity, the binarized image may be subjected to edge processing to extract the outline of the image, the integration of the vector distance from a predetermined position to the pixels of the outline may be calculated, and the ratio of the left-right vector integration values ​​may be used to determine that the images are symmetrical if the ratio is within a predetermined range. Alternatively, the horizontal pixel distribution of the detected flare image may be calculated, and the kurtosis or skewness of the distribution may be compared.

[0047] In this way, the erroneous detection determination unit 341 determines the possibility of erroneous detection of an abnormal flare based on the shape of the flare extracted from the flare image determined to be an abnormal flare. This method may result in an erroneous determination when the flare is blown horizontally by the influence of the wind, or when the shape of the range having the RGB luminance values ​​corresponding to the flare formed by sunset clouds is nearly symmetrical. Therefore, the erroneous detection determination unit 341 may further determine whether or not there is an erroneous detection by the method described below.

[0048] 2. False positive detection based on hue FIG. 7 is a first diagram illustrating the erroneous detection determination process based on hue in the first embodiment of the present invention. Image 7a is an image of an abnormal flare, and image 7b is an image in which morning clouds were erroneously detected as an abnormal flare. In this case, it is difficult to determine whether or not the detection was erroneous using the shape determination method described above. Therefore, the applicant analyzed the hue, saturation, and brightness of the pixels in these images that were determined to be flares. The results are shown in Figure 8.

[0049] FIG. 8 is a second diagram illustrating the erroneous detection determination process based on hue in the first embodiment of the present invention. The table in FIG. 8 shows the hue, saturation, and brightness of the image 7a in which a flare abnormality was correctly detected, and the hue, saturation, and brightness of the image 7b in which a flare abnormality was erroneously detected. As shown in the figure, it was found that there was no difference in saturation and brightness, but there was a difference in hue. In addition, it was found that it is easier to determine whether or not a false detection has occurred by focusing on the average value of all pixels rather than the maximum and minimum values ​​among all pixels having RGB luminance values ​​corresponding to a flare. The applicant also tried a method of determining whether or not a false detection has occurred by focusing on the frequency distribution of the hue in the monitoring area, but no significant difference was found between the abnormal flare and the morning glow clouds. Note that the maximum, minimum, and average values ​​of the hue in the table in FIG. 8 are values ​​based on a hue of 45°. In other words, the average value of the hue of the flare part of the image 7a in FIG. 7 is 39.9° (45-5.1), and the average value of the hue of the flare part of the image 7b is 45.4°. FIG. 9 shows the hue and saturation to be detected.

[0050] FIG. 9 is a third diagram illustrating the erroneous detection determination process based on hue in the first embodiment of the present invention. As shown in the figure, the saturation of the flare, sunrise clouds, and sunset clouds is around 25%. Here, a hue of 45° at a saturation of around 25% corresponds to a deep yellow, and the lower the degree of the hue, the closer it is to red. Also, at a saturation of around 25%, a hue of 45° is a whitish yellow. As shown in Figure 8, the sunrise clouds are close to a whitish yellow (hue 45.4°), while the flare is a yellow that is closer to orange (hue 39.9°).

[0051] Therefore, the erroneous detection determination unit 341 extracts all pixels having RGB luminance values ​​in a predetermined range corresponding to the flare from the flare image in which the erroneous flare is detected by the erroneous flare determination unit 34, and analyzes the hues of all the extracted pixels. Then, the erroneous detection determination unit 341 calculates a corrected hue by subtracting 45° from the analyzed hue, and calculates an average value of the corrected hues after subtraction. If the calculated average value is, for example, ±1°, the erroneous detection determination unit 341 may determine that the detected object is a morning glow cloud or a sunset cloud, and that the detection of the erroneous flare is an erroneous detection, and may determine other cases as not being an erroneous detection. Alternatively, the erroneous detection determination unit 341 may determine that the detected object is an erroneous flare and not an erroneous detection, and may determine other cases as not being an erroneous detection, if the calculated average value is smaller than a predetermined set value (for example, −4).

[0052] In addition to the judgment based on shape, the judgment based on hue can be performed to more accurately prevent erroneous detection of abnormal flares. Also, since the average hue value of all pixels detected as flares is used, it is possible to distinguish between clouds and flares based on whether the overall appearance is whitish (clouds) or deep yellow (flares). Note that by making the judgment based on the average hue value of all pixels, it is possible to make a judgment similar to when a person visually judges whether it is a cloud or a flare.

[0053] 3. False positive detection based on time-dependent change characteristics Furthermore, the erroneous detection determination section 341 may perform the erroneous detection determination based on a change over time in a pixel group detected as an abnormal flare. FIG. 10 is a diagram illustrating the process of determining whether or not there is an erroneous detection based on the characteristics of changes over time in the first embodiment of the present invention. FIG. 10(a) is a graph showing the change over time in the ratio of the number of pixels detected as flares to the number of pixels in the entire flare image when sunset clouds are detected as abnormal flares. FIG. 10(b) is a graph showing the change over time in the ratio of the number of pixels detected as flare to the number of pixels in the entire flare image when abnormal flare is correctly detected. As shown in Figures 10(a) and 10(b), the variation in the number of pixels of sunset clouds is small, while the variation in the number of pixels per unit time of flares is large. In other words, when observed at short time intervals, flares exhibit drastic changes in behavior, such as fluctuating in shape or disappearing for a moment, whereas clouds have the characteristic of maintaining a constant size and shape over time. The false detection determination unit 341 compares the behavior of flares contained in multiple images with the behavior of clouds when the images contain clouds, and determines the possibility of a false alarm.

[0054] Specifically, the erroneous detection determination unit 341 calculates the number of pixels having RGB luminance values ​​corresponding to a flare for a flare image at each predetermined time (for example, every few seconds to every few minutes) when an abnormal flare is detected by the abnormal flare determination unit 34. Next, the erroneous detection determination unit 341 calculates a change over time in the calculated number of pixels. If the change over time in the calculated number of pixels is greater than a predetermined range, the erroneous detection determination unit 341 determines that what was detected was a flare and not a false detection, and if the change over time in the number of pixels is within the predetermined range, the erroneous detection determination unit 341 determines that what was detected was a sunset cloud or the like and that there is a high possibility of a false detection. For example, if the change over time in the number of pixels is equal to or greater than a threshold value twice consecutively, the erroneous detection determination unit 341 determines that what was detected was a flare and not a false detection. In other words, if the number of pixels at time T1 is P1, the number of pixels at the next time T2 is P2, and the number of pixels at the next time T3 is P3, if the values ​​of |P1-P2| and |P2-P3| are equal to or greater than a threshold, the erroneous detection determination unit 341 determines that there is no erroneous detection. Note that the number of consecutive occurrences used as the criterion for the determination is not limited to two, and may be any number. Furthermore, it is not necessary for the change in the number of pixels over time to be continuously observed to exceed the predetermined range. For example, if the number of times that the change in the number of pixels over time exceeds the predetermined range in a predetermined period of time (e.g., two minutes) is equal to or exceeds a threshold value, the erroneous detection determination unit 341 may determine that the detection of the abnormal flare is not an erroneous detection. The calculation of the change over time may be performed as follows. First, the erroneous detection determination unit 341 performs a normalization process by calculating the standard deviation of the number of flare pixels in a predetermined time and dividing the calculated standard deviation by the average value of the number of flare pixels in the predetermined time. If the result of the division is equal to or greater than a predetermined threshold, the erroneous detection determination unit 341 may determine that the detection of the abnormal flare is not an erroneous detection, and if the result is equal to or less than the threshold, may determine that the detection of the abnormal flare is an erroneous detection. This normalization process is performed in order to accurately determine whether the object is a cloud or an abnormal flare by dividing the amount of change by the number of pixels of the detected object and judging the proportion of the amount of change to the entire object, since the proportion of the amount of change to the entire object is small when the detected object is a cloud. In the case of a cloud, the proportion of the amount of change to the entire object is very small, so the value after division is, for example, several percent or less. On the other hand, in the case of an abnormal flare, the proportion of the amount of change to the entire flare is large. In the normalization process, division is performed by the average value of the number of pixels, but division may be performed by using other representative values ​​such as the maximum value, minimum value, or median value of the number of pixels in a predetermined time.

[0055] In addition, erroneous detection determination section 341 may calculate the ratio of the number of pixels corresponding to flare to the total number of pixels for the flare image for each predetermined time period, and determine whether the detection of abnormal flare by abnormal flare determination section 34 is correct or erroneous, based on a change over time in the calculated ratio.

[0056] In this way, it is possible to distinguish between abnormal flares and sunset or sunrise clouds based on the characteristics of the changes in the flare over time shown by the results of analyzing the video for a specified period including the time when the abnormal flare was detected.

[0057] Next, an example of a false alarm determination process that combines the above three determination methods will be described. FIG. 11 is a flowchart showing an example of the false alarm determination process in the first embodiment of the present invention. It is assumed that the false detection determination unit 341 can be set to perform or not perform the false detection determination processes 1 to 3 described above. For example, in an environment where the wind is constantly blowing, the observer may set it so that "1. false detection determination based on shape" is not performed, or conversely, in an environment where there is no influence of wind and there is little influence of morning and evening clouds due to the installation position and direction of the camera 10 and surrounding buildings, the observer may set it so that only "1. false detection determination based on shape" is performed and other determination processes are not performed. Of course, the observer may set it so that only "2. false detection determination based on hue" is performed, or so that only "3. false detection determination based on characteristics of change over time" is performed. Alternatively, it can be set so that all of the false detection determination processes 1 to 3 are performed. These settings are set and registered in a predetermined setting file, and the false detection determination unit 341 performs the following false detection determination process by referring to the setting contents registered in this setting file. Note that the processing order of 1 to 3 shown below is for convenience, and the false detection determination processes 1 to 3 can be performed in any order.

[0058] When the abnormal flare determining section 34 detects an abnormal flare, the false detection determining section 341 starts a false alarm determining process. First, erroneous detection determination section 341 reads a setting file and acquires setting information regarding which of determination processes 1 to 3 is to be executed. Next, the erroneous detection determination unit 341 determines whether or not to perform a determination based on the shape (step S1). If a determination is to be performed (step S1; Yes), the erroneous detection determination unit 341 performs the process of "1. Determination of erroneous detection based on shape" described above (step S2). For example, the erroneous detection determination unit 341 extracts pixels indicating a flare from the flare image, and calculates the degree of left-right symmetry (similarity) based on a vertical line passing through the burner unit 56. If it can be considered symmetrical (similarity is equal to or greater than a threshold), the erroneous detection determination unit 341 determines that there is no erroneous detection. If it is not symmetrical (similarity is less than a threshold), the erroneous detection determination unit 341 determines that there is a high possibility of erroneous detection. The erroneous detection determination unit 341 records the erroneous detection method and the determination result in the storage unit 37 in association with each other. If the determination based on the shape is not to be performed (step S1; No), the process proceeds to step S3.

[0059] Next, the erroneous detection determination unit 341 determines whether or not to perform a determination based on the hue (step S3). If a determination based on the hue is to be performed (step S3; Yes), the erroneous detection determination unit 341 performs the process of "2. Determination of erroneous detection based on the hue" described above (step S4). For example, the erroneous detection determination unit 341 extracts pixels that indicate flare from the flare image and calculates the average value of the hue. For example, if the average value is near 45° (for example, ±1°), the erroneous detection determination unit 341 determines that there is a high possibility of erroneous detection, and if the average value is a value smaller than 45° by a predetermined value (for example, 4°) or more, the erroneous detection is not a erroneous detection. The erroneous detection determination unit 341 records the erroneous detection method and the determination result in the storage unit 37 in association with each other. On the other hand, if a determination based on the hue is not to be performed (step S3; No), the process proceeds to step S5.

[0060] Next, the erroneous detection determination unit 341 determines whether or not to perform a determination based on the characteristics of the time-dependent change (step S5). If the determination based on the characteristics of the time-dependent change is to be performed (step S5; Yes), the erroneous detection determination unit 341 performs the process of "3. Determination of erroneous detection based on the characteristics of the time-dependent change" described above (step S6). For example, the erroneous detection determination unit 341 acquires a flare image taken at a predetermined time after the detection of the abnormal flare, and monitors the transition of the number of pixels having RGB luminance values ​​corresponding to the flare. If the variation in the number of pixels exceeds a predetermined range, the erroneous detection determination unit 341 determines that it is not an erroneous detection, and if the variation in the number of pixels is not observed, there is a high possibility of an erroneous detection. The erroneous detection determination unit 341 records the method of erroneous detection and the determination result in the storage unit 37 in association with each other. On the other hand, if the determination based on the characteristics of the time-dependent change is not to be performed (step S5; No), the process proceeds to step S7.

[0061] Next, erroneous detection determination section 341 makes a final determination as to whether or not the detection of the abnormal flare by abnormal flare determination section 34 is a false alarm (step S7). False detection determination section 341 makes a final determination based on the determination results recorded in storage section 37 in steps S2, S4, and S6. For example, when erroneous detection determination has been performed using all of methods 1 to 3, and when it is determined that there is a possibility of erroneous detection in each of processes 1 to 3, erroneous detection determination section 341 determines that the detection of the abnormal flare is a "false alarm." On the other hand, when it is determined that there is no erroneous detection using all of methods 1 to 3, erroneous detection determination section 341 determines that the detection of the abnormal flare is "not a false alarm."

[0062] Also, for example, when erroneous detection is judged by only one of methods 1 to 3, erroneous detection judgment unit 341 judges it as "false alarm" if it judges that there is a possibility of erroneous detection by that method, and judges it as "not a false alarm" if it judges that there is no erroneous detection. Also, when one of the methods is "1. Judgment of erroneous detection based on shape" and it is judged that there is a possibility of erroneous detection, it may judge that there is no false alarm and add supplementary information saying "possibility of false alarm" to the judgment result.

[0063] In addition, for the judgment results obtained by methods 1 to 3, if it is judged that there is a possibility of a false detection based on "1. Judgment of false detection based on shape", a score of +1 point is given, if it is judged that there is a possibility of a false detection based on "2. Judgment of false detection based on hue", a score of +2 points is given, if it is judged that there is a possibility of a false detection based on "3. Judgment of false detection based on characteristics of change over time", a score of +3 points is given, and the false detection judgment unit 341 may judge it to be a "false alarm" if the total score is equal to or exceeds a predetermined number, and may judge it to be "not a false alarm" if not.

[0064] Furthermore, assuming that measures must be taken if an abnormal flare actually occurs and therefore that the determination of abnormal flare detection should be made with caution, for example, when it is determined that there is no erroneous detection using any one of methods 1 to 3, even if the remaining method determines there is a erroneous detection, erroneous detection determination section 341 may give weight to the detection of abnormal flare by abnormal flare determination section 34 and determine that there is “no erroneous detection,” while adding supplementary information that there is “possibility of erroneous detection” to the determination result.

[0065] 11, the execution or non-execution of each erroneous detection determination process is set, but the time period may be added as an execution condition. For example, steps S1 to S7 may be executed only during seasonal time periods when sunrise clouds or sunset clouds are observed (e.g., 5:00 to 6:00, 17:30 to 18:30).

[0066] (Monitoring process flow) Next, the flow of processing in the monitoring device 30 will be described. 12 is a flow chart showing an example of a monitoring process in one embodiment of the present invention. Camera 10 continuously captures images of the periphery of the burner part of the flare stack, and outputs the images to image acquisition device 15. In monitoring device 30, image acquisition unit 31 acquires flare images via image distribution device 20.

[0067] First, the observer sets the positional relationship among the normal flare region 51, the abnormal flare region 52, the black smoke detection region 53, and the background region 54 in the flare image 50 (step S11). For example, the observer inputs information (e.g., the length of each side of the normal flare region 51, the distance of each side from a predetermined position of the burner region, etc.) that defines the positional relationship between the burner part and the normal flare region 51 to the monitoring device 30. Alternatively, the observer may directly perform an operation to specify the range of the normal flare region 51 (e.g., specify a rectangular frame indicating the normal flare region 51 on the image using an input means such as a mouse) on the image of the flare image 50 shown in Figs. 3 and 4 that is displayed on the display of the monitoring device 30, thereby setting the positional relationship. The same applies to the other abnormal flare regions 52, the black smoke detection regions 53, and the background region 54. The setting information acquisition unit 32 acquires this information and records it in the storage unit 37.

[0068] Next, video analysis unit 33 sets each monitoring area (normal flare area 51, abnormal flare area 52, black smoke detection area 53, background area 54) for flare image 50 based on information defining the positional relationship between normal flare area 51, abnormal flare area 52, black smoke detection area 53, and background area 54 recorded in storage unit 37 (step S12). Video analysis unit 33 recognizes the position of burner unit 56 in flare image 50 by pattern recognition processing, and sets a predetermined position of burner unit 56 as a reference point. Specifically, video analysis unit 33 first calculates the positional relationship between the reference point and each monitoring area, and sets each area of ​​normal flare area 51, abnormal flare area 52, black smoke detection area 53, and background area 54. Then, the output control unit 38 generates an image of a monitoring screen 60 as exemplified in FIG. 4 in which an image in which the boundaries of a normal flare region 51, an abnormal flare region 52, a black smoke detection region 53, and a background region 54 are superimposed on a flare image 50 captured by the camera 10 and displayed at predetermined positions, and displays the image on the display.

[0069] Next, the observer inputs each setting information of the monitoring screen 60 illustrated in FIG. 4. The setting information acquisition unit records each input setting information in the storage unit 37. The image analysis unit 33 sets each setting information recorded in the storage unit 37 as a threshold value used for analyzing the flare image 50 (step S13). Next, the observer starts monitoring (step S14). First, the image analysis unit 33 recognizes the burner part of the flare image 50, and resets the reference point and each monitoring area (normal flare area 51, abnormal flare area 52, black smoke detection area 53, background area 54) as necessary (step S15). In this process, the shooting direction of the camera 10 may be changed by the observer's operation, and then adjusted again to the direction of the flare stack, or the shooting direction of the camera 10 may be shifted due to some external factor. In that case, each monitoring area is set with respect to the previous reference point, so there is a possibility that it is not in an appropriate positional relationship with the current reference point. The process of step S15 is performed for the purpose of correcting this and adjusting the range of the monitoring area in the flare image 50 captured by the camera 10.

[0070] Next, the image analysis unit 33 performs an analysis process on the flare image 50 (step S16). For example, the image analysis unit 33 detects pixels in the normal flare region 51 in which the flare 55 appears, based on the thresholds (e.g., red luminance threshold: 243, green luminance threshold: 152, luminance width: 30) input in the setting information input fields 63c to 63e. The image analysis unit 33 outputs the number of pixels in the normal flare region 51 in which the flare 55 appears (number of flare pixels (normal)) to the misfire determination unit 35. Furthermore, image analysis unit 33 detects pixels in which flare is reflected in abnormal flare region 52 based on the same threshold value. Image analysis unit 33 outputs the number of pixels in which flare is reflected in abnormal flare region 52 (number of flare pixels (abnormal)) to abnormal flare determination unit 34. 4 illustrates an example in which a red luminance value and a green luminance value are set, the image analysis unit 33 may detect pixels in which a flare is reflected in the normal flare region 51 or the abnormal flare region 52 by adding a blue luminance value to the detection. Also, the monitoring screen 60 may be provided with an input field for setting a blue luminance value.

[0071] Furthermore, the video analysis unit 33 converts the RGB image into a black-and-white image by grayscaling the multiple pixels in the background region 54, and calculates a representative value of the black-and-white luminance value in the converted black-and-white image (such as the average or median of the luminance values ​​of the multiple pixels). The video analysis unit 33 subtracts the value (e.g., 100) input in the setting information input field 63f from the calculated representative value of the black-and-white luminance value (256 gradations) to calculate a threshold value of the pixel value for determining black smoke. The video analysis unit 33 converts all pixels in the black smoke detection region 53 from an RGB image to a black-and-white image. The video analysis unit 33 compares the black-and-white luminance values ​​of all pixels in the black smoke detection region 53 in the converted black-and-white image with the threshold value for determining black smoke. The video analysis unit 33 calculates the number of pixels having a luminance value smaller than the threshold value for determining black smoke, and outputs this number of pixels (number of black smoke pixels) to the black smoke determination unit 36.

[0072] Furthermore, the video analysis unit 33 may detect the flare 55 as follows. That is, instead of using the threshold value or brightness width of the RGB brightness value set by the observer, the brightness value of the flare 55 is set based on the pixels of the normal flare region 51. For example, the video analysis unit 33 may consider a pixel that is in a predetermined positional relationship from the reference point (a pixel in a portion where the probability that a flare is reflected in a normal driving state is highest) to be a part of the image in which the flare 55 is reflected, calculate the RGB brightness value of the pixel, and set it as the threshold value of the flare 55. The video analysis unit 33 sets an arbitrary brightness width to the calculated RGB brightness value, and calculates the number of flare pixels (normal) in the normal flare region 51 and the number of flare pixels (abnormal) in the abnormal flare region 52.

[0073] Next, the abnormal flare determination unit 34 determines whether the number of flare pixels (abnormal) is equal to or greater than a threshold for abnormal flare determination (step S17). The threshold for abnormal flare determination is a value obtained by multiplying the total number of pixels included in the abnormal flare region 52 by the ratio input in the setting information input field 63a. For example, when the total number of pixels is 100 and 20(%) is input in the setting information input field 63a, the abnormal flare determination unit 34 determines whether the number of flare pixels (abnormal) acquired from the video analysis unit 33 is equal to or greater than 20. When the number is equal to or greater than the threshold for abnormal flare determination (step S17; Yes), the abnormal flare determination unit 34 outputs a determination result that the flare is abnormal (the flare 55 is large) to the erroneous detection determination unit 341. The erroneous detection determination unit 341 performs the erroneous alarm determination process described with reference to FIG. 11 (step S171). The erroneous detection determination unit 341 outputs a processing result of the erroneous alarm determination process to the abnormal flare determination unit 34. If the determination result by erroneous detection determination section 341 is a false alarm (step S172; Yes), abnormal flare determination section 34 determines that the flare is normal, and outputs this determination result to output control section 38. Then, the process proceeds to step S19.

[0074] If the determination result by erroneous detection determination section 341 is not a false alarm (step S172; No), abnormal flare determination section 34 determines that there is an abnormal flare (step S18). Abnormal flare determination section 34 outputs the determination result that an abnormal flare has occurred to output control section 38. Note that if the determination result by erroneous detection determination section 341 is not a false alarm but has added additional information indicating “possibility of false alarm”, abnormal flare determination section 34 outputs the additional information indicating “possibility of false alarm” to output control section 38 together with the determination result that an abnormal flare has occurred.

[0075] When the number of flare pixels (abnormal) included in the abnormal flare region 52 is less than the threshold for abnormal flare determination (step S17; No), the abnormal flare determination unit 34 determines that the flare is normal, and outputs this determination result to the output control unit 38. Then, the misfire determination unit 35 determines whether or not the number of flare pixels (normal) is equal to or less than the threshold for misfire determination (step S19). The threshold for misfire determination is, for example, 0. In this case, the misfire determination unit 35 determines whether or not there is a pixel having an RGB luminance value that can be regarded as a flare in the normal flare region. When the number of flare pixels (normal) is equal to or less than the threshold for misfire determination (step S19; Yes), the misfire determination unit 35 determines that the flare 55 has misfired (step S20). The misfire determination unit 35 outputs the determination result that the flare 55 has misfired to the output control unit 38. The threshold for misfire determination is recorded in advance in the storage unit 37.

[0076] In addition, the misfire judgment unit 35 may judge that a misfire has occurred when the number of flare pixels (normal) acquired from the image analysis unit 33 remains below the threshold value for misfire judgment for a predetermined period of time, or when the number of flare pixels (normal) acquired from the image analysis unit 33 is below the threshold value for misfire judgment for a predetermined number of consecutive times or more.

[0077] When the number of flare pixels (normal) included in the normal flare region 51 exceeds the threshold for misfire judgment (step S19; No), the misfire judgment unit 35 judges that misfire has not occurred and outputs this judgment result to the output control unit 38. Then, the black smoke judgment unit 36 ​​judges whether the number of black smoke pixels is equal to or greater than the threshold for black smoke generation judgment (step S21). The threshold for black smoke generation judgment is a value obtained by multiplying the total number of pixels included in the black smoke detection region 53 by the ratio input in the setting information input field 63b. For example, when the total number of pixels is 100 and 5 (%) is input in the setting information input field 63b, the black smoke judgment unit 36 ​​judges whether the number of black smoke pixels acquired from the video analysis unit 33 is 5 or more. When the number is equal to or greater than the threshold for black smoke generation judgment (step S21; Yes), the black smoke judgment unit 36 ​​judges that black smoke has occurred (step S22). The black smoke judgment unit 36 ​​outputs the judgment result that black smoke has occurred to the output control unit 38.

[0078] If the number of black smoke pixels included in the black smoke detection area 53 is less than the threshold for determining black smoke occurrence (step S21; No), the black smoke determination unit 36 ​​determines that black smoke is not occurring (step S23) and outputs this determination result to the output control unit 38.

[0079] Next, the output control unit 38 reflects the determination results based on each determination process on the monitoring screen 60. For example, if the determinations in steps S17, S19, and S21 are No, the output control unit 38 sets the character data "Normal" in the result display field 61 of the monitoring screen 60. If the determination is made in step S18 that the flare is abnormal, the output control unit 38 sets, for example, the character data "Flare Abnormal" in the result display field 61 of the monitoring screen 60. If the determination is made in step S171 that the flare is not a false alarm but that there is a possibility of a false alarm, the output control unit 38 sets, for example, the character data "Flare Abnormal (Possible False Alarm)" in the result display field 61 of the monitoring screen 60. If the determination is made in step S20 that the flare is a misfire, the output control unit 38 sets, for example, the character data "Misfire" in the result display field 61 of the monitoring screen 60. If the determination is made in step S22 that black smoke is generated, the output control unit 38 sets, for example, the character data "Black Smoke Generation" in the result display field 61 of the monitoring screen 60. In addition, if an abnormal flare is determined in step S18 and then black smoke is determined to have occurred in step S22, the output control unit 38 sets, for example, text data "Flare abnormality and black smoke occurrence" or "Flare abnormality (possibility of false alarm) and black smoke occurrence" in the result display field 61 on the monitoring screen 60.

[0080] Furthermore, the output control unit 38 displays the red luminance value of the pixel in which the flare 55 is reflected in the normal flare region 51 in the flare luminance display field 62a, and the green luminance value in the flare luminance display field 62b. For example, these values ​​are calculated by the video analysis unit 33 for the determination in step S19. The output control unit 38 generates an image of the monitoring screen 60 in which the above values ​​are set, and displays it on the display (step S24). If the monitoring of the flare 55 is to be continued (step S25; No), the processing from step S15 onwards is repeated, and if the monitoring of the flare 55 is to be ended (step S25; Yes), the series of processing is ended. Note that when it is determined that there is a "flare abnormality", "misfire", "black smoke generation", etc., the time and the flare image at that time may be recorded as a log in the storage unit 37.

[0081] When abnormal flare region 52 is set to include normal flare region 51, in the process of step S16, image analysis unit 33 may output a value obtained by subtracting the number of pixels in the region where flare is reflected in normal flare region 51 from the number of pixels in the region where flare is reflected in abnormal flare region 52 as the flare pixel number (abnormal) to abnormal flare determination unit 34. Alternatively, the threshold value for determining abnormal flare may be set to a large value. Furthermore, when the black smoke detection region 53 is set to include the abnormal flare region 52, the video analysis unit 33 may, in the process of step S16, output a value obtained by subtracting the number of pixels of the region in which black smoke is reflected in the abnormal flare region 52 from the number of pixels of the region in which black smoke is reflected in the black smoke detection region 53 as the black smoke pixel number to the black smoke determination unit 36. Similarly, when the black smoke detection region 53 is set to include the abnormal flare region 52 and the normal flare region 51, the video analysis unit 33 may, in the process of step S16, output a value obtained by subtracting the number of pixels of the region in which black smoke is reflected in the abnormal flare region 52 or the normal flare region 51 from the number of pixels of the region in which black smoke is reflected in the black smoke detection region 53 as the black smoke pixel number to the black smoke determination unit 36. Alternatively, the threshold value for black smoke determination may be set to a large value.

[0082] In general, in a flare stack, a flare image is captured and monitored. According to the monitoring system 1 of the present embodiment, by simply inputting the flare image into the monitoring device 30, a flare abnormality, misfire, and generation of black smoke can be detected in one system. As shown in FIG. 1, the monitoring system 1 displays the flare image on the monitoring monitor 25 and monitors the flare as in the past, while referring to the detection results such as "normal", "flare abnormal", "flare abnormal (possible false alarm)", "misfire", and "black smoke generation" notified by the monitoring device 30. In addition, the monitor can arbitrarily set the ranges of the normal flare region 51, the abnormal flare region 52, the black smoke detection region 53, and the background region 54, which are the criteria for judging a flare abnormality, etc., and the brightness value corresponding to the flare, according to the properties of the surplus gas and the installation environment of the flare stack, and therefore can be applied to any facility. In addition, if a flare is once determined to be abnormal, then determined to be a false alarm, and then eventually turns out to be normal, the output control unit 38 may set text data such as "Normal (flare-like cloud detected)" in the result display field 61 on the monitoring screen 60.

[0083] Furthermore, if sunset or sunrise clouds hang behind the flare stack, there is a possibility that these clouds will be erroneously detected as anomalous flares, but the monitoring system 1 of this embodiment can prevent confusion with sunset clouds and the like, and can accurately detect anomalous flares early. Alternatively, if there is a possibility of erroneous detection, a notice of this possibility can be sent to the monitor, so that the monitor can deal with the anomalous flare while recognizing the possibility of erroneous detection.

[0084] Second Embodiment A monitoring device 30a according to a second embodiment of the present invention will now be described with reference to FIGS. The monitoring device 30a according to the second embodiment has a function of learning images resulting from the determination of whether or not they are abnormal flares and whether or not they are falsely detected in the process of the first embodiment, and constructing a trained model that distinguishes whether the image is an image of a normal flare, an image of an abnormal flare, or an image that will be falsely detected as an abnormal flare. Also, the monitoring device 30a according to the second embodiment has a function of using the trained image thus constructed to determine which of the above three types of flare is included in the flare image acquired by the image acquisition unit 31. Hereinafter, among the configurations according to the second embodiment of the present invention, the same configurations and processes as those according to the first embodiment of the present invention are denoted by the same reference numerals, and the description thereof will be omitted.

[0085] FIG. 13 is a functional block diagram showing an example of a monitoring device in the second embodiment of the present invention. Monitoring device 30a according to the second embodiment includes abnormal flare determination unit 34a instead of abnormal flare determination unit 34. Abnormal flare determination unit 34a has a function of discriminating between normal flare and abnormal flare based on a trained model, in addition to the function of abnormal flare determination unit 34 of the first embodiment. Monitoring device 30a also includes a learning unit 39. These functions will be described below with reference to FIG. 14.

[0086] FIG. 14 is a diagram illustrating the monitoring process in the second embodiment of the present invention. First, the function of the learning unit 39 will be described with reference to the upper diagram of FIG. 14. The learning unit 39 accumulates learning data in the storage unit 37, in which the images judged as "normal", "flare abnormal", and "false alarm" are labeled with labels such as "normal", "abnormal", "false alarm", and "similar" by the process of the flowchart of FIG. 12. Note that these labels may be added after the monitor actually confirms "normal", "flare abnormal", and "false alarm" after the process of FIG. 12. Next, the monitor instructs the monitoring device 30a to perform learning. Then, the learning unit 39 reads out the learning data accumulated in the storage unit 37 and performs deep learning to build a trained model (for example, a multi-layered neural network) that distinguishes between images of "normal", "abnormal", and "false alarm".

[0087] Next, the function of the abnormal flare determination unit 34a will be described with reference to the lower diagram of FIG. 14. The abnormal flare determination unit 34a inputs a flare video of a monitoring target into a trained model and determines whether the flare video is "normal", "abnormal", or "false alarm". The abnormal flare determination unit 34a may perform an abnormal flare detection process based on the determination result by the trained model instead of the process of step S17 in FIG. 12. For example, if the determination result of the trained model is "normal" or "false alarm", the process of step S19 and subsequent steps is performed in the same manner as when "NO" is determined in step S17. If the determination result of the trained model is "abnormal", the process of step S18 and subsequent steps is performed. Alternatively, if the determination result of the trained model is "normal", the process may proceed to step S19 and subsequent steps, and if the determination result of the trained model is "abnormal" or "false alarm", the process of step S171 and subsequent steps may be performed. Alternatively, the determination using the trained model and each determination process described in the flowchart of Fig. 12 may be performed together. For example, the abnormal flare determination unit 34a performs determination using the trained model while performing the determination in step S17. Then, if the determination in step S17 is YES or if the trained model determines that the flare is "abnormal" or "false alarm", the processes from step S171 onward may be performed. Moreover, while performing such processing, the learning data may be further accumulated, and the learning unit 39 may reconstruct the trained model at a predetermined timing. This makes it possible to improve the accuracy of the trained model.

[0088] According to the present embodiment, it is possible to distinguish between an abnormal flare and an image similar to an abnormal flare using a trained model, thereby preventing erroneous detection of an abnormal flare and enabling detection of an abnormal flare with high accuracy. In addition, a trained model that can distinguish between misfires and the occurrence of black smoke may be constructed by learning images of misfires and the occurrence of black smoke.

[0089] Each process in the monitoring device 30 can be realized, for example, by a CPU or the like of the monitoring device 30 executing a program. The program executed by the monitoring device 30 may be recorded in a computer-readable recording medium, and may be realized by reading and executing the program recorded in the recording medium. The monitoring device 30 includes hardware such as an OS and peripheral devices. The computer-readable recording medium is, for example, a portable medium such as a flexible disk, an optical magnetic disk, a ROM, or a CD-ROM, and a storage device such as a hard disk built into the monitoring device 30. The computer-readable recording medium may also include a medium that dynamically holds a program for a short period of time, such as a communication line when a program is transmitted via a network such as the Internet or a communication line such as a telephone line, and a medium that holds a program for a certain period of time, such as a volatile memory inside a computer system that is a server or client in that case. The above program may be a program for realizing a part of the above-mentioned functions, or may be a program that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

[0090] In addition, the components in the above-described embodiment may be replaced with known components as appropriate without departing from the spirit of the present invention. Furthermore, the technical scope of the present invention is not limited to the above-described embodiment, and various modifications may be made without departing from the spirit of the present invention. Note that an RGB luminance value is a value expressed by a combination of red (R), green (G), and blue (B) luminance values ​​(0 to 255) in the RGB color space, and a black and white luminance value is a value expressed in 256 levels from 0 (black) to 255 (white). [Explanation of symbols]

[0091] 1. Surveillance system 10. Camera 15. Image acquisition device 20. Video distribution device 25. Surveillance monitor 30...Monitoring device 31...Image acquisition unit 32 Setting information acquisition unit 33. Video Analysis Section 34, 34a: Abnormal flare determination section 341, 341a: False detection determination unit 35... Misfire determination section 36 Black smoke detector 37...Storage section 38 Output control section 39. Learning Section 50 Flare video 51 Normal flare area 52. Anomalous flare region 53 Black smoke detection area 54...Background area 60...Monitoring screen 61...Result display field 62a, 62b Flare brightness display field 63a~63f Setting information input field 64. Register button

Claims

1. an image acquisition unit that acquires an image including a burner unit that emits a flare; A video analysis unit that analyzes the acquired video; an abnormal flare determination unit that detects an abnormal flare based on a size of the flare in a predetermined abnormal flare region that is determined in advance for the analyzed image; an erroneous detection determination unit that determines a possibility of erroneous detection of the detected abnormal flare by analyzing the image in which the abnormal flare is detected; Equipped with The erroneous detection determination unit performing a hue-based determination and another determination on the image in which the abnormal flare is detected; determining a possibility of erroneous detection of the abnormal flare by giving more importance to a result of the determination based on the hue than to the results of the other determinations; monitoring equipment.

2. A storage unit stores a first score according to a result of the judgment based on the hue, a second score according to a result of the other judgment, and a predetermined threshold value, the first score when it is determined that there is a possibility of erroneous detection in the determination based on the hue is set to a value larger than the second score when it is determined that there is a possibility of erroneous detection in the other determination, The erroneous detection determination unit determining that the detected abnormal flare is a false alarm when the sum of the first score and the second score is equal to or greater than the threshold; determining that the detected abnormal flare is not a false alarm when the sum of the first score and the second score is less than the threshold value; The monitoring device according to claim 1.

3. The erroneous detection determination unit, as one of the other determinations, determining a possibility of erroneous detection based on a degree of symmetry of the shape of the flare included in the image in which the abnormal flare is detected; The monitoring device according to claim 1 or 2.

4. The erroneous detection determination unit performs the determination based on the hue by calculating an average value of values ​​related to hues of pixels having RGB luminance values ​​within a predetermined range corresponding to the flare in the abnormal flare region included in the image in which the abnormal flare is detected; determining the possibility of a false positive based on the calculated average value; The monitoring device according to any one of claims 1 to 3.

5. The erroneous detection determination unit calculating a corrected hue by subtracting a predetermined reference hue from a hue of each pixel having an RGB luminance value within a predetermined range corresponding to the flare in the abnormal flare region included in the image in which the abnormal flare has been detected; Calculating an average value of the corrected hue of each of the pixels; determining a possibility of erroneous detection of the abnormal flare based on the calculated average value; 5. The monitoring device according to claim 4.

6. the erroneous detection determination unit determines the possibility of erroneous detection of the abnormal flare only during a pre-specified time period. A monitoring device according to any one of claims 1 to 5.

7. the erroneous detection determination unit, when it is determined that there is no erroneous detection in at least one of the determination based on the hue and the other determination, and when it is determined that there is an erroneous detection in the remaining method, adds supplementary information indicating that there is a possibility of an erroneous detection in addition to the determination result that there is no erroneous detection and that there is an erroneous detection in the remaining method; A monitoring device according to any one of claims 1 to 6.

8. an imaging device for capturing video of the flare emission portion of the flare stack and its surroundings; A monitoring device according to any one of claims 1 to 7 that acquires the video; A surveillance system comprising:

9. acquiring an image including a burner portion emitting a flare; Analysing the captured image; detecting an abnormal flare based on a size of the flare in a predetermined abnormal flare region set for the analyzed image; determining a possibility that the abnormal flare detected is an erroneous detection by analyzing the image in which the abnormal flare is detected; having In the step of determining the possibility of erroneous detection, performing a hue-based determination and another determination on the image in which the abnormal flare is detected; determining a possibility of erroneous detection of the abnormal flare by giving more importance to a result of the determination based on the hue than to the results of the other determinations; Monitoring method.

Citation Information

Patent Citations

  • Visible-smoke monitor

    JP1998232198A

  • Method and device for detecting fire by image processing

    JP2000268270A

  • System for detecting black smoke

    JP2001256475A

  • Flame / gas smoke detecting system, and flame / gas smoke detecting method

    JP2006268200A

  • Pilot flame monitoring method and its apparatus

    JP2008199173A