JP7134918B2 - Image processing device and image processing method - Google Patents

Description

The present invention relates to an image processing apparatus and an image processing method for processing an image of a raceway photographed through tuyeres of a blast furnace.

The raceway in the blast furnace is photographed with a camera through the tuyeres. As a technique using the image of the raceway thus obtained, for example, there is a blast furnace abnormality detection method disclosed in Patent Document 1. This method is a blast furnace abnormality detection method for detecting an abnormality in which a tuyere portion of a blast furnace is in a blocked state, wherein the raceway portion is imaged through a furnace monitoring window provided in the tuyere portion, and the captured image is obtained. is equal to or less than a preset luminance threshold value and the luminance decrease rate is equal to or less than a preset luminance decrease rate threshold value, it is determined that an abnormality has occurred in which the tuyeres are closed.

WO2014/203509

The image of the raceway includes a coke burning area, a pulverized coal burning area, a PC lance area, and a background area. When no distinction is made between the coke combustion zone and the pulverized coal combustion zone, the combustion zone is described. The PC lance area is an area indicating a portion of a PC (Pulverized Coal) lance inserted into the blast furnace that is positioned within the raceway. The background area is an area indicating the inner peripheral surface of the blower pipe connected to the tuyere.

In order to monitor the temperature of the combustion zone within the raceway, the inventors have studied a technique for measuring the temperature of the combustion zone. Although the temperature of the combustion zone is said to be about 2300° C., it is difficult with current technology to directly measure the temperature of the combustion zone. Therefore, it is conceivable to estimate the temperature of the combustion region based on the luminance value of the combustion region by utilizing the positive correlation between the temperature of the combustion region and the luminance value.

It is assumed that the brightness values of the PC lance and background regions are smaller than the brightness values of the burn region. In the evaluation of the luminance value of the burn area, if the luminance values of the PC lance area and the background area are included, the luminance value of the burn area becomes smaller than the true value. As a result, the combustion temperature becomes lower than the true value.

SUMMARY OF THE INVENTION An object of the present invention is to provide an image processing apparatus and an image processing method capable of improving the accuracy of extracting a burned area from an image of a raceway.

The image processing device according to the first aspect of the present invention uses a plurality of images of the raceway obtained by a camera photographing the raceway through the tuyeres of a blast furnace at a plurality of times, and extracts pixels at the same position. a first generation unit for generating a representative value image representing a representative value of the first mask region corresponding to the background region and the PC lance region by binarizing the representative value image with a first threshold; a second generator for generating a first mask image having a first non-masked area corresponding to the burn area; and the raceway obtained by the camera photographing the raceway after photographing the plurality of images. and an extracting unit that extracts the burn region by masking the first mask image on the image of .

An image processing apparatus according to a first aspect of the present invention uses a first mask image generated based on a representative value image to extract a burnt region from an image of a raceway. The reason for not adopting the method of extracting the burnt region from the raceway image using the threshold is as follows. Among the combustion regions, the brightness value of the pulverized coal combustion region is higher than the brightness value of regions other than the combustion region (background region, PC lance region, etc.), but the difference is relatively small. Since the brightness value of the pulverized coal combustion region has a large amount of time variation, it may become about the same as the brightness value of the region other than the combustion region. Therefore, the method of extracting the combustion area from the raceway image using the threshold cannot extract the pulverized coal combustion area among the combustion areas.

Each pixel of the representative value image represents the representative value (mean, median, mode) of the pixels at the same position in these images generated using multiple images of the raceway. Therefore, even if the luminance value of the combustion region of pulverized coal fluctuates with time, the luminance value (representative value) of the combustion region of pulverized coal is always higher than the luminance value of the region other than the combustion region in the representative value image. It is possible to make it higher. A first mask image is generated using such a representative value image. Therefore, by masking the first mask image, it is possible to prevent areas other than the burned area from being extracted together when the burned area is extracted from the image of the raceway. Therefore, according to the image processing device according to the first aspect of the present invention, it is possible to improve the accuracy of extracting the combustion area included in the image of the raceway.

In the above configuration, the first generator generates the representative value image by moving along the time axis the range of the plurality of images from which the representative values of the pixels at the same position are selected.

Taking the average value as an example of the representative value, this configuration uses a moving average to generate the average value image. With this configuration, the first non-masked area and the first masked area can be changed over time. Therefore, even if the combustion area, the background area, and the PC lance area change over time, this can be dealt with.

The above configuration may further include a calculator that calculates a representative value of luminance of the combustion region extracted by the extractor.

The operator can evaluate the state of the combustion region based on the representative values (average, median, mode).

In the above configuration, the second generating section shrinks the first unmasked region.

Since temperature changes are rapid in the vicinity of the tuyere of the blast furnace, the refractive index of the space between the camera and the tuyere constantly changes. As a result, in a plurality of images (time-series images) of the raceway, the combustion area fluctuates vertically and horizontally. Therefore, the first unmasked area extends beyond the combustion area to include the PC lance area and the background area. By contracting the first non-masked region by the second generator, it is possible to prevent the first non-masked region from reaching the PC lance region and the background region. Therefore, according to this configuration, it is possible to further improve the extraction accuracy of the combustion region.

In the above configuration, a third generation unit for generating a variation value image representing the variation value of pixels at the same position using the plurality of images, and binarizing the variation value image with a second threshold value. Thus, a fourth generator for generating a second mask image having a second mask region corresponding to a region with a small temporal variation in luminance value and a second unmasked region corresponding to a region with a large temporal variation and a fifth generation unit that generates a third mask image having a third non-mask area indicating a logical sum of the first non-mask area and the second non-mask area, wherein the extraction unit comprises: For the image of the raceway, the third mask image is used as a mask instead of the first mask image to extract a raw ore drop region indicating the combustion region and the region where the unmelted ore falls.

Since the amount of time variation in the brightness value of the raw ore drop region is large, the raw ore drop region can be distinguished from the PC lance region and the background region according to the variation image (for example, the standard deviation image). Therefore, by using the second mask image generated based on the variation image, it is possible to extract the raw ore drop region. On the other hand, as described above, using the first mask image generated based on the representative value image enables extraction of the burn region.

With this configuration, a third mask image is generated having a third unmasked area that is a logical sum of the first unmasked area of the first masked image and the second unmasked area of the second masked image. The first unmasked area corresponds to the combustion area, and the second unmasked area corresponds to the raw ore drop area. Therefore, according to this configuration, it is possible to extract both the burned region and the raw ore fall region from the raceway image using the third mask image.

In the above configuration, the third generation unit generates the variation value image by moving along the time axis the plurality of images from which the variation value of the pixel at the same position is selected.

Taking the example of the variability value being the standard deviation, this configuration uses the moving standard deviation to generate the standard deviation image. According to this configuration, the second non-masked area and the second masked area can be changed over time. Therefore, even if an area with a large temporal variation in luminance value and an area with a small temporal variation in luminance value change with the lapse of time, this can be dealt with.

The image processing method according to the second aspect of the present invention uses a plurality of images of the raceway obtained by a camera photographing the raceway through the tuyeres of the blast furnace at a plurality of times, and extracts pixels at the same position. a first generating step of generating a representative value image showing a representative value of; a first mask region corresponding to the background region and the PC lance region by binarizing the representative value image with a first threshold; , a second generating step of generating a first masked image having a first unmasked area corresponding to a burned area; and said raceway obtained by said camera capturing said raceway after capturing said plurality of images. and an extraction step of masking the first mask image on the image of and extracting the burn region.

The image processing method according to the second aspect of the present invention defines the image processing apparatus according to the first aspect of the present invention from the viewpoint of the method, and has the same effects as the image processing apparatus according to the first aspect of the present invention. have

ADVANTAGE OF THE INVENTION According to this invention, the precision which extracts a combustion area from the image of a raceway can be improved.

It is a cross-sectional schematic diagram near the lower part of an example of a blast furnace. FIG. 4 is a schematic diagram showing an example of 20 consecutive frames included in a raceway moving image; It is a schematic diagram which shows an example of the image of a raceway. FIG. 4 is a diagram showing sample areas (1) to (4) respectively set in the coke combustion area, pulverized coal combustion area, PC lance area, and background area in the image shown in FIG. 4 is a graph showing the relationship between the average luminance value of each of sample areas (1) to (4) and time. 1 is a block diagram of an image processing device according to an embodiment; FIG. FIG. 5 is an explanatory diagram for explaining the relationship between a plurality of images of a raceway and an average value image; FIG. 4 is a schematic diagram showing an example of an average value image and a first mask image generated from this image; FIG. 4 is an explanatory diagram for explaining the relationship between a plurality of raceway images based on a first mask image and the raceway image masked by the first mask image; 3 is a schematic diagram showing a state in which a first mask image is superimposed on each of the 20 consecutive frames shown in FIG. 2; FIG. FIG. 4 is a schematic diagram showing an example of an image of a raceway taken with deposits formed on the inner peripheral surface of the blower pipe. FIG. 4 is a schematic diagram showing an example of an average value image in a state where deposits are formed on the inner circumferential surface of the blower tube, and a first mask image generated from this image; FIG. 10 is a schematic diagram showing an example of 20 continuous frames included in a raceway movie in which deposits are formed on the inner peripheral surface of the blower pipe; FIG. 14 is a schematic diagram showing a state in which a first mask image is superimposed on each of the 20 consecutive frames shown in FIG. 13; FIG. 4 is a schematic diagram showing an example of an image of a raceway taken with dirty tuyere viewing windows. FIG. 4 is a schematic diagram showing an example of a mean value image with a dirty tuyere viewing window and a first mask image generated from this image; FIG. 10 is a schematic diagram showing an example of 20 consecutive frames included in a raceway movie with dirty tuyere viewing windows. FIG. 18 is a schematic diagram showing a state in which a first mask image is superimposed on each of the 20 consecutive frames shown in FIG. 17; 4 is a flowchart for explaining the operation of the image processing apparatus according to the embodiment; FIG. 4 is a schematic diagram showing an example of an average value image, a first mask image generated from this image, and a first mask image that has undergone contraction processing. FIG. 3 is a schematic diagram showing a state in which a first mask image subjected to contraction processing is masked for each of the 20 consecutive frames shown in FIG. 2 ; FIG. 4 is a schematic diagram showing an example of 20 consecutive frames included in a raceway movie in which unmelted ore is falling. FIG. 11 is a block diagram of an image processing device according to a second modified example; FIG. 4 is an explanatory diagram for explaining the relationship between a plurality of images of a raceway and a standard deviation image; FIG. 4 is a schematic diagram showing an example of a mean value image, a first mask image generated from this image, a standard deviation image, a second mask image generated from this image, and a third mask image; FIG. 23 is a schematic diagram showing a state in which a third mask image is superimposed on each of the 20 consecutive frames shown in FIG. 22; FIG. 23 is a schematic diagram showing a state in which a first mask image is superimposed on each of the 20 consecutive frames shown in FIG. 22; A first mask image generated from the average value image with no unmelted ore falling, a second mask image generated from the standard deviation image with no unmelted ore falling, and a third FIG. 4 is a schematic diagram showing an example of a mask image; 3 is a schematic diagram showing a state in which a third mask image is superimposed on each of the 20 consecutive frames shown in FIG. 2; FIG. 10 is a flowchart for explaining the operation of the second modified example;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail based on the drawings. In each figure, the configurations denoted by the same reference numerals indicate the same configuration, and the description of the content of the configuration that has already been described will be omitted.

FIG. 1 is a schematic cross-sectional view of an example of a blast furnace 100 near its lower portion. A tuyere 101 is formed on the lower wall surface of the blast furnace 100 . High-temperature air 103 is sent into the blast furnace 100 at high speed from a blast pipe 102 arranged outside the blast furnace 100 via a tuyere 101 . Thereby, a space called a raceway 104 is formed from the tuyeres 101 toward the inside of the blast furnace 100 . Pulverized coal 106 is sent from a PC lance 105 through a tuyere 101 into a raceway 104 at high speed.

The blast pipe 102 is branched into two, one of which is connected to the source of the high-temperature air 103, and a tuyere viewing window 107 is attached to the other open end. A camera 108 captures a moving image V (time-series images) of the raceway 104 through the tuyere viewing window 107 .

FIG. 2 is a schematic diagram showing an example of 20 consecutive frames included in the moving image V of the raceway 104. As shown in FIG. The moving image V is in color, and one frame of the moving image V is one of the images Im1 of the raceway 104 . FIG. 3 is a schematic diagram showing an example of an image Im1 of the raceway 104. As shown in FIG. The image Im1 includes a coke combustion region R1-a, a pulverized coal combustion region R1-b, a PC lance region R2, and a background region R3. A boundary between the background region R3 and other regions indicates a contour R4 of the tuyere 101. FIG. When collectively referring to the coke combustion region R1-a and the pulverized coal combustion region R1-b, they are referred to as the combustion region R1.

The combustion region R1 (coke combustion region R1-a, pulverized coal combustion region R1-b) is shown in gray. PC lance region R2 and background region R3 are shown in black. 2, 4, 11, 13, 15, 17 and 22 are the same.

FIG. 4 shows sample regions (1) to ( 4). The sample region (1) is set to the coke combustion region R1-a. The sample region (2) is set to the pulverized coal combustion region R1-b. The sample area (3) is set in the PC lance area R2. A sample area (4) is set in the background area R3.

FIG. 5 is a graph showing the relationship between the average luminance value of each of sample areas (1) to (4) and time. The vertical axis indicates the average luminance value. The horizontal axis indicates the order of the 1st to 100th frames (image Im1 of raceway 104) as time.

The average brightness values are the sample area (1) of the coke combustion area R1-a, the sample area (2) of the pulverized coal combustion area R1-b, the sample area (3) of the PC lance area R2, and the sample of the background area R3. Higher in order of area (4). This indicates that the average luminance value is higher in the order of coke combustion region R1-a, pulverized coal combustion region R1-b, PC lance region R2, and background region R3.

The sample region (1) of the coke combustion region R1-a and the sample region (2) of the pulverized coal combustion region R1-b have a large amount of temporal variation in the average luminance value. This indicates that the coke combustion region R1-a and the pulverized coal combustion region R1-b have a large temporal variation in average brightness value.

The sample area (3) of the PC-lance area R2 has a slight temporal variation in the average luminance value. This indicates that the amount of temporal variation in the average luminance value of the PC-lance region R2 is slight.

The sample area (4) of the background area R3 has a substantially constant average luminance value. This indicates that the average luminance value of the background region R3 is substantially constant.

The image processing apparatus 1 according to the embodiment performs image processing on the moving image V of the raceway 104 captured by the camera 108 . FIG. 6 is a block diagram of the image processing device 1 according to the embodiment. The image processing apparatus 1 includes a control processing section 2 , an IF section 3 , an input section 4 and an output section 5 . The control processing unit 2 controls the entire image processing apparatus 1 and performs control and processing necessary for the operation of the image processing apparatus 1 . The control processing unit 2 includes hardware such as a CPU (Central Processing Unit), RAM (Random Access Memory), ROM (Read Only Memory), and HDD (Hard Disk Drive) to execute the functions of the control processing unit 2. It is realized by the program and data of

The IF section 3 is connected to the control processing section 2 and inputs/outputs images, information, etc. to/from external devices under the control of the control processing section 2 . For example, the IF unit 3 receives a moving image V of the raceway 104 captured by the camera 108 and sends it to the control processing unit 2 . The IF section 3 is implemented by an input/output interface circuit.

The input unit 4 is a device connected to the control processing unit 2 and used by an operator to input various information, data, commands, and the like. The input unit 4 is implemented by a mouse, keyboard, touch panel, or the like. The output unit 5 is a device that is connected to the control processing unit 2 and outputs the commands and data input from the input unit 4, the moving image V after image processing, etc., according to the control of the control processing unit 2. FIG. The output unit 5 is realized by a liquid crystal display, an organic EL display (Organic Light Emitting Diode display), or the like.

The control processing unit 2 includes a storage unit 21, a first generation unit 22, a second generation unit 23, an extraction unit 24, and a calculation unit 25 as functional blocks.

The storage unit 21 stores a moving image V of the raceway 104 captured by the camera 108 and the like.

The first generating unit 22 uses a plurality of images Im1 of the raceway 104 obtained by the camera 108 photographing the raceway 104 through the tuyere 101 of the blast furnace 100 at a plurality of times, and generates the pixels at the same position. An average value image Im2 representing the average value is generated. The first generation unit 22 moves the range of the plurality of images Im1 from which the average values are selected along the time axis (moving average) to generate the average value image Im2. The average value image Im2 is an example of a representative value image. A median value or a mode value may be used instead of the average value.

A plurality of images Im1 of the raceway 104 are images Im1 arranged in time series. In the case of the moving image V, a portion (a plurality of frames) of the moving image V becomes a plurality of images Im1 of the raceway 104 . The plurality of images Im1 of the raceway 104 may be a predetermined number of continuous frames, or a predetermined number of frames randomly selected from the continuous frames.

FIG. 7 is an explanatory diagram illustrating the relationship between a plurality of images Im1 of the raceway 104 and the mean value image Im2. Assume that the number of multiple images Im1 of the raceway 104 is N (N is an integer equal to or greater than 2). Assume that the number of pixels forming each image Im1 is M (M is an integer). In the 1st image Im1-1 to the Nth image Im1-N, the average value of the values of the 1st pixels becomes the value of the 1st pixel constituting the average value image Im2, and the value of the 2nd pixel. The average value becomes the value of the second pixel forming the average value image Im2, and the average value of the Mth pixel values becomes the value of the Mth pixel forming the average value image Im2.

The second generator 23 uses the average value image Im2 to generate the first mask image M1. FIG. 8 is a schematic diagram showing an example of the average value image Im2 and the first mask image M1 generated from this image. The second generator 23 converts the average value image Im2 into a grayscale image. Next, the second generation unit 23 binarizes the grayscale-converted average value image Im2 with the first threshold value to correspond to the low-luminance region including the background region R3 and the PC lance region R2. A first masked image M1 having a first masked region R11 and a first unmasked region R12 corresponding to the burn region R1 is generated. The first threshold value is second generated in advance by the operator (user) based on the respective average luminance values of the combustion region R1, the PC lance region R2, and the background region R3 included in the pre-generated average value image Im2. Set in part 23 .

The extracting unit 24 masks the image Im1 of the raceway 104 obtained by the camera 108 capturing images of the raceway 104 after the multiple images Im1 of the raceway 104 are captured, with the first mask image M1, thereby forming a combustion region. Extract R1. FIG. 9 is an explanatory diagram illustrating the relationship between a plurality of images Im1 of the raceway 104 that form the basis of the first mask image M1 and the image Im1 of the raceway 104 that is masked with the first mask image M1.

A plurality of images Im1 of the raceway 104 taken from time t1 to time t7 show seven consecutive frames. It is assumed that the number of multiple images Im1 is five. When the image Im1 of the raceway 104 captured by the camera 108 at time t6 is sent to the control processing unit 2, the first generation unit 22 generates five images Im1 of the raceway 104 captured most recently, that is, at time t6. An average value image Im2 is generated using five images Im1 of the raceway 104 photographed from t1 to time t5. The second generator 23 generates the first mask image M1 using this average value image Im2. Using this first mask image M1 as a mask, the extraction unit 24 extracts the burn region R1 from the image Im1 of the raceway 104 taken at time t6.

Next, when the image Im1 of the raceway 104 captured by the camera 108 at time t7 is sent to the control processing unit 2, the first generation unit 22 generates five images Im1 of the raceway 104 captured most recently, That is, the five images Im1 of the raceway 104 captured from time t2 to time t6 are used to generate the average value image Im2. The second generator 23 generates the first mask image M1 using this average value image Im2. Using this first mask image M1 as a mask, the extraction unit 24 extracts the burn region R1 from the image Im1 of the raceway 104 taken at time t7. FIG. 10 shows a schematic diagram showing a state in which the first mask image M1 masks each of the 20 consecutive frames shown in FIG. In FIG. 10, the masked region (first mask region R11) is shown in white. Regions that are not masked (first unmasked region R12) are shown in gray.

As described above, the image processing apparatus 1 extracts the combustion region R1 from each frame (each image Im1 of the raceway 104) that constitutes the moving image V of the raceway 104 captured by the camera 108 in real time.

Referring to FIG. 6, calculation unit 25 calculates the average luminance value of combustion region R1 extracted by extraction unit 24. FIG. The value is not limited to the average brightness value, and may be a representative value of the combustion region R1 (for example, median value or mode value.) The output unit 5 outputs the calculated average brightness value.

Even if deposits are formed on the inner peripheral surface of the blower pipe 102 shown in FIG. 1, according to the embodiment, the combustion region R1 can be extracted without being affected by this. This will be explained. FIG. 11 is a schematic diagram showing an example of an image Im1 of the raceway 104 captured with deposits formed on the inner peripheral surface of the blower pipe 102. As shown in FIG. As can be seen by comparing FIG. 11 and FIG. 3, the area of the background region R3 is large in the lower part of FIG. This is caused by deposits formed on the inner peripheral surface of the blower pipe 102 .

FIG. 12 is a schematic diagram showing an example of an average value image Im2 with deposits formed on the inner peripheral surface of the blower tube 102 and a first mask image M1 generated from this image. FIG. 13 is a schematic diagram showing an example of 20 continuous frames included in the moving image V of the raceway 104 in a state where deposits are formed on the inner peripheral surface of the blower pipe 102. As shown in FIG. FIG. 14 is a schematic diagram showing a state in which the first mask image M1 is superimposed on each of the 20 consecutive frames shown in FIG. In FIG. 14, the masked region (first mask region R11) is shown in white. Regions that are not masked (first unmasked region R12) are shown in gray. As can be seen by comparing FIG. 14 and FIG. 10, in each frame in FIG. 12). Therefore, it is possible to extract the combustion region R1 without being affected by deposits formed on the inner peripheral surface of the blower pipe 102 .

In the tuyere viewing window 107 shown in FIG. 1 , the surface opposite to the side on which the camera 108 is arranged becomes dirty due to soot or the like generated in the blast furnace 100 . Even if the tuyere viewing window 107 is dirty, according to the embodiment, the combustion region R1 can be extracted without being affected by this. This will be explained. FIG. 15 is a schematic diagram showing an example of an image Im1 of the raceway 104 captured with the tuyere peephole 107 dirty. As can be seen by comparing FIG. 15 and FIG. 3, in FIG. 15, the area of the combustion region R1 is small and the luminance value is low. This is because the tuyere viewing window 107 is dirty.

FIG. 16 is a schematic diagram showing an example of the mean value image Im2 with the tuyere viewing window 107 dirty and the first mask image M1 generated from this image. FIG. 17 is a schematic diagram showing an example of 20 continuous frames included in the moving image V of the raceway 104 with the tuyere viewing window 107 dirty. FIG. 18 is a schematic diagram showing a state in which the first mask image M1 is superimposed on each of the 20 consecutive frames shown in FIG. In FIG. 18, the masked region (first mask region R11) is shown in white. Regions that are not masked (first unmasked region R12) are shown in gray. As can be seen by comparing FIG. 18 and FIG. 10, the tuyere viewing window 107 is dirty, so in each frame of FIG. Although the area of is small, the first unmasked region R12 is secured. Therefore, the combustion region R1 can be extracted without the tuyere viewing window 107 being affected by dirt.

Even if deposits are not initially formed on the inner peripheral surface of the blast pipe 102 and the tuyere viewing window 107 is not dirty, deposits are formed on the inner peripheral surface of the blast pipe 102 with the passage of time. or the tuyere viewing window 107 may become dirty. According to the embodiment, since the average value image Im2 on which the first mask image M1 is based is generated using moving average, the first mask region R11 and the second non-mask region R22 can be changed over time. can be done. Therefore, according to the embodiment, even if such a situation occurs, the combustion region R1 can be stably extracted.

Operations of the image processing apparatus 1 according to the embodiment will be described. FIG. 19 is a flow chart explaining this operation. 6 and 19, control processing unit 2 causes storage unit 21 to store moving image V of raceway 104 sent from camera 108 . The operator uses the input unit 4 to input a command to the control processing unit 2 to output the average luminance value of the combustion region R1. The image processing device 1 thereafter extracts the combustion region R1 for each frame (image Im1 of each raceway 104) of the moving image V of the raceway 104 sent from the camera 108, and outputs the average luminance value of the combustion region R1. do. explain in detail.

The control processing unit 2 determines whether or not a frame (image Im1 of the raceway 104) has been input to the IF unit 3 (S1). When the control processing unit 2 determines that no frame has been input to the IF unit 3 (No in S1), the processing S1 is repeated.

When the control processing unit 2 determines that a frame has been input to the IF unit 3 (Yes in S1), the first generation unit 22 generates consecutive frames starting from the order n-1, which is one before the order n of this frame. An average value image Im2 is generated using the frames (a plurality of images Im1 of the raceway 104) of order n−1 to nk (S2).

The second generator 23 generates the first mask image M1 using the average value image Im2 generated in process S2 (S3). The extraction unit 24 uses the first mask image M1 generated in the process S3 as a mask to extract the burn region R1 from the n-th frame (S4).

The calculator 25 calculates the average luminance value of the combustion region R1 extracted in step S4 (S5). The output unit 5 outputs the average luminance value calculated in step S5 (S6).

The operator can use the input unit 4 to input a command to the control processing unit 2 to stop outputting the average luminance value of the combustion region R1. The control processing unit 2 determines whether or not an instruction to stop outputting the average luminance value of the combustion region R1 has been input (S7). When the control processing unit 2 determines that an instruction to stop outputting the average brightness value of the combustion region R1 has been input (Yes in S7), the image processing device 1 performs processing to output the average brightness value of the combustion region R1. finish. When the control processing unit 2 determines that an instruction to stop outputting the average luminance value of the combustion region R1 has not been input (No in S7), the process returns to S1. The above is the operation of the image processing apparatus 1 according to the embodiment.

With reference to FIG. 3, the present inventors studied a method of extracting the burn region R1 from the image Im1 of the raceway 104 using circle fitting. Here, a fitting region obtained by circular fitting the contour R4 of the tuyere 101 included in the image Im1 of the raceway 104 is defined as the burning region R1. It is conceivable to use the representative value of the luminance values of the fitting region as the luminance value of the burning region R1. Circle fitting can remove the background from the fitting region, but not the PC-lance region R2. Therefore, the luminance value of the burn region R1 becomes smaller than the true value under the influence of the luminance value of the PC lance region R2.

In order to avoid this, it is conceivable to set the maximum value or the third quartile of the luminance values in the fitting region as the luminance value of the burning region R1. In the former case, when the luminance value of one pixel in the fitting region is extremely large, the luminance value is set to the luminance value of the combustion region R1, which is not preferable as the luminance value of the combustion region R1. The latter is affected by the area of the PC lance 105 (Fig. 1) and the area of the pulverized coal 106 (Fig. 1) in the fitting area. .

Moreover, the method of extracting the combustion region R1 using circle fitting also causes the following problems. If deposits are formed on the inner peripheral surface of the blower pipe 102 connected to the tuyere 101, or if the tuyere viewing window 107 becomes dirty, the outline R4 of the tuyere 101 will not appear circular, but will appear distorted (Fig. 11, Fig. 15). Therefore, circle fitting cannot eliminate the background from the fitting region.

As described above, the method of extracting the combustion region R1 using circle fitting does not provide good extraction accuracy for the combustion region R1.

4 and 5, based on the fact that the brightness value of the combustion region R1 is greater than the brightness value of the regions other than the combustion region R1 (the PC lance region R2, the background region R3, etc.), the inventors studied a method of extracting the burn region R1 from the image Im1 of the raceway 104 using a threshold value. However, among the combustion regions R1, the brightness value of the pulverized coal combustion region R1-b is greater than the brightness values of the regions other than the combustion region R1 (PC lance region R2, background region R3, etc.), but the difference is relatively small. Since the brightness value of the pulverized coal combustion region R1-b has a large amount of time variation, it may become approximately the same as the brightness value of the regions other than the combustion region R1. Therefore, the method of extracting the combustion region R1 from the image Im1 of the raceway 104 using a threshold value cannot extract the pulverized coal combustion region R1-b from the combustion region R1.

As described above, in the method of extracting the combustion region R1 from the image Im1 of the raceway 104 using circle fitting and the method of extracting the combustion region R1 from the image Im1 of the raceway 104 using a threshold value, the combustion region R1 The extraction accuracy of is not good.

In contrast, according to the embodiment, it is possible to improve the extraction accuracy of the combustion region R1. explain in detail. Each pixel of the average value image Im2 (an example of the representative value image, FIGS. 7 and 8) is generated using a plurality of images Im1 (time-series images) of the raceway 104, and is located at the same position in these images Im1. Indicates the average value of a pixel. Therefore, even if the luminance value of the pulverized coal combustion region R1-b varies with time, the luminance value of the pulverized coal combustion region R1-b in the average value image Im2 is always the region other than the combustion region R1. It is possible to make the luminance value higher than that of (PC lance region R2, background region R3, etc.). The first mask image M1 is generated using the mean value image Im2. Therefore, when the burn region R1 is extracted from the image Im1 of the raceway 104 using the first mask image M1 as a mask, it is possible to prevent regions other than the burn region R1 from being extracted together.

In the combustion of the pulverized coal 106 (FIG. 1), the central portion is not burned or the combustion momentum is weak, and the peripheral portion has strong combustion momentum. Therefore, it is assumed that the central portion of the pulverized coal combustion region R1-b has a lower luminance value than the peripheral portion. Therefore, if the brightness value of the combustion region R1-b of pulverized coal is included in the evaluation of the brightness value of the combustion region R1, the brightness value of the combustion region R1 may not be evaluated correctly. According to the embodiment, for the reason described above, when extracting the combustion region R1 from the image Im1 of the raceway 104, it is possible to mask the central portion of the pulverized coal combustion region R1-b.

The embodiment has a first modified example and a second modified example. About these, it demonstrates centering on a different point from embodiment. The first modified example will be described. Referring to FIG. 6, the second generating unit 23 of the first modified example shrinks the first unmasked region R12. This will be described with reference to FIG. 20 . FIG. 20 is a schematic diagram showing an example of the mean value image Im2, the first mask image M1 generated from this image, and the first mask image M4 that has undergone contraction processing. The mean value image Im2 and the first mask image M1 are the same as the mean value image Im2 and the first mask image M1 shown in FIG. Since the first unmasked region R12 of the first mask image M1 has been eroded, the first unmasked region R12 of the eroded first masked image M4 is the first unmasked region of the first masked image M1. Smaller area than R12.

The extraction unit 24 extracts the burn region R1 from the image Im1 of the raceway 104 using the contracted first mask image M4 as a mask.

Referring to FIG. 1, the temperature changes rapidly near the tuyere 101 of the blast furnace 100, so the refractive index of the space between the camera 108 and the tuyere 101 constantly changes. As a result, the combustion region R1 captured in the moving image V fluctuates vertically and horizontally. Therefore, the first non-masked region R12 of the first masked image M1 extends beyond the burn region R1 to include the PC lance region R2 and the background region R3. Therefore, in the first modified example, the shrinking process of the first non-masked region R12 is performed a predetermined number of times so that the first non-masked region R12 does not reach the PC lance region R2 or the background region R3. The number of contraction processes is determined from this point of view.

The extraction unit 24 extracts the burn region R1 from the image Im1 of the raceway 104 using the contracted first mask image M4 as a mask. FIG. 21 shows a schematic diagram showing a masked state of the first mask image M4 subjected to the contraction process for each of the 20 consecutive frames shown in FIG. In FIG. 21, the masked region (first mask region R11) is shown in white. Regions that are not masked (first unmasked region R12) are shown in gray. According to this first mask image M4, it is possible to suppress the first non-mask region R12 (FIG. 20) from spreading over the PC lance region R2 and the background region R3, compared to the first mask image M1 shown in FIG. Therefore, it is possible to further improve the extraction accuracy of the combustion region R1.

A second modification will be described. In the second modification, in addition to the burning region R1, a raw ore drop region indicating a region where unmelted ore (unreduced ore) falls is extracted. Referring to FIG. 1 , in blast furnace 100 , unmelted ore (not shown) may fall onto raceway 104 . This may last a few seconds or may last for several minutes. The operator can immediately recognize when unmelted ore is falling onto the raceway 104, as unmelted ore falling onto the raceway 104 can damage or clog the air ducts 102. It is necessary to

FIG. 22 is a schematic diagram showing an example of 20 consecutive frames included in the moving image V of the raceway 104 in which unmelted ore is falling. As can be seen by comparing FIG. 22 and FIG. 2, when the unmelted ore falls on the raceway 104, the luminance value decreases in the entire image Im1. The brightness value of the raw ore drop region is lower than the brightness value of the burning region R1 and close to the brightness values of the PC lance region R2 and the background region R3. For this reason, it is not possible to extract the raw ore fall region from the first mask image M1 generated using the average value image Im2.

Since the raw ore drop region has a large temporal variation in brightness, the raw ore fall region can be distinguished from the PC lance region R2 and the background region R3 according to the standard deviation image Im3. A second modification uses a third mask image M3 generated based on the mean value image Im2 and the standard deviation image Im3 in order to extract both the burnt region R1 and the raw ore fall region.

FIG. 23 is a block diagram of an image processing device 1a according to the second modification. In addition to the storage unit 21, the first generation unit 22, the second generation unit 23, the extraction unit 24, and the calculation unit 25, the control processing unit 2 of the image processing device 1a includes the third generation unit 26, the fourth generation unit 27, and the A fifth generator 28 is provided.

The third generating unit 26 generates a standard deviation image Im3 representing the standard deviation of pixels at the same position using a plurality of images Im1 of the raceway 104 used to generate the average value image Im2. The third generating unit 26 generates a standard deviation image Im3 by moving the plurality of images Im1 from which standard deviations of pixels at the same position are selected along the time axis (moving standard deviation). The standard deviation image Im3 is an example of a variation image. Variance may be used instead of standard deviation.

FIG. 24 is an explanatory diagram illustrating the relationship between multiple images Im1 of the raceway 104 and the standard deviation image Im3. Assume that the number of multiple images Im1 of the raceway 104 is N (N is an integer equal to or greater than 2). Assume that the number of pixels forming each image Im1 is M (M is an integer). In the 1st image Im1-1 to the Nth image Im1-N, the standard deviation of the value of the 1st pixel is the value of the 1st pixel constituting the standard deviation image Im3, and the value of the 2nd pixel. The standard deviation becomes the value of the 2nd pixel forming the standard deviation image Im3, and the standard deviation of the value of the Mth pixel becomes the value of the Mth pixel forming the standard deviation image Im3. The third generation unit 26 generates the standard deviation image Im3 in parallel with the generation of the average value image Im2 shown in FIG. 7 by the first generation unit 22 .

The fourth generator 27 uses the standard deviation image Im3 to generate the second mask image M2. FIG. 25 schematically shows an example of a mean value image Im2, a first mask image M1 generated from this image, a standard deviation image Im3, a second mask image M2 generated from this image, and a third mask image M3. It is a diagram. 23 and 25, the fourth generator 27 converts the standard deviation image Im3 into a grayscale image. Next, the fourth generator 27 binarizes the grayscale-converted standard deviation image Im3 with the second threshold value to obtain a second mask region corresponding to a region with a small temporal variation in luminance value. A second mask image M2 having R21 and a second unmasked region R22 corresponding to this region with a large amount of time variation is generated. The second threshold value is determined by the operator (user) based on the average brightness values of each of the raw ore drop region, the burn region R1, the PC lance region R2, and the background region R3 included in the pre-generated standard deviation image Im3. is set in advance in the fourth generator 27 .

The first generation unit 22 generates the mean value image Im2 in parallel with the generation of the standard deviation image Im3 by the third generation unit 26 . The second generator 23 generates the first mask image M1 by binarizing the average value image Im2 with the first threshold value.

The fifth generator 28 generates the third mask image M3. The third mask image M3 includes a third non-mask region R32 representing the logical sum of the first non-mask region R12 included in the first mask image M1 and the second non-mask region R22 included in the second mask image M2. have. A region other than the third non-mask region R32 becomes the third mask region R31.

FIG. 26 is a schematic diagram showing a state in which the third mask image M3 is superimposed on each of the 20 consecutive frames shown in FIG. FIG. 27 is shown for comparison. FIG. 27 is a schematic diagram showing a state in which the first mask image M1 is superimposed on each of the 20 consecutive frames shown in FIG. In FIGS. 26 and 27, the masked region (third mask region R31) is shown in white. The unmasked regions (third unmasked regions R32) are shown in gray. It can be seen that if the third mask image M3 is used, the burned region R1 (light gray region) and the raw ore drop region R6 (dark gray region) can be extracted.

By using the second mask image M2 (FIG. 25), the raw ore drop region R6 can be extracted. In the case where the luminance value of the combustion region R1 varies greatly with time, the combustion region R1 can also be extracted from the second mask image M2. However, in a state in which the amount of time variation of the luminance value of the combustion region R1 is small, the combustion region R1 cannot be extracted from the second mask image M2. Therefore, in the second modified example, the third mask image M3 is used so that the combustion region R1 can be extracted even when the luminance value of the combustion region R1 has a small temporal variation.

A description will be given of the third mask image M3 created in a state in which the unmelted ore has not fallen. FIG. 28 shows a first mask image M1 generated from the average value image Im2 with no unmelted ore falling, and a second mask image M1 generated from the standard deviation image Im3 with no unmelted ore falling. FIG. 10 is a schematic diagram showing an example of a mask image M2 and a third mask image M3; FIG. 29 is a schematic diagram showing a state in which the third mask image M3 is superimposed on each of the 20 consecutive frames shown in FIG. In FIG. 29, the masked region (third mask region R31) is shown in white. The unmasked regions (third unmasked regions R32) are shown in gray. 29 and 10, the third mask image M3 can also extract the burn region R1 similar to the first mask image M1. Therefore, in the third mask image M3, when the unmelted ore has not fallen onto the raceway 104, it can be seen that the extraction accuracy of the combustion region R1 is high as in the first mask image M1.

The operation of the second modified example will be described. FIG. 30 is a flow chart explaining this operation. Referring to FIGS. 23 and 30, control processing unit 2 causes storage unit 21 to store moving image V of raceway 104 sent from camera 108 . The operator uses the input unit 4 to input a command to the control processing unit 2 to output the average brightness values of the combustion region R1 and the raw ore drop region R6. The image processing device 1a extracts the combustion area R1 and the raw ore drop area R6 for each frame of the moving image V of the raceway 104 sent from the camera 108, and outputs the average luminance value of these areas. explain in detail.

The control processing unit 2 determines whether or not a frame (image Im1 of the raceway 104) has been input to the IF unit 3 (S11). When the control processing unit 2 determines that no frame has been input to the IF unit 3 (No in S11), the processing S11 is repeated.

When the control processing unit 2 determines that a frame has been input to the IF unit 3 (Yes in S11), the first generation unit 22 and the third generation unit 26 each generate the frames one before the order n of this frame. A mean value image Im2 and a standard deviation image Im3 are generated using consecutive frames (a plurality of images Im1 of the raceway 104) of n−1 to nk starting from n−1 (S12).

23, 25 and 30, second generating unit 23 generates first mask image M1 using average value image Im2 generated in process S12, and fourth generating unit 27 processes A second mask image M2 is generated using the standard deviation image Im3 generated in S12 (S13). The fifth generation unit 28 performs a logical sum of the first non-mask region R12 of the first mask image M1 generated in process S13 and the second non-mask region R22 of the second mask image M2 generated in process S13. (S14).

The extraction unit 24 uses the third mask image M3 generated in the process S14 as a mask to extract the combustion region R1 and the raw ore drop region R6 from the n-th frame (S15). The raw ore drop region R6 is extracted when unmelted ore is falling on the raceway 104 .

The calculator 25 calculates the average brightness value of the area extracted in step S16 (S16). The output unit 5 outputs the average luminance value calculated in step S16 (S17). The operator refers to the average brightness value to determine the state of the combustion area R1 and whether unmelted ore has fallen onto the raceway 104 or not.

The operator can use the input unit 4 to input a command to the control processing unit 2 to stop outputting the average brightness values of the combustion region R1 and the raw ore drop region R6. The control processing unit 2 determines whether or not a command to stop outputting the average brightness values of the combustion region R1 and the raw ore drop region R6 has been input (S18). When the control processing unit 2 determines that an instruction to stop outputting the average brightness values of the combustion region R1 and the raw ore drop region R6 has been input (Yes in S18), the image processing device 1a controls the combustion region R1 and the raw ore The process of outputting the average luminance value of the falling region R6 is ended. When the control processing unit 2 determines that an instruction to stop outputting the average luminance values of the combustion region R1 and the raw ore drop region R6 has not been input (No in S18), the process returns to step S11.

1, 1a Image processing device 100 Blast furnace 101 Tuyere 102 Air duct 103 High temperature air 104 Raceway 105 PC lance 106 Pulverized coal 107 Tuyere viewing window 108 Camera Im1 image (raceway image)
Im2 average value image (an example of a representative value image)
Im3 standard deviation image (an example of variation image)
M1 First mask image M2 Second mask image M3 Third mask image M4 Contracted first mask image R1 Combustion region R1-a Coke combustion region R1-b Pulverized coal combustion region R2 PC lance region R3 Background region R4 Tuyere contour R6 Raw ore drop region R11 First mask region R12 First non-mask region R21 Second mask region R22 Second non-mask region R31 Third mask region R32 Third non-mask region V Animation

Claims (7)

First, a representative value image showing representative values of pixels at the same position is generated using a plurality of images of the raceway obtained by photographing the raceway through the tuyeres of the blast furnace at a plurality of times. a generator;
A first masked image having a first masked region corresponding to the background region and the PC lance region and a first unmasked region corresponding to the burn region by binarizing the representative value image with a first threshold value a second generator that generates
an extraction unit that masks the first mask image in the image of the raceway obtained by photographing the raceway with the camera after photographing the plurality of images, and extracts the combustion region. , image processor. 2. The image according to claim 1, wherein said first generation unit generates said representative value image by moving along a time axis a range of said plurality of images from which said representative values of pixels at said same positions are selected. processing equipment. 3. The image processing apparatus according to claim 1, further comprising a calculator that calculates a representative value of the brightness of the combustion region extracted by the extractor. The image processing apparatus according to any one of claims 1 to 3, wherein the second generator performs contraction processing on the first non-masked area. a third generation unit that generates a variation value image representing a variation value of pixels at the same position using the plurality of images;
By binarizing the variation value image with a second threshold value, a second mask region corresponding to a region where the amount of luminance value variation over time is small and a second non-mask region corresponding to a region where the amount of variation over time is large. a fourth generation unit that generates a second mask image having a mask area;
further comprising a fifth generation unit that generates a third mask image having a third non-mask region indicating a logical sum of the first non-mask region and the second non-mask region;
The extracting unit uses the third mask image instead of the first mask image as a mask for the image of the raceway to extract a raw ore drop region indicating the combustion region and the region where the unmelted ore falls. The image processing apparatus according to any one of claims 1 to 4, which extracts. 6. The image processing according to claim 5 , wherein said third generation unit generates said variation value image by moving said plurality of images from which said variation values of pixels at said same positions are selected along a time axis. Device. First, a representative value image showing representative values of pixels at the same position is generated using a plurality of images of the raceway obtained by photographing the raceway through the tuyeres of the blast furnace at a plurality of times. a generation step;
A first masked image having a first masked region corresponding to the background region and the PC lance region and a first unmasked region corresponding to the burn region by binarizing the representative value image with a first threshold value a second generating step that generates
an extracting step of masking the first mask image in the raceway image obtained by photographing the raceway with the camera after photographing the plurality of images, and extracting the combustion region. , image processing methods.

Images