JPH06346130A - Detection of slag foaming by image processing - Google Patents

Abstract

PURPOSE:To quickly execute the detection of slag foaming and to stably execute desiliconizing treatment on a casting floor and molten iron pre-treatment without needing large reconstruction to equipment. CONSTITUTION:After digitizing a video signal obtd. by taking the picture of an iron receiving vessel opening part with an image pick-up means 1 using an image processing means 2, this digitized signal is stored in an image memory and also, by taking the difference between the present flame image and a time average image concerning the deveral images of few flame shapes before this flame image, the differential gradiation image is obtd., and further, only the peak brightness part is extracted from this differential gradiation image. Then, by detecting the peak part having >= fixed area and >= fixed distance from the opening part of the iron receiving vessel with a deciding means 3 as slag splash, the slag foaming can be detected automatically and accurately, and the operator is unnecessary to always observe the picture of the opening part of the iron receiving vessel.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a slag foaming detection method by image processing for detecting jetting of slag during hot metal casting bed desiliconization treatment or hot metal pretreatment.

[0002]

2. Description of the Related Art Generally, blast furnace hot metal is subjected to a treatment for removing specific components (Si, P, S, etc.) in the hot metal by adding a predetermined treating agent before the steelmaking process in a converter. . That is, the cast bed desiliconization treatment is performed by introducing a desiliconizing agent into the hot metal flowing through the inclined pouring gutter, thoroughly mixing the hot metal and the desiliconizing agent with a stirrer, and injecting the mixture into a hot metal wheel. For hot metal pretreatment such as dephosphorization and desulfurization, a lance is inserted into the reaction vessel, and a pretreatment material such as oxygen gas, nitrogen gas, solid oxide, or lime is introduced into the hot metal in the reaction vessel from the tip of the lance. Do by blowing.

At this time, the CO gas generated by the decarburization reaction may cause slag foaming in which the slag in the container is rapidly foamed, and the slag may flow out of the container. Further, when not only the slag but also the hot metal overflows, the yield decreases. Further, when the slag and the hot metal flow out from the hot metal car, there is a risk that the outflowing slag and the hot metal fill the rail, and the hot metal car cannot be pulled out.

Therefore, conventionally, an image from a TV camera for picking up a hot metal injection port of a hot metal car is displayed on a monitor, and an operator constantly monitors the image. When the operator determines that slag foaming has occurred, the foaming suppressant is introduced into the hot metal car to suppress the slag foaming and suppress the overflow of the slag and the hot metal.

[0005]

However, since the pigeon receiving place is a high temperature and is in a bad environment where dust, smoke, sparks, etc. are generated, the image from the TV camera is very difficult to see, and slag splashes, smoke, and It was difficult to distinguish it from flames. Therefore, detection of slag foaming requires judgment based on operator experience.

Further, as an example of a method for detecting slag foaming, there is a method described in JP-A-60-169762. In this method, first, the sound in the piggy-hole container is input with a microphone through a waveguide arranged above the inlet of the piggy-hole container. Then, from the input acoustic signal, a signal in a specific frequency range corresponding to the foaming sound generated when slag foaming occurs is extracted, and when this signal exceeds a certain level, it is determined that slag foaming has occurred, and an alarm is issued. .

However, in this detection method, the waveguide must be placed in the immediate vicinity of the piggy-hole container in order to detect the foaming sound, so that the waveguide is clogged with the slag or the like that is ejected from the pig-iron container. Will end up. Therefore, there is a problem in that the detection capability is deteriorated and the waveguide is melted due to the decrease in the sound pressure level.

The present invention has been made in order to solve the above-mentioned problems, and it is not necessary for the operator to constantly monitor the mouth portion of the receiving iron container, and the slag foaming can be detected promptly without requiring the operator's experience. Can be done
It is an object of the present invention to provide a slag foaming detection method that enables stable cast bed desiliconization treatment and hot metal pretreatment without requiring a large facility modification.

[0009]

SUMMARY OF THE INVENTION The gist of the present invention is to provide a slag foaming detection method for detecting slag foaming that occurs during pig iron reception in a pig iron container, wherein an image of the mouth of the pig iron container is imaged. Slag forming is detected by processing and detecting slag droplets.

Further, the image obtained by imaging the mouth portion of the piggyback container by the image pickup means is digitized as a two-dimensional gray image and then stored in the image memory, and the current frame image and the previous frame image are stored. A difference grayscale image is generated by taking the difference between the images of several frames and the time average image, and only the peak portion of the luminance is extracted from this difference grayscale image, and the area of the extracted peak portion has a constant value. The peak portion whose distance from the mouth portion of the pig iron container is a certain distance or more may be detected as slag droplets.

[0011]

The operation of the present invention will be described below with reference to the drawings. As shown in FIG. 1, the slag foaming detection method according to the present invention performs image processing on slag droplets blown out from the mouth of the piggy-receiving container based on the video signal obtained by the image pickup means 1 for picking up the mouth of the pig-making container. The image processing by means 2 is used for detection. Then, the judgment means 3 judges the occurrence of slag foaming based on the detected occurrence status of slag droplets, thereby detecting the slag foaming.

The reason why the image processing means 2 detects only the slag droplets to determine the occurrence of slag foaming is as follows. In other words, in the video signal obtained by imaging the mouth of the ladle container such as a ladle or a hot metal truck, monitor the video for what phenomenon is determined to cause slag foaming. The survey was conducted for operators who are performing slag foaming detection. As a result, it was found that most of the operators saw the slag droplets ejected from the injection port of the torch car and determined that slag foaming had occurred.

However, in addition to the slag splash, some operators judge by looking at the movement of smoke, the color of the flame, the amount of the flame, etc. Therefore, in order to detect the slag foaming, it is necessary to detect these as well. An experiment was conducted to determine whether there was any. This experiment is a field operation experiment in which only the slag droplets are observed and the slag foaming monitoring work is performed, and the experiment result is evaluated by the index of "pigment performance / pigment target". As a result, the detection rate of slag foaming was 90% in the conventional monitoring operation, but the detection rate was 9% when the monitoring operation was performed only by looking at the slag droplets.
An improved result of 2.5% was obtained.

Therefore, from this result, the movement of smoke, the color of flame, the amount of flame, etc. are only used as an auxiliary criterion by the operator, and only the slag droplets blown out from the mouth portion of the pig iron container should be detected. Therefore, it was found that slag foaming detection was possible. For the above reason, the determination of the occurrence of slag foaming is made by detecting the slag droplets.

In FIG. 1, the image pickup means 1 is directed to the mouth part of the piggy-hole container at a position apart from the mouth part of the pig-tail container so as not to be damaged by the slag splashed out from the pig-tail container. Will be placed. Then, the injection state of the hot metal is imaged and input as a video signal.

The image processing means 2 for detecting slag splashes
As shown in FIG. 2, an image input method 4 for digitizing a video signal acquired by the image pickup means 1 as a grayscale image, a preprocessing method 5 for performing preprocessing such as smoothing on the digitized grayscale image, An image storage unit 6 for storing a preprocessed grayscale image, and a difference processing method for obtaining a difference between an input grayscale image at a certain time and a grayscale image of a frame before that time stored in the image storage unit 6 7, a method 8 of extracting the peak portion of the brightness from the difference grayscale image obtained thereby and binarizing the peak portion, and spatially connected pixels are collected together and Labeling processing method 9 with labels such as 1,2,3 ... Method of calculation 0 Metropolitan consists of.

A method of detecting a slag droplet candidate by calculating the difference between a grayscale image at a certain time point and a grayscale image of a frame before that will be described with reference to FIGS. 2 and 3.

When the slag foaming does not occur, the slag droplets do not fly out from the pig iron container.
The image obtained by the difference processing method 7 of FIG.
It hardly changes like 11 and 12. However,
If slag splashes are coming out, 13, 1 in FIG.
As in Nos. 4 and 15, a portion considered to be slag splash is detected from the image obtained by subtracting the difference. In this way, the difference between the grayscale image at a certain point of time and the grayscale image before that is obtained by the difference processing method 7, whereby the slag droplet candidate can be detected.

However, in the difference grayscale image,
In addition to slag droplets, it is necessary to distinguish them from slag droplets because they contain flames, smoke, etc. at the receiving place. By the way, comparing the luminance distribution of "slag droplets" with the luminance distribution of "flame and smoke", there is a peak in the luminance distribution for slag droplets, and this peak is sharp, whereas Regarding smoke, it can be seen that the luminance distribution is gentle and no peak is generated.

Therefore, according to the method 8 of extracting the peak part of the luminance distribution and binarizing the peak part as a slag droplet candidate, first, the difference grayscale image obtained by the difference processing method 7 is subjected to the shrinkage process of the grayscale level. Is performed n times. Subsequently, after performing the density-level expansion process n times on the image subjected to the contraction process, the difference between the image subjected to these processes and the image not subjected to these processes is calculated.

Then, a binarization processing method in which a portion having a brightness larger than a preset threshold value is set to 1 and the other portion is set to 0 on the difference grayscale image The peak portion of the luminance of can be extracted as a slag droplet candidate. Then, the extracted slag droplet candidate is labeled by the labeling processing method 9, and the area of the slag droplet candidate and the distance from the mouthpiece of the pig iron container are calculated by the feature amount calculation method.

Next, the judging means 3 in FIG. 1 determines the area of the slag droplet candidate output from the image processing means 2 as described above and the distance from the mouth of the pig iron container as a characteristic quantity.
Determine whether slag foaming has occurred. That is, in order to distinguish the slag splash from the noise component, it is confirmed that the area of the slag splash candidate is larger than the predetermined threshold value.
Also, in order to distinguish the slag droplets from the hot metal that droops from the tilting gutter and attaches around the mouth of the piggy-back container, make sure that it is at least a certain distance from the mouth of the piggy-back container.

Further, in order to prevent the slag droplet candidates detected in the image of the previous frame from being newly detected due to the change in the brightness, the slag droplet candidates are located in the same place in the images of the previous frame within the predetermined number. Make sure it is not detected. Then, when these three conditions are satisfied, the slag droplet candidate is detected as a slag droplet, and it is determined that slag foaming has occurred.

[0024]

EXAMPLE Next, as an example of the present invention, an example of a method for detecting slag forming by applying the above-mentioned image processing to an image at the time of pouring hot metal into a mixed pig metal in a cast bed desiliconizing process Will be explained. In this embodiment, as shown in FIG.
A TV camera 17 is used as the image pickup means 1 in FIG. Then, the TV camera 17 is installed diagonally above the tow metal 18 so as to capture an image of the slag droplets 19 ejected from the hot metal injection port 20 of the tow metal.

In FIG. 2, the image input method 4 converts the video signal of the hot metal injection port obtained from the TV camera 17 installed as described above into a grayscale image of 256 gradations with 512 x 480 dots, The image is stored in an image storage area (not shown) in the image processing means 2. Next, by preprocessing method 5, the image is smoothed by applying a median filter to the grayscale image, and the preprocessed image is stored in the image storage means 6 as an input smoothed image.

Then, by the difference processing method 7, first, a time average image is obtained for the smoothed images of five frames already stored in the image storage means 6 as the image of the frame before the frame of the input smoothed image. . further,
The difference between the input smoothed image and the time average image is obtained, and the image obtained by this difference is used as the difference grayscale image.

Next, according to the peak detection / binarization processing method 6, the grayscale contraction processing is performed 6 times and the grayscale contraction processing is performed on the differential grayscale image 6 times.
The difference between the image that has been subjected to the grayscale expansion processing once and the image that has not been subjected to these processing is calculated. Then, a portion having a density of 12 or more is extracted as a luminance peak portion from the gray image thus obtained, and binarization processing is performed to divide the gray image into a peak portion and a portion other than the peak portion.

Next, a labeling process is performed by a labeling process method 9 so that 1,2,
Label with 3 ... Then, the feature amount calculation method 10 calculates the area and position of the peak portion (distance from the hot metal injection port) for this labeling image.

An electronic computer is used as the judging means 3 in FIG. Then, the feature amount of the area and position of the peak portion of the brightness of the grayscale image obtained by the image processing means 2 is input, and it is confirmed whether or not the feature amount satisfies a predetermined condition. That is, when the position of the center of gravity of the peak part is 20 pixels or more away from the mouth part of the hot metal and the area of the peak part is 50 pixels or more, the peak part is regarded as slag droplets,
Judge that slag foaming has occurred.

According to the above-mentioned method, when the slag foaming occurrence is judged by image-processing the surveillance image of the pig iron inlet of the torpedo car, 14 slag droplets can be detected out of 16 slag droplets. did it. In this way, it was possible to achieve detection accuracy that was in good agreement with the operator's judgment. Furthermore, by making the determination in consideration of the movement of the slag droplets and the fluctuation of the illumination, it becomes possible to perform the slag foaming detection with less erroneous detection and over-detection due to the disturbance such as flame and smoke.

[0031]

As described above, according to the present invention, it is possible to realize high detection accuracy comparable to the judgment by a skilled operator. Further, by considering the movement of the slag droplets and the fluctuation of the illumination depending on the state of the detection place, it is possible to perform the slag foaming detection in which the false detection and the over-detection due to the disturbance such as flame and smoke are greatly reduced. In addition, image processing can be used to detect slag foaming by detecting slag droplets, which was a method that was conventionally performed by the operator's eyes. Therefore, it is possible to save labor. Further, since slag foaming detection can be realized with high detection accuracy, it is possible to prevent slag and hot metal overflow satisfactorily by quickly adding the foaming inhibitor after detecting slag foaming, thus improving yield. And stable operation can be achieved.

[Brief description of drawings]

FIG. 1 is a block diagram showing an apparatus for realizing a slag foaming detection method of the present invention.

FIG. 2 is a diagram showing a flow of processing in an image processing means.

FIG. 3 is a diagram showing the principle of image processing for detecting slag droplets.

FIG. 4 is a diagram showing an arrangement example of image pickup means.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Imaging means 2 Image processing means 3 Judgment means 4 Image input method 5 Pre-processing method 6 Image storage means 7 Difference processing method 8 Peak detection / binarization processing method 9 Labeling processing method 10 Feature amount calculation method 11-16 Difference grayscale image 17 TV camera 18 Hot metal car 19 Slag splash 20 Hot metal injection port

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Hiroyuki Tanaka 1 Kimitsu, Kimitsu-shi Shin Nippon Steel Co., Ltd. Kim Tsu Works (72) Inventor Tomoaki Maruoka Kimitsu 1, Kimitsu Shin Nippon Steel Co., Ltd. Tsu Steel Works (72) Inventor Tatsuhiko Kondo 1 Kimitsu, Kimitsu City Shin Nippon Steel Co., Ltd. Kimitsu Works

Claims (2)

[Claims] 1. A slag foaming detection method for detecting slag foaming that occurs during pig iron reception in a pig iron container, wherein slag foaming is detected by image-processing a video image of the mouth of the pig iron container. A method for detecting slag foaming by image processing characterized by detecting. 2. A two-dimensional gray-scale image obtained by picking up an image of the mouth of the piggy-hole container by an image pickup means is digitized and then stored in an image memory, and a current frame image and a previous frame image are stored. A difference grayscale image is generated by taking the difference between the images of several frames and the time average image, and only the peak portion of the luminance is extracted from this difference grayscale image, and the area of the extracted peak portion has a constant value. The method for detecting slag foaming by image processing according to claim 1, wherein the peak portion whose distance from the mouth portion of the pig iron container is a certain distance or more is detected as slag droplets.