JPH08138058A - Analysis moving image data - Google Patents

Abstract

PURPOSE: To recognize the situation change of a detection object with high accuracy not by handling image data of time series as a still picture, respectively and effectively using the feature of a moving image, etc. CONSTITUTION: This method is constituted in such a way that binarization processing is applied to inputted moving image data (S1-S3), and sequential differential image data is found to two binarized moving image data (S4) adjacent to each other in time, and the degree of change of the image data is found based on the change of the differential image data (S5). Also, it can be recommended that a threshold value for binarization processing is updated based on the change with time of the binarized image data or differential image data. By employing such constitution, the situation change of the detection object can be recognized with high accuracy.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving image data analysis method and, more particularly, to a moving image data processing technique applicable to grasping the condition of sparks at a blast furnace tap.

[0002]

2. Description of the Related Art FIG. 16 shows a schematic processing flow chart in an example of a conventional moving image data analysis method. This conventional method is explained here by taking as an example the case where it is applied to ascertain the condition of sparks at the blast furnace tap. That is, conventionally, color image data is first input (S1), and the image is expressed only by the brightness by performing grayscale conversion (S2). An example thereof is shown in FIG. Then, paying attention to the brightness of each pixel of the gray scale image, a binarization process is performed in which if it is above a certain threshold value, it is converted to white, and if it is below the threshold value, it is converted to black (S3). An example thereof is shown in FIG. next,
The number of white luminance pixels is calculated for the binarized image. The situation is grasped by the change in the number of pixels over time (S5 '). The time series data of the gray scale image obtained at this time are shown in FIG. 4, and the time series data of the binarized image are shown in FIG. 5, respectively. Here, 0, 100, 200, 300, 4 in each figure
00, 500 frames are 0, 3.3, 6.6, 1 respectively
This corresponds to the data after 0.0, 13.3, and 16.6 seconds.
Calculations of the number of pixels of white luminance for the time-series data of the binarized image shown in FIG. 5 are shown in FIGS. 17 (a) and 17 (b). In the past, the state of sparks at the blast furnace taphole was grasped based on this calculation result. As for the threshold value for obtaining the binarized image, a method for still image data is usually used. That is, the brightness distribution of the entire image is calculated and, for example, a value at which the white brightness distribution and the black brightness distribution are equal is determined as the threshold value according to the distribution status. However, there was no such established method for moving image data.

[0003]

In the conventional moving image data analysis method as described above, even if the image changes greatly between 200 and 300 frames in the time series data of the grayscale image of FIG. Regardless, there is almost no change in the number of pixels in the white part (see FIGS. 17A and 17B). This makes it impossible to detect a drastic change in the situation at the tap of the blast furnace and a sudden increase in the amount of sparks. Also, with the above-mentioned method in which the threshold value is also used for still image data, each moving image data has its own threshold value for each grayscale image, which is common to all grayscale images. It is not clear which threshold should be used. On the other hand, if the threshold value is changed for each grayscale image, the reference level becomes variable, which is a problem.
Therefore, a threshold value determination method for binarizing moving image data has become necessary. The present invention has been made in view of the above circumstances. A first object of the present invention is not to treat time-series image data as still images as in the conventional example, but to analyze the characteristics of moving images. The objective is to provide a moving image data analysis method that can accurately detect changes in the condition of a detection target by making the most of it.
A second object is to provide a moving image data analysis method capable of appropriately determining a threshold value used for binarization processing of moving image data.

[0004]

In order to achieve the first object, the first aspect of the present invention performs binarization processing on input image data, and outputs two binary values temporally adjacent to each other. It is configured as an image data analysis method for sequentially obtaining difference image data for the converted image data and obtaining the degree of change of the image data based on the change of the difference image data. In addition, in order to achieve the second object, the second invention performs binarization processing on the input moving image data with a preset initial threshold value. The threshold value is updated until the temporal change of the binarized moving image data becomes a predetermined value or more, and the threshold value is updated based on the change of the binarized moving image data by the updated threshold value. It is configured as a moving image data analysis method for obtaining the degree of change of input moving image data. In addition, in order to achieve the first and second objects, the third invention performs binarization processing on input moving image data with a preset initial threshold value and temporally adjoins them. The difference image data is sequentially obtained for the two binarized moving image data, and the threshold value is updated until the temporal change of the difference image data becomes a predetermined value or more, and the updated threshold value is updated. It is configured as a moving image data analysis method for obtaining the degree of change of the input moving image data based on the change of the difference image data obtained by using. Furthermore, the above moving image data is
This is a moving image data analysis method in which color image data is converted to gray scale. Furthermore, it is a moving image data analysis method in which the color image data is time-series image data of the blast furnace taphole, and the degree of change is the appearance of sparks.

[0005]

According to the first aspect of the present invention, the input moving image data is binarized, and differential image data is sequentially obtained from two temporally adjacent binarized moving image data. Then, the degree of change of the input moving image data is obtained based on the change of the difference image data. As described above, the time-series image data is not treated as a still image as in the conventional example, but by utilizing the characteristics of the moving image, it is possible to accurately grasp the situation change of the detection target. According to the second aspect of the invention, the input moving image data is binarized by an initial threshold value set in advance, and the temporal change of the binarized moving image data is predetermined. The above threshold is updated until the value is exceeded,
The degree of change of the input moving image data is obtained based on the change of the moving image data binarized by the updated threshold value. In this case, since the change in the binarized moving image data can be regarded as a function of the threshold value, an appropriate threshold value for the moving image data should be obtained by using an optimization method (for example, the golden section method). You can
Therefore, the situation change of the detection target can be captured more accurately than in the case of using the threshold for still image data as in the conventional example. According to the third invention, binarization processing is performed on the input moving image data with an initial threshold value set in advance, so that two pieces of binary moving image data temporally adjacent to each other are binarized. On the other hand, the difference image data is sequentially obtained, the threshold value is updated until the temporal change of the difference image data becomes a predetermined value or more, and the change of the difference image data obtained by using the updated threshold value. Based on the above, the degree of change of the input moving image data is obtained. In this case, due to the synergistic action of the first and second inventions, it is possible to more accurately capture the situation change of the detection target.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments embodying the present invention will be described below with reference to the accompanying drawings for the understanding of the present invention. The following embodiments are examples of embodying the present invention and are not intended to limit the technical scope of the present invention. Here, FIG. 1 is a schematic process flow chart of a moving image data analysis method according to an embodiment (first embodiment) of the first invention, and FIG. 2 shows grayscale conversion of image data at a blast furnace taphole. Explanatory diagram (shared with conventional example), FIG. 3 is an explanatory diagram showing binarization processing of a grayscale image (shared with conventional example), and FIG. 4 is explanatory diagram of time series data of grayscale image (shared with conventional example). ，
5 is an explanatory diagram of time-series data of a binarized image (shared with a conventional example), FIG. 6 is an explanatory diagram showing difference image processing, FIG. 7 is an explanatory diagram of a difference image of binarized time-series data, FIG. 8 Is a chart showing the transition of the number of white pixels in the difference image of the binarized time-series data, and FIG. 9 is a schematic flow chart of a moving image data analysis method according to an embodiment (second embodiment) of the second invention, FIG. 10 shows the third
FIG. 11 is a schematic flow chart of a moving image data analyzing method according to one embodiment (third embodiment) of the invention of FIG. 11, FIG. 11 is an explanatory diagram of a binarized moving image when the threshold value is 64, and FIG. 13 is an explanatory diagram of the difference image when the value is 64, FIG. 13 is a chart showing the transition of the number of white pixels in the difference image when the threshold value is 64, and FIG. 14 is white in the difference image when the threshold value is 128 FIG. 15 is a chart showing the transition of the number of pixels, and FIG. 15 is a chart showing the transition of the number of white pixels in the difference image when the threshold value is 192.

The same symbols are used for the same elements as those in the schematic process flow chart of the example of the conventional moving image data analysis method shown in FIG. As shown in FIG. 1, a moving image data analysis method according to an embodiment (first embodiment) of the first invention performs binarization processing on input moving image data (S1 to S3), Difference image data is sequentially obtained for two binarized moving image data that are temporally adjacent to each other (S
4), the change degree of the input moving image data is obtained based on the change of the difference image data (S5). Hereinafter, also in the first embodiment, a case will be described in which a change in the appearance of sparks generated at the blast furnace tap hole similar to the conventional example is detected. First, color image data is input (S1), and grayscale conversion is performed on the color image data to express only image brightness (S2). Then, paying attention to the brightness of each pixel of the gray scale image, a binarization process is performed in which if it is above a certain threshold value, it is converted to white, and if it is below the threshold value, it is converted to black (S3). Up to this point, it is completely the same as the conventional example, and the obtained time-series grayscale image and binarized image are also the same as the conventional example (see FIGS. 2 to 5). However, in the first embodiment, as shown in FIG. 6, the difference (exclusive OR) between the two images is taken to create a difference image (S4). Regarding the time series data of the binarized image of FIG. 5,
FIG. 7 shows the result of taking a difference image for two adjacent frames. FIGS. 8A and 8B show the calculated number of white luminance pixels for the time-series image data of the difference image.
As described above, by using the difference image, it is possible to easily detect a situation in which the screens of 200 frames and 300 frames are largely changed, which cannot be detected in the conventional example. As a result, it was possible to detect that the condition of the taphole of the blast furnace changed significantly at this point, and the amount of sparks increased rapidly. That is, according to the first embodiment, it is possible to accurately detect the situation change of the detection target by utilizing the characteristics of the moving image, instead of treating the time-series image data as still images as in the conventional example. I was able to.

By the way, in the conventional example, the threshold value is obtained by the method used for still image data. However, if this value is used as it is for moving image data, there is a problem so that a method for determining the threshold value for moving image data is used. I have already mentioned what is needed. The second and third inventions have been made with this point in mind, and the second and third inventions will be described below. As shown in FIG. 9, a moving image data analysis method according to an embodiment (second embodiment) of the second aspect of the invention uses a preset initial threshold value for input moving image data. Quantization processing is performed (S11 to S1
4) The threshold value is updated until the temporal change of the binarized moving image data becomes a predetermined value or more (S15-S).
17), the change degree of the input moving image data is obtained based on the change of the moving image data binarized by the updated threshold value (S18). In this case, since the change in the binarized moving image data can be regarded as a function of the threshold value, the maximum value of the change amount is calculated by using a well-known optimization method such as the golden section method. As a result, an appropriate threshold value for moving image data can be obtained. Therefore, the situation change of the detection target can be captured more accurately than in the case of using the threshold for still image data as in the conventional example.
Further, as shown in FIG. 10, the moving image data analyzing method according to one embodiment (third embodiment) of the third invention uses an initial threshold value set in advance for an input moving image. Binarization processing is performed (S21 to S24), differential image data is sequentially obtained from two temporally adjacent binarized moving image data (S25), and the temporal change of the differential image data is a predetermined value. The above threshold is updated until it becomes the following (S2
6 to S28), the degree of change of the input moving image data is obtained based on the change of the difference image data obtained using the updated threshold value (S29).

In this case, the situation of the object to be detected can be more accurately grasped by the synergistic action of the first and second embodiments. By the way, the second embodiment is applied to the case where the effect can be obtained only by updating the threshold value without performing the differential imaging processing of the third embodiment. It can be said that this corresponds to the case where the differential imaging processing of the embodiment is omitted. Therefore, since the contents of the second embodiment will be clarified by making the third embodiment more specific, the third embodiment will be mainly described below. In the third embodiment as well, the moving image data analysis is performed on the moving image of the spark appearance at the blast furnace taphole. That is, a case will be described in which the place where the appearance of spark changes most is detected in the grayscale moving image shown in FIG. First, for example, 64 is preset as an initial threshold value for binarization processing (S21). Next, input color image data (S2
2) Then, this is gray scale converted to be expressed only by the image brightness (S23). Then, paying attention to the brightness of each pixel of the gray scale image, a binarization process is performed in which if it is equal to or more than the initial threshold value, it is converted to white, and if it is less than the threshold value, it is converted to black (S24). The binarized moving image at this time is as shown in FIG. Next, when the difference image of two adjacent frames is taken from this binarized moving image, the difference image as shown in FIG. 12 is obtained (S25). In order to extract the change amount of the difference image, FIG. 13 is a plot of the time change of the number of white pixels in each difference image (S26). According to the figure, even though the moving image has a large change in the 100th frame, the 200th frame, and the 300th frame,
There is not much change in the number of white pixels between the -200 difference image and the 200-300 difference image (S27).

Next, the threshold value is changed to 128 (S28), and the processes of steps S24 to S26 are performed again to obtain FIG. From the figure, it can be seen that the number of white pixels greatly changes between the difference image of 100-200 and the difference image of 200-300 corresponding to the moving image (S27). As a result of repeating the same processing, it was found that the threshold value at which the largest change in the number of white pixels appears was 192, as shown in FIG. However, as a specific method of determining the threshold value, one of the optimization methods described in the second embodiment was used. This determined threshold value is used as a threshold value for the binarization process for the moving image of the spark appearance at the taphole. If this threshold value is used, it is possible to easily and automatically extract the place where the appearance of the spark changes the most in the same moving image of the taphole (S29). As described above, according to the third embodiment, as in the second embodiment, by updating the threshold value, an appropriate threshold value for moving image data can be obtained. Further, according to the third embodiment, instead of treating the time-series image data as still images as in the conventional example, the difference image data that makes use of the characteristics of the moving image as in the first embodiment is used. By using, it was possible to detect the change in the situation of the detection target more accurately.

In the first to third embodiments described above, the input image data is color image data converted into gray scale, but the present invention is not limited to this. The data may be input image data. In that case, the grayscale processing is simply omitted, and it is easy to apply this method.
In the first to third embodiments, the case where the color image data is time-series image data of the blast furnace tap hole and the degree of change is the appearance of sparks has been described. ~
The third invention is not limited to this, and can be applied to moving image data analysis in other fields. It goes without saying that it is particularly effective when the situation of the detection target changes in time series. In the first embodiment, in the moving image data analysis in step S5,
In the embodiment, the number of white luminance pixels of the difference image is calculated in the extraction of the change amount of the difference image data in step S26, but the number of black pixels may be calculated in actual use. In the second and third embodiments, when the threshold value is updated, the maximum change amount of the binarized image data or the difference image data is obtained. When inputting one after another, it is necessary to obtain the maximum value of the change amount for every certain number of data or time. In this case, the number of data or the time may be set according to the expected change status of the detection target.

[0012]

The moving image data analysis method according to the present invention comprises:
Since it is configured as described above, instead of treating time-series image data as still image data as in the conventional example, by utilizing the characteristics of moving images, and /
Alternatively, it is possible to accurately detect the situation change of the detection target by obtaining an appropriate threshold value for the moving image data.

[Brief description of drawings]

FIG. 1 is a schematic process flow diagram of a moving image data analysis method according to an embodiment (first embodiment) of the first invention.

FIG. 2 is an explanatory view showing grayscale conversion of image data of a blast furnace taphole (shared with a conventional example).

FIG. 3 is an explanatory view showing binarization processing of a grayscale image (shared with a conventional example).

FIG. 4 is an explanatory diagram of time-series data of a grayscale image (shared with a conventional example).

FIG. 5 is an explanatory diagram of time-series data of a binarized image (shared with a conventional example).

FIG. 6 is an explanatory diagram showing difference image processing.

FIG. 7 is an explanatory diagram of a difference image of binarized time series data.

FIG. 8 is a chart showing changes in the number of white pixels in a difference image of binarized time-series data.

FIG. 9 is a schematic flowchart of a moving image data analysis method according to an embodiment (second embodiment) of the second invention.

FIG. 10 is a schematic flowchart of a moving image data analysis method according to an embodiment (third embodiment) of the third invention.

FIG. 11 is an explanatory diagram of a binarized moving image when the threshold value is 64.

FIG. 12 is an explanatory diagram of a difference image when the threshold value is 64.

FIG. 13 is a chart showing changes in the number of white pixels in a difference image when the threshold value is 64.

FIG. 14 is a chart showing changes in the number of white pixels in a difference image when the threshold value is 128.

FIG. 15 is a chart showing changes in the number of white pixels in a difference image when the threshold value is 192.

FIG. 16 is a schematic process flow chart in an example of a conventional moving image data analysis method.

FIG. 17 is a chart showing changes in the number of white pixels in binarized time series data.

[Explanation of symbols]

 S1 ... Color image data input step S2 ... Gray scale processing step S3 ... Binarization processing step S4 ... Difference image processing step S5 ... Moving image data analysis step

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Tetsuo Okuda 1 Kanazawa Town, Kakogawa City, Hyogo Prefecture Kadogawa Works Kakogawa Works

Claims (5)

[Claims] 1. A binarization process is performed on input moving image data, differential image data is sequentially obtained from two temporally adjacent binarized moving image data, and the difference image data is changed. A moving image data analysis method for obtaining the degree of change of the input moving image data based on the above. 2. Binarization processing is performed on the input moving image data with a preset initial threshold value. The threshold value is updated until the temporal change of the binarized moving image data becomes a predetermined value or more, and the threshold value is updated based on the change of the binarized moving image data by the updated threshold value. A moving image data analysis method for obtaining the degree of change of input moving image data. 3. Binarization processing is performed on input moving image data with a preset initial threshold value, and differential images are sequentially added to two binarized moving image data that are temporally adjacent to each other. Data is obtained, the threshold value is updated until the temporal change of the difference image data becomes a predetermined value or more, and the input is made based on the change of the difference image data obtained by using the updated threshold value. Video data analysis method for obtaining the degree of change in the generated video data. 4. The moving image data analysis method according to claim 1, wherein the moving image data is gray scale of color image data. 5. The moving image data analyzing method according to claim 4, wherein the color image data is time-series image data of the blast furnace taphole, and the degree of change is the degree of sparking.