JPH0944671A - High brightness object recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a high-luminance object recognizing device with which an object having high luminance almost equal to the background (such as overflowing molten iron, for example), is exactly recognized out of the high- luminance background (such as fire, for example), changing its position and shape with the passage of time in a short time by image processing without being affected by the change of background. SOLUTION: Concerning an input image to recognize a high-luminance object out of a high-luminance background, edge extraction is performed by differentiating processing (S3), each unit is emphasized, binarizing processing is performed for distinguishing whether it is edge or not (S4), labeling processing is performed to each unit for distinguishing each unit (S5), the number of units is counted (S6) and when the count value is higher than a threshold value, the high- luminance object existent in the high-luminance background is recognized (S7). When the count value is less than the threshold value, it is judged that any high-luminance object does not exist.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for recognizing a high-luminance object having a luminance level similar to that of a background from an ever-changing high-luminance background by using an image processing method.

[0002]

2. Description of the Related Art In the hot metal desulfurization process in the steelmaking process, it has been conventionally performed by an operator to visually check whether or not hot metal spillage has occurred once every several hours. There is. Therefore, the work of monitoring hot metal spillage by such visual confirmation is one of the obstacles in advancing the complete automation of the steelmaking process. As an automatic system for monitoring hot metal spillage, a surveillance camera periodically takes a picture of the surroundings of the pot, and the resulting captured image is subjected to image processing to identify the spilled hot metal that is the object of recognition from the background flame. A method of finding can be considered.

However, monitoring of hot metal spilled by image processing has not been realized. This is because the background flame has a high brightness and the hot metal to be recognized has a brightness as high as that of the flame. When the number of pixels having a brightness level equal to or higher than a certain threshold level is counted for the entire background (flame) of one screen, and an object to be recognized (blowover hot metal) is recognized based on the counted value, the background (flame) ) Itself has high brightness, even when there is no recognized object (hot metal spill), the count value increases and overrecognition occurs as if the recognized object exists.

A method of recognizing hot metal spillage using pattern matching is also conceivable. However, neither the flame as the background nor the hot metal spillage as the object to be recognized has an irregular shape over time. Changes to
It is impossible to recognize spilled hot metal.

By the way, one application example of the image processing technology in the molten steel process is the material "small cross section BTC" of the 3rd research presentation meeting of the measurement technology subcommittee of the equipment engineering department of NKK equipment technology center.
C Development of monitoring device in mold ”. This system is equipped with a TV camera for each strand, photographs the inside of the mold of continuous casting equipment, removes the flame image generated by the combustion of powder from the photographed image, and the unmelted powder layer becomes thin. This is to detect changes in the amount of radiant light that is observed. In the image processing in this system, a method of obtaining the minimum luminance value in the time series direction and replacing the image information with this minimum luminance value to remove the flame image is used.

[0006]

In the above-mentioned conventional technique, it is possible to remove a flame image whose position and shape are changing every moment, so that this conventional technique can be applied to image processing when recognizing hot metal spilling over. However, there is a problem as described below, so it is impossible to apply it. Objects that remain in the image after removal of the flame can only support those with a much lower brightness level than the high-intensity flame image, and with the method of replacing the image information with the minimum brightness value in the time series direction,
High-intensity objects (boiled hot metal) that are generated in a short time as the flame is removed are also removed.

As described above, it has been difficult to recognize an object of high brightness such as hot metal spilled from a background of high brightness such as flame by image processing.

The present invention has been made in view of the above circumstances, and an object having a brightness as high as the background is changed from a background having a high brightness whose position and shape change every moment. An object of the present invention is to provide a high-luminance object recognition device that can be recognized accurately and in a short time by image processing without being affected.

[0009]

A high-brightness object recognition apparatus according to the present invention is an apparatus for recognizing a high-brightness object having a brightness level similar to that of a background from a photographed high-brightness background image. Then, an image processing means for performing image processing including edge enhancement processing, binarization processing at a predetermined brightness level, and counting processing of the number of individuals on the photographed image is provided.

In the high-luminance object recognition apparatus of the present invention, the input image is edge-extracted to emphasize the individual, and then binarized,
The number of divided individuals is counted, and whether or not there is a high-luminance object to be recognized is determined based on the counted value. Since the high-brightness background is not separated and the whole area is connected,
When the above image processing is performed, the number of individuals becomes “0” or “1”. On the other hand, with respect to the high-intensity object to be recognized, the edges are emphasized, so that the individual objects are reliably separated and the number of individuals is counted, so that the count value becomes plural. Therefore, even when a high-brightness background and a high-brightness object that cannot be distinguished only by the brightness level are mixed, the high-brightness object can be recognized based on the count value of the number of individuals described above. When the count value is "0" or "1", the high brightness object does not exist, and when the count value is "2" or more, the high brightness object exists.

FIG. 1 is a flow chart showing the logic of image processing in the high brightness object recognition apparatus of the present invention.
The concept of the method in the high-luminance object recognition device of the present invention will be described below with reference to this flowchart.

First, an image taken by the camera is input (step S1). It is determined whether or not there is a high-luminance body having a high luminance level in the input image (step S2). In this case, it is checked whether or not the number of pixels having the luminance level exceeding the predetermined threshold level is equal to or more than a predetermined number, and when the number of pixels exceeding the threshold level is equal to or more than the predetermined number, it is determined that a high-luminance body exists and the step is performed. When the number of pixels exceeding the threshold level is less than the predetermined number, the process proceeds to S3, and it is determined that there is no high-luminance body, and the process returns to step S1.

Next, in order to emphasize each individual high-luminance object to be recognized, edge extraction by differential processing is performed (step S3). Then, binarization processing is performed to distinguish whether it is an edge (step S4), and labeling processing is performed on each individual in order to distinguish each individual according to the binarization result (step S4). S5). After that, the number of labeled individuals is counted (step S6), and it is determined whether or not the count value is equal to or more than a threshold value (a predetermined value of 2 or more) (step S7). If the count value is greater than or equal to the threshold value, the presence of a high-luminance object is recognized (step S
8), end. On the other hand, when the count value is less than the threshold value, it is determined that there is no high brightness object, and the process returns to step S1.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, with respect to an embodiment of a high-intensity object recognizing device of the present invention, in a hot metal desulfurization process in a steelmaking process, a flame is used as a high-intensity background, and a hot metal is used as a high-intensity object. This will be specifically described.

FIG. 2 is a schematic diagram showing the structure of the high-luminance object recognition device of the present invention. In FIG. 2, reference numeral 11 denotes an underground pit for performing hot metal desulfurization treatment, and a pot 12 for containing the hot metal to be treated is provided in the underground pit 11. A CCD camera 1 for photographing the vicinity of the upper surface of the pot 12 is provided diagonally above the pot 12 in the underground pit 11.

The image taken by the CCD camera 1 is sent to the image processing device 2. The image processing device 2 performs image processing including differential processing, labeling processing, and the like, which will be described later, on the input image to recognize whether or not a high-intensity object (boiled hot metal B) exists, and the recognition result is obtained. Is output to the recognition result display monitor 3. The image processing device 2 is
A CRT monitor 2a for displaying an input image and various images that change in accordance with image processing is provided. The recognition result display monitor 3 displays the recognition result by the image processing device 2 to notify the operator. It should be noted that when the hot metal spillage exists, the recognition result display monitor 3 may issue a warning.

FIG. 3 shows the CCD camera 1 during the hot metal desulfurization process.
It is a schematic diagram which shows the image imaged in. The captured image contains flame A rising from the pot 12 and hot metal B spilling from the pot 12. FIG. 4
[Fig. 4] is a diagram showing images of each frame for flame and hot metal. Flames do not always occur, and their shape also changes over time. Also,
The position and shape of the hot metal that spills over the air changes every moment. Further, in addition to this, scattered particles are generated by hitting an obstacle while the spilled hot metal falls. Therefore,
In this embodiment, a plurality of hot metal spilled particles, which are high-intensity objects, are recognized from the flame, which is a high-intensity background.

Since the flame A and the spilled hot metal B are photographed, the CCD camera 1 is adjusted for imaging a high-intensity body. In a fully automatic camera, adjustment takes time depending on the brightness level of a captured image, and erroneous recognition occurs in image processing for a saturated image. Under these circumstances, the CCD camera 1 was set as shown in Table 1 below.

[0019]

[Table 1]

FIG. 5 is a flow chart showing the procedure of image processing in the image processing apparatus 2 in this example, and the image processing which is a characteristic part of the present invention will be described below with reference to this flow chart.

First, an image taken by the CCD camera 1 is input to the image processing device 2 (step S11). The input image is displayed on the CRT monitor 2a of the image processing device 2. The image processing device 2 performs image processing as described below, but the processed images are sequentially displayed on the CRT monitor 2a as the image processing is executed.

The input image is binarized for each pixel at a predetermined brightness level (both the brightness level of flame and spilled hot metal is lower) (step S12).
It is determined whether or not the number of pixels exceeding the predetermined brightness level is equal to or larger than a predetermined number (step S13). This makes it possible to determine whether or not there is a high-intensity body (flame and / or hot metal spilling over). When there are a predetermined number or more of pixels that exceed the predetermined brightness level, it is determined that a high-intensity body exists, and the process proceeds to step S14. When the number of pixels that exceed the predetermined brightness level is less than the predetermined number, the high-intensity body is detected. When it is determined that there is not, the process returns to step S11. The process in step S13 corresponds to frame 4 in FIG.

Next, after waiting for input by the amount corresponding to the frames 5 to 9 in FIG. 3 (step S14), the image corresponding to the frame 10 in FIG. 3 (photographed image of scattered hot metal particles spilled over) is input. (Step S15). Smoothing processing is performed on the input image (step S16). By this smoothing processing, the noise component generated in the combination of the single-chip CCD camera and the monochrome image processing board is removed.
By removing this noise component, the subsequent counting process of the number of individuals (the number of particles of spilled hot metal scattered) can be accurately performed.

Next, a high-intensity object (scattering particles) to be recognized
In order to emphasize each individual, the edges are extracted by the differential processing in the two orthogonal directions (step S17). And
Binarization processing is performed to distinguish whether it is an edge (step S18). According to the binarization result, each scattered particle present on the right side of the pot 12 is labeled (step S19), and the number of individuals labeled on the right side of the pot 12 is counted (step S20). Similarly, part 2
According to the result of the quantification, each scattered particle on the left side of the pot 12 is labeled (step S21), and the pot 12 is heated.
The number of labeled individuals is counted on the left side of (step S22).

Next, the count values on the right side and the left side of the obtained pot 12 are respectively compared with predetermined threshold values to judge whether or not each count value is equal to or more than the threshold value (step S23). In this case, the predetermined thresholds for the count values on the right and left sides of the pot 12 are different. For example, the threshold value for the count value on the right side is 5
And the threshold value for the count value on the left side is 13. The reason why the threshold values are different in this way is that even if the hot metal spills on the right and left sides of the pot 12 are similar, the right side (or left side) of the pot 12 may be different depending on the mounting position of the CCD camera 1. This is because it may become a shadow part.

When the count value on the right side and / or the left side is equal to or more than the threshold value, the hot metal spilling on the right side and / or the left side is recognized (step S24), and the process ends. On the other hand, when all the counted values are less than the threshold value, it is determined that there is no spilled hot metal, and the process returns to step S11.

FIG. 6 is a schematic diagram of an image on the right side of the pot 12 after the binarization process in step S18. The edge is emphasized by the differential processing, and the differential image is set to 2 at a predetermined brightness level.
Since the value is converted into a value, for example, when the brightness level is expressed by 8 bits, the edge of the scattered particles is “255”, and the other is “0”. In the example shown in FIG. 6, six scattered particles are detected. Then, labeling is performed on such a binarized image. In this labeling, "0" to "255" are used, and the brightness value is sequentially set as the labeling number from the background ("0") to each scattered particle for each individual.
This state is shown in FIG. The size of each individual distinguished by the attached brightness value (labeling number) is represented by the number of pixels (vertical axis in FIG. 7). Then, in FIG. 7, when the maximum value of the brightness value (labeling number) is obtained for the pixel number larger than 0, the maximum value becomes the number of scattered particles. Since the obtained number of individuals (6 in this example) exceeds a predetermined threshold value (5 in this example), it is recognized that hot metal spillage occurred on the right side of the pot 12.

The same processing as in the case of the right side is performed on the image on the left side of the pot 12 to recognize whether or not hot metal spillage occurred on the left side of the pot 12.

The recognition result obtained by the above image processing as to whether or not hot metal spillage has occurred is output and displayed on the recognition result display monitor 3 to notify the operator.

Table 2 below shows the results of the actual detection of hot metal spilled by the above-mentioned method. As shown in Table 2, a detection rate of hot metal of spilled spillage of 100% and an over detection rate of 0% due to the influence of a flame, which is a high-intensity background, could be achieved.

[0031]

[Table 2]

In the above-mentioned example, only the presence or absence of hot metal spillage was recognized according to the number of the obtained splatter particles. However, when a large-scale hot metal spillage of hot metal occurs, Since the number is also large, it is possible to recognize not only the presence or absence of hot metal spillage but also the degree of spillage based on the number of scattered particles.

In the above example, the flame is used as the high-intensity background and the hot metal is used as the high-intensity object, and the hot metal is recognized in the hot metal desulfurization process in the steelmaking process. However, this is an example. Of course, the present invention can be applied to other examples in which an object having a brightness as high as the background is recognized from the high-brightness background.

[0034]

As described above, according to the present invention, the image processing including the edge enhancement processing, the binarization processing by the predetermined threshold brightness level and the counting processing of the number of individuals is performed, and the high brightness object is obtained based on the obtained number of individuals. Is recognized from the high-intensity background, an object with a high brightness as high as the background is selected from the high-intensity background whose position and shape change every moment, without being affected by the background change. It is possible to recognize accurately and in a short time by image processing.

[Brief description of drawings]

FIG. 1 is a flowchart showing a procedure of image processing in a high-luminance object recognition device of the present invention.

FIG. 2 is a schematic diagram showing the configuration of an embodiment of a high-luminance object recognition device of the present invention.

FIG. 3 is a schematic diagram showing a photographed image in an example of the present invention.

FIG. 4 is a diagram showing an example of an image in each frame of flame and spilled hot metal in an embodiment of the present invention.

FIG. 5 is a flowchart showing a procedure of image processing in the embodiment of the present invention.

FIG. 6 is a schematic diagram showing an image after binarization processing in the embodiment of the present invention.

FIG. 7 is a diagram showing an example of labeling processing according to the embodiment of the present invention.

[Explanation of symbols]

 1 CCD camera 2 Image processing device 2a CRT monitor 3 Recognition result display monitor 11 Underground pit 12 Pot A Flame B Blowout hot metal

Claims (1)

[Claims] 1. A device for recognizing a high-intensity object having a brightness level similar to that of the background of a captured high-intensity background image, wherein edge enhancement processing and predetermined processing are performed on the captured image. A high-brightness object recognition device comprising image processing means for performing image processing including binarization processing according to the brightness level and counting processing of the number of individuals.