TWI721901B - Method for estimating state of liquid level, computer program product, and computer readable recording medium - Google Patents

Abstract

A method for estimating a state of a liquid level, a computer program product, and a computer-readable recording medium are disclosed. The method includes: setting each of a plurality of thermal images as a target image one by one, and performing the target image on: searching at least one container edge around a liquid level area as at least one positioning baseline; positioning a region of interest in the liquid level area close to the positioning baseline; setting a region of at least one image adjacent to the target image as at least one comparison hot zone, wherein the region of interest maps the region of at least one image adjacent to the target image; and estimating a degree of difference between the comparison hot zone and the region of interest; and generating a state index according to the degree of difference.

Description

Liquid level estimation method, computer program product and computer readable recording medium

The present invention relates to a liquid level state evaluation technology, in particular to a liquid level state estimation method based on the image stream establishment of the liquid level of a steelmaking mold, a computer program product and a computer readable recording medium.

In the field of iron and steel smelting, the liquid level disturbance of the mold liquid level (such as the liquid level tumbling) is the cause of many steel-making defects. The conventional steel-making defect detection technology mostly uses eddy current or thermal phase meter. (thermal phase analyzer) for measurement.

Taking an infrared thermal imager (or called an infrared thermal imager) as an example, the liquid surface of the mold is photographed to generate thermal image streaming data, and then the liquid surface tumbling in the thermal image is manually observed. However, in addition to being unable to quantify the results of observing the image with the human eye, it is also prone to problems such as inconsistent standards and time-consuming operations.

In practice, defects caused by such tumbling, such as slag entrapment defects, often occur in the side cover area. For example, the conventional eddy current method is used to detect liquid level disturbances, because eddy currents can only measure a single point in space. The state of the position cannot be continuously measured for the side cover area that is prone to slag entrapment defects.

In view of this, it is necessary to provide a technical solution that is different from the past to solve the problems existing in the conventional technology.

One object of the present invention is to provide a method for estimating the liquid level state, based on the location of the region of interest in the thermal image near the area prone to slag entrapment defects, and the difference in the position of the region of interest of adjacent images is compared in order to be effective Reduce steelmaking defects.

The second purpose of the present invention is to provide a computer program product, based on the location of the region of interest in the thermal image near the area prone to slag entrapment defects, and compare the position of the region of interest in adjacent images to effectively reduce the slag. Steel defects.

Another object of the present invention is to provide a computer-readable recording medium with a built-in program, which is based on the location of the region of interest in the thermal image near the area prone to slagging defects, and compares the location of the region of interest in adjacent images Difference, in order to effectively reduce the defects of slag entrapment.

To achieve the above objective, one aspect of the present invention provides a liquid level state estimation method, which is configured to be executed by a processor to execute instructions stored in a memory. The method includes the steps of: acquiring a thermal image stream data The thermal image stream data includes a plurality of thermal images taken in a discrete time manner toward a heated liquid surface, each of the thermal images includes a liquid surface area and at least one container edge around the liquid surface area; Each of the thermal images is set as a target image one by one, and the target image is: search for the at least one container edge around the liquid surface area as at least a positioning baseline; and locate a region of interest in the liquid surface area In the vicinity of the positioning baseline; set the region of at least one adjacent image that is mapped to the target image by the region of interest as at least one hot area for comparison; and estimate the distance between the hot area for comparison and the area of interest A degree of difference; and a status indicator is generated based on the degree of difference.

In an embodiment of the present invention, the state indicators of the thermal images undergo a homogenization process to generate a homogenization state curve, determine whether the homogenization state curve is greater than a warning threshold, and if the determination is yes, output a warning signal, If the judgment is no, a record signal is output.

In an embodiment of the present invention, the step of searching the at least one container edge around the liquid level area as the at least one positioning baseline includes: detecting at least one edge line segment on at least a straight line around the liquid level area; And generating the at least one positioning baseline based on all edge line segments on the same straight track.

In an embodiment of the present invention, the at least one positioning baseline is generated according to the at least one edge line segment and a noise filtering algorithm.

In an embodiment of the present invention, the number of the positioning baselines is two, and the two positioning baselines intersect at one point.

In an embodiment of the present invention, the area of the previous image and/or the subsequent image that the region of interest is mapped to the target image is set as the comparison hot area.

In an embodiment of the present invention, at least one edge of the region of interest is parallel to at least one edge of the target image, and an intersection point between two adjacent edges of the region of interest is set as a positioning point. A positioning distance from the positioning point to one of the at least one positioning baseline is not greater than the sum of the lengths of two adjacent sidelines forming the positioning point.

In an embodiment of the present invention, the state indicator is a difference probability based on the degree of difference.

To achieve the above objective, another aspect of the present invention provides a computer program product. After the computer program is loaded and executed, the computer can execute the above-mentioned liquid level state estimation method.

In order to achieve the above objective, another aspect of the present invention provides a computer-readable recording medium. The computer can read a program stored in the recording medium. After the computer loads and executes the program, the computer can complete the above-mentioned Estimation method of liquid level.

The liquid level state estimation method, computer program product and computer readable recording medium of the present invention locate the region of interest in the liquid level area near the positioning baseline; and map the region of interest to the target image The area of at least one adjacent image of is set as at least one comparison hot area; and the degree of difference between the comparison hot area and the region of interest is estimated, and the state index of the target image is generated according to the degree of difference, which can be used for Track the position of the side cover in the steelmaking process to adjust the measurement area, and use the position of the side cover to locate the distance between the measurement area and the side cover, which can effectively detect the surrounding conditions of the side cover that are more prone to problems, so as to effectively reduce steel-making defects .

P1: Acquisition steps

P2: Edge measurement step

P3: Positioning steps

P4: Mapping steps

P5: Estimation steps

P6: Warning steps

P: interval

R: interval

R1: area of interest

D1: Positioning distance

D2: Positioning distance

L1: Position the baseline

L2: Position the baseline

M: target image

A: Anchor point

S ₁ : similarity

S ₂ : similarity

S ₃ : similarity

S ₄ : similarity

S ₅ : similarity

S ₆ : similarity

t: timeline

Figure 1: A schematic flow chart of a method for estimating a liquid level according to an embodiment of the present invention.

Figure 2: A schematic diagram of the location of the region of interest in the liquid level state estimation method according to the embodiment of the present invention.

Figure 3: A schematic diagram of the method for estimating the state of the liquid level according to the embodiment of the present invention for the numerical analysis of the difference in the image of the first region.

Fig. 4 is a schematic diagram of the liquid level state estimation method of the embodiment of the present invention searching for an area similar to an image in the first area of the first frame in a second area of a different frame.

Fig. 5: A schematic diagram of the difference estimation of the region of interest of adjacent frames in the liquid level state estimation method of the embodiment of the present invention.

Figure 6: A schematic diagram of continuous time rolling data obtained after continuous comparison and continuous calculation processing by the liquid level state estimation method of the embodiment of the present invention.

Figure 7: A schematic diagram of the liquid level state estimation method according to the embodiment of the present invention converted into a state index after numerical analysis processing.

In order to make the above and other objectives, features, and advantages of the present invention more obvious and understandable, the preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. Furthermore, the directional terms mentioned in the present invention, such as up, down, top, bottom, front, back, left, right, inside, outside, side, surrounding, center, horizontal, horizontal, vertical, vertical, axial, The radial direction, the uppermost layer or the lowermost layer, etc., are only the direction of reference to the attached drawings. Therefore, the directional terms used are used to describe and understand the present invention, rather than to limit the present invention.

An aspect of the present invention provides a method for estimating a liquid level. A processor can execute instructions stored in a memory. The memory is coupled to the processor, and the memory can also store Related data (such as data to be compared and comparison results, etc.), for example, the processor can be configured as a data processing device, platform or server, etc., the memory can be configured as a data storage medium or database, etc., the memory The body and the processor can be integrated as a whole, and there is no limitation here. It should be understood that the liquid level state can refer to the static or dynamic characteristics of the liquid surface, for example: liquid level fluctuations or disturbances (fluctuation), such as the liquid level tumbling state of the liquid being heated, or, in the liquid Gas is introduced into the liquid surface to generate bubbles or bubble collapsed state, etc., but not limited to this. As shown in Figure 1, the method includes: an image capturing step P1, an edge measurement step P2, a positioning step P3, a mapping step P4, and an estimation step P5.

In the image capturing step P1, a thermal image stream data can be obtained. The thermal image stream data includes a plurality of thermal images captured in a discrete time manner toward a heated liquid surface, and each of the thermal images includes a liquid surface area and At least one container edge around the liquid level area.

In the edge measurement step P2, each of the thermal images can be set as a target image one by one, and the target image is performed: searching for the at least one container edge around the liquid surface area as at least a positioning baseline.

In the positioning step P3, a region of interest (ROI) in the liquid surface area can be positioned near the positioning baseline.

In the mapping step P4, an area of at least one adjacent image that maps the region of interest to the target image may be set as at least one hot zone for comparison.

In the estimation step P5, a degree of difference between the comparison hot zone and the region of interest can be estimated; and a status indicator can be generated according to the degree of difference.

Optionally, in an embodiment, as shown in FIG. 1, the method may further include a warning step P6, which may perform an averaging process (for example, using a window averaging algorithm) to generate a warning step P6. The averaged state curve is used to determine whether the averaged state curve is greater than a warning threshold, if the judgment is yes, a warning signal is output, and if the judgment is no, a record signal is output.

The following is an example of the implementation of the above-mentioned liquid level state estimation method embodiment. Here, only the disturbance state of the mold liquid level for steelmaking is estimated as an example. For example, it can be used to estimate the mold liquid for steelmaking. The gas (such as argon) generates bubbles or bubble collapse state on the liquid surface, but not limited to this, but not limited to this. The embodiment of the liquid surface state estimation method can also be used to estimate other liquids (such as The state of the liquid level of water).

For example, the processor can obtain the thermal image stream data from the memory, and the thermal image stream data can be stored in the memory online or offline. For example, the thermal image stream data can be directed by an infrared thermal imager. The heating liquid surface (such as the mold liquid surface for steelmaking) is formed by capturing multiple thermal images in a discrete time manner. For example, the thermal images are captured intermittently on the time axis. The thermal images It can be a color thermal image, or a gray-scale thermal image of the color thermal image after gray-scale processing, and its color depth corresponds to different temperature values.

As shown in Figure 2, the processor can set each of the thermal images as a target image M one by one. Taking the gray-scale thermal image as an example, each thermal image includes a liquid surface area (see the shallower in the figure). Color strip area, such as the area where the mold liquid level is located, etc.) and at least one container edge around the liquid level area (see the dark background area in the figure, such as the side cover, etc.).

As shown in Figure 2, the processor can perform the target image M: search the at least one container edge (such as L1) around the liquid surface area as at least one positioning baseline, but not limited to this; for example, the positioning The number of baselines can also be two (such as L1, L2), and the two positioning baselines L1, L2 can intersect at one point to facilitate positioning and serve as a basis for the subsequent estimation method of the liquid level.

Optionally, in an embodiment, as shown in Figure 2, the step of searching the at least one container edge around the liquid level area as the at least one positioning baseline (such as L1, L2) may include: Detect at least one edge line segment on at least a straight track around the area, for example, an edge detection method (such as Hough transform method or RANSCA random search algorithm, etc.) can be used to detect the edge line segment; and based on the same straight track The at least one positioning baseline can be generated from all the edge line segments of, for example, all the edge line segments on a first straight line track can be connected and extended to form the positioning baseline L1, and all the edge line segments on a second straight line track can be connected and combined. Extend to form the positioning baseline L2. In this way, the at least one positioning baseline can be used as a positioning basis for evaluating the liquid level state (such as tumbling state, etc.) near the side line (such as the side cover), and can assist in determining whether the mold liquid surface near the side cover is tumbling. In order to effectively reduce steelmaking defects.

Optionally, in an embodiment, the at least one positioning baseline may also be generated according to the at least one edge line segment and a noise filtering algorithm (such as a recursive filtering algorithm such as Kalman filtering). In this way, if the picture is blurred (for example, it is derived from the vibration and other shaking during the molding process), the noise filtering algorithm can be used to continuously track the edge of the container (such as the side cover), and the sudden and drastic change of the edge of the container can be detected. The measurement result is removed to reduce the influence of false detection results on subsequent estimation results.

Subsequently, the processor may perform the target image: locate a region of interest in the liquid level area near the positioning baseline (for example, a single or multiple positioning baselines), as a subsequent estimation of the liquid level state Basis of the method.

Optionally, in an embodiment, as shown in FIG. 2, at least one side line of the region of interest R1 is parallel to at least one side line of the target image M, and a distance between two adjacent sides of the region of interest R1 is An intersection is set as a positioning point A, and a positioning distance from the positioning point A to one of the at least one positioning baseline (such as L1) (such as the vertical distance between A and L1) D1 is not greater than forming the positioning point A The sum of the lengths of two adjacent edges of is used to locate the region of interest R1. Specifically, for example, the region of interest can select a location that is likely to cause steelmaking defects due to turbulence of the liquid surface, such as next to the side cover; alternatively, multiple regions of interest can also be used to fill the liquid surface area for liquid surface Global detection, but not limited to this.

But not limited to this, the location rule of the region of interest R1 can also be changed according to the actual situation, for example, a location distance (such as A and L2) from the positioning point A to another of the at least one positioning baseline (such as L2) The distance between the mid-perpendicular line of L2) D2 is not greater than the two neighbors forming the positioning point A The sum of the lengths of the edges.

Subsequently, the processor may perform the target image: set the region of at least one adjacent image that is mapped to the target image by the region of interest as at least one hot spot for comparison, for example, as shown in FIG. 3, Based on a time axis t, the region of interest (the rectangular frame in the thermal image shown in the figure) mapped to the previous image and/or the subsequent image of the target image is set as the comparison hot area ( The rectangular frame in the thermal image as shown in the figure), where the rectangular frame is only used to indicate that the comparison hot area and the region of interest are regions with the same position in different thermal images.

Subsequently, the processor may perform the target image: estimate a degree of difference between the comparison hot zone and the region of interest, and generate a status indicator based on the degree of difference, for example, a probability of a difference generated by the degree of difference .

An example is as follows, taking the thermal image stream data including three frames composed of the target image and its front or back adjacent images as an example, but it is not limited to this.

For example, take the tumbling state of the hydrothermal fluid (such as the mold liquid level for steelmaking) as an example. Because the color depth of the pixels in the thermal image corresponds to different temperature values, when the liquid level area in the thermal image (such as the mold liquid level) ) When the tumbling phenomenon occurs, the changes in the color depth of the pixels in the entire liquid surface area represent changes in the local temperature. Therefore, the greater the variation between pixels of continuous thermal images, the higher the probability of the tumbling phenomenon. Among them, the range of the considered estimation is related to its accuracy. In order to prevent the range from being too large to affect the reliability of the estimation, the region of interest can be used as an estimation range. Among them, the tumbling probability of the liquid level at a position can be expressed by the following formula 1:

其中B(

,t)代表在某時刻t在某位置x之滾盪機率，該滾盪機率正比於溫度T的變化。

Where B(

, t ) represents the tumbling probability at a certain position x at a certain time t , and the tumbling probability is proportional to the change in temperature T.

In addition, if you want to measure the rolling probability in a region (such as the region of interest), you can integrate the rolling probability in the region according to the above equation 1, as shown in the following equation 2:

其中，

表示一特定的區域(如該感興趣區)。

among them,

Represents a specific area (such as the area of interest).

In addition, to compare the difference value of a specific area, you can Minute. The above-mentioned differentiation can use high-pass filtering to achieve the effect of differentiation. In the first frame, the second frame, and the third frame in several time intervals before and after in the image data, for a first region (such as a pixel grid in the region of interest), the image and the front and back The frame is compared to obtain the difference value of the image in the first area.

For example, please refer to FIG. 3 to perform a difference numerical analysis on the images in the first area of the first frame, the second frame, and the third frame. The above-mentioned difference numerical analysis can be expressed by the following formula:

or

According to the numerical analysis of the difference in the above formula, the degree of rollover represented by the above formula 2 can be rewritten as the following formula 3:

該第一區域內的滾盪程度在一定的溫度區間內，可以正比於下式：

The degree of tumbling in the first zone within a certain temperature range can be proportional to the following formula:

其中T(x,t)表示在時間t時在x位置的溫度變化。

可以被解讀為連續的數個時間區間內的該第一畫幅、該第二畫幅及該第三畫幅中的

區域的相似程度。影像相似程度越高(差異性越小)代表其滾盪機率越低。

Where T(x, t ) represents the temperature change at position x at time t.

Can be interpreted as the first frame, the second frame, and the third frame in several consecutive time intervals

The degree of similarity of the regions. The higher the degree of image similarity (the smaller the difference), the lower the probability of its rolling.

，例如作為該感興趣區與比對熱區。 Then, a second area may be defined around the first area of each of the first frame, the second frame, and the third frame, wherein the second area is larger than the first area. In other words, define the second area as

, For example, as the region of interest and comparison hot zone.

Referring to FIG. 4, in the second area between the first frame, search for an area similar to the image in the first area of the second frame to calculate a first similarity. In other words, based on the image in the first area of the second frame, search for an area similar to the image in the first area of the second frame in the R area between the first frame, and determine its The degree of similarity, for example, in Figure 4, S ₁ , S ₂ , S ₃ , S ₄ , S ₅ , and S ₆ can be used to represent a middle frame and its first three frames, two frames, one frame, and one The similarity of the frame, the last two frames, and the last three frames.

Then, referring to Figure 4, the first degree of similarity can also be exchanged for a degree of difference, and then the rolling probability can be obtained, so as to determine the rolling degree in the second area of the first frame, which can be used for As the state indicator of the target image in the region of interest, in this example, the state indicator can be used to indicate the tumbling state of the liquid level, but it is not limited to this.

Optionally, in the second area of the third frame, search for an area similar to the image in the first area of the second frame to calculate a second degree of similarity; and the second degree of similarity After converting to another degree of difference, another rolling probability is obtained to determine the rolling degree in the second area of the third frame. By analogy, n (for example, n may be a positive integer greater than 2) frames can be compared to the similarity of the second region in the manner described above.

Specifically, searching for an area similar to the image in the first area of the first frame in the second area of the front and rear frames, the above formula 3 can be rewritten as the following formula 4:

其中Sim(P,P _n,r )表示該第一畫幅在該第二畫幅中的該第二區域的一相似度；Diff(P,P _n,r )表示該第一畫幅在該第二畫幅中的該第二區域的一差異度；及

表示該第二區域內的範圍。

Where Sim ( P, P _{n, r} ) represents a similarity of the first frame in the second area in the second frame; Diff ( P, P _{n, r} ) represents the first frame in the second frame A degree of difference in the second region in; and

Indicates the range within the second area.

The foregoing calculation method can be as shown in Figure 5, where the P interval (for example, the first area) of one frame (for example, the first frame), and the R interval (for example, the second frame) of another frame (for example, the second frame) In the second area), search for the area with the greatest similarity. The above calculation method can be used to judge images in multiple consecutive time segments.

Optionally, in an embodiment, a fourth area may be defined around a third area of each of the first frame, the second frame, and the third frame, wherein the fourth area is larger than the third area , For example, the third area and the first area do not overlap, and the fourth area and the first area The two areas do not overlap; it is also possible to search for an area similar to the image in the third area of the second frame in the fourth area between the first frame to calculate a third degree of similarity; In the fourth area among the three frames, search for an area similar to the image in the third area of the second frame to calculate a fourth similarity; and the third similarity and the fourth similarity can be converted After the difference is another degree, another tumbling probability is obtained to determine another degree of tumbling in the fourth region, and then tumbling monitoring is performed on the entire liquid surface of the mold liquid surface.

Finally, the first degree of similarity can be replaced by a degree of difference, and then a rollover probability (can be used as a difference probability) can be obtained to determine a rollover degree in the second area (can be used as a degree of difference) As an indicator of the state of the target image in the area of interest, it can be used to indicate the difference state of the area of interest over time, and can be used as a basis for personnel or machines to perform related analysis or evaluation operations.

Optionally, the area with a high degree of rolling can also be marked. Please refer to Figure 6, after continuous comparison and continuous calculation processing, continuous time rolling data can be obtained. For example, the horizontal axis represents the sequence of continuous frames, and the vertical axis represents the degree of rolling.

After the continuous time rolling data obtained in Figure 6 is processed by numerical analysis (such as window averaging), the rolling probability can be converted into a graph as shown in Figure 7, for example, the horizontal axis represents time (in seconds), The vertical axis represents the degree of rolling, where the left box (for example, the area of interest in the first frame) represents the low probability of rolling in the period, and the right box (for example, the area of interest in the third frame) represents the period of time The probability of rolling is higher. In other words, the second area of the third frame has a tendency to increase the probability of rolling.

It should be noted that because the liquid level tumbling is a continuous phenomenon, and the disturbance of the liquid level during the tumbling process will last for a period of time, the large data of the state indicator (such as the magnitude of the value and the occurrence of higher values) can be used. Frequency, etc.), as an early warning whether there is rolling (that is, the difference in the liquid level in the continuous image) occurs.

Because the above calculation method involves the plane calculation of the estimated area (such as the region of interest), when the estimated interval is too large, the local change will be too small, making the numerical difference not enough to highlight the rolling state; on the contrary, when the estimated interval is too small, Small changes will cause great changes in values, and make the system overly sensitive. So the size of the comparison interval will affect the sensitivity. Therefore, the embodiments of the present invention can also adjust the size of the estimated area according to requirements. For example, when needed For large-scale operation detection, it is necessary to cover the calculated interval with the same size operation block for independent operation to obtain the difference value of each local area. Avoid using too large arithmetic area to perform calculations, and reduce the sensitivity to differences (such as rollover) changes.

On the other hand, the present invention also provides a computer program product. After the computer program is loaded and executed, the computer can execute the above-mentioned liquid level state estimation method. For example: the computer program product may contain several program instructions, which can be implemented in an existing programming language (such as C or Python, etc.) in order to perform the above-mentioned liquid level estimation method, but not Is limited.

On the other hand, the present invention also provides a computer-readable recording medium, such as an optical disc, a flash drive, or a hard disk, etc. The computer can read a program stored in the recording medium (such as the above-mentioned computer program), and when the computer loads the program After execution, the computer can complete the above-mentioned liquid level state estimation method.

Continuing from the foregoing, the liquid level estimation method, computer program product, and computer readable recording medium of the above-mentioned embodiments of the present invention are based on the location of the region of interest in the thermal image near the area prone to slag entrapment defects. The position of the area of interest of the adjacent images is compared to the difference in order to effectively reduce the slag entrapment defects.

For example, the measurement area of the thermal image needs to be slightly adjusted for the side cover area. This is because the size of each steel blank is different and needs to be adjusted in different product manufacturing processes. However, when the side cover is adjusted, the measurement area The entire movement is also required. One of the reasons is that the original measurement area may appear outside the adjusted measurement area, resulting in measurement failure; the other reason is that the measurement area of interest often appears on the die edge. So when the side cover is adjusted, the measurement area also needs to be adjusted.

Therefore, the liquid level state estimation method, computer program product, and computer readable recording medium of the above-mentioned embodiments of the present invention locate the region of interest in the liquid level area near the positioning baseline; and map the region of interest. The area of at least one adjacent image to the target image is set as at least one comparison hot zone; and the degree of difference between the comparison hot zone and the region of interest is estimated, and the target image is generated according to the degree of difference The status indicator can be used to track the position of the side cover in the steelmaking process to adjust the measurement area. Using the position of the side cover to locate the distance between the measurement area and the side cover, it can effectively detect the surrounding conditions of the side cover that are more prone to problems. Effectively reduce steelmaking defects.

Although the present invention has been disclosed in preferred embodiments, it is not intended to limit the present invention. Anyone familiar with the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the present invention The scope of protection shall be subject to the scope of the attached patent application.

P1: Acquisition steps

P2: Edge measurement step

P3: Positioning steps

P4: Mapping steps

P5: Estimation steps

P6: Warning steps

Claims (10)

A method for estimating the state of a liquid level is configured to be executed by a processor to execute instructions stored in a memory. The method includes the steps: Acquire a thermal image stream data, the thermal image stream data includes a plurality of thermal images taken in a discrete time manner toward a heated liquid surface, each of the thermal images includes a liquid surface area and surrounding areas of the liquid surface area. At least one edge of the container; each of the thermal images is set as a target image one by one, and the target image is: Searching the at least one container edge around the liquid level area as at least one positioning baseline; Positioning a region of interest in the liquid level area near the positioning baseline; Setting the region of at least one adjacent image that maps the region of interest to the target image as at least one hot spot for comparison; and Estimating a degree of difference between the comparison hot zone and the region of interest; and generating a status indicator based on the degree of difference. The liquid level state estimation method according to claim 1, wherein the state indicators of the thermal images are homogenized to generate a homogenization state curve, and it is judged whether the homogenization state curve is greater than a warning threshold, and if it is judged as yes , Output a warning signal, if judged as no, output a record signal. The liquid level state estimation method according to claim 1, wherein the step of searching the at least one container edge as the at least one positioning baseline around the liquid level area includes: Detect at least one edge on at least a straight track around the liquid surface area Line segment; and The at least one positioning baseline is generated according to all edge line segments on the same straight track. The liquid level estimation method according to claim 3, wherein the at least one positioning baseline is generated according to the at least one edge line segment and a noise filtering algorithm. The liquid level state estimation method according to claim 1, wherein the number of the positioning baselines is two, and the two positioning baselines intersect at one point. The liquid level state estimation method according to claim 1, wherein the region of the previous image and/or the subsequent image that maps the region of interest to the target image is set as the comparison hot zone. The liquid level state estimation method according to claim 1, wherein at least one side line of the region of interest is parallel to at least one side line of the target image, and an intersection point between two adjacent side lines of the region of interest is set as For a positioning point, a positioning distance from the positioning point to one of the at least one positioning baseline is not greater than the sum of the lengths of two adjacent sidelines forming the positioning point. The liquid level state estimation method according to claim 1, wherein the state indicator is a difference probability based on the degree of the difference. A computer program product, when the computer program is loaded and executed by the computer, the computer can execute the liquid level state estimation method as described in any one of claim items 1 to 8. A computer-readable recording medium. The computer can read a program stored in the recording medium. After the computer loads and executes the program, the computer can complete the liquid level state estimation as described in any one of claims 1 to 8 method.

Images