Rock wool electric melting furnace liquid level detection method and system based on machine vision

Abstract

The invention provides a rock wool electric melting furnace liquid level detection method and system based on machine vision, and the detection method comprises the steps: extracting liquid level features through an image preprocessing technology: firstly, eliminating Gaussian noise and salt and pepper noise in an image in a high-temperature environment through a median filtering algorithm, and protecting liquid level edge information; then extracting a liquid level through an image pixel processing technology, and carrying out multiple times of expansion and corrosion on the image to realize image enhancement; then extracting an edge contour by adopting edge detection based on a Canny operator, and taking the distance from the extracted farthest point of the liquid level edge to the image edge as a liquid level feature value; and finally, calculating the extracted liquid level characteristic value to obtain a real liquid level value. The detection system comprises a hardware part and a software part, the hardware part comprises an industrial control computer, an industrial camera, a relay alarm device and a display, and the software part comprises an image acquisition module, an image processing module, a height calculation module, a data storage module, an alarm module and a height module.

Description

Technical Field

[0001] The present invention relates to the technical field of furnace monitoring, and in particular to a machine vision-based rock wool electric furnace liquid level detection method and system. Background Art

[0002] Rock wool is an efficient, fireproof and environmentally friendly inorganic thermal insulation material, which is widely used in many fields such as construction, electricity, metallurgy, and shipbuilding. The liquid level of rock wool in the rock wool electric furnace is an important parameter, which directly affects the melting and production efficiency of the furnace. In the electric furnace, the height of rock wool water (molten rock) needs to be accurately measured for effective furnace temperature management and raw material replenishment. Due to the high ambient temperature (1000℃) and dust interference during the production of rock wool liquid, the rock is not completely burned, and the liquid level will fluctuate. The existing liquid level detection technology cannot accurately measure the liquid level, such as laser, radar, etc. At present, liquid level detection is carried out by manual measurement, which has many disadvantages, such as high temperature exposure, harmful gases and high-temperature steam, and manual height depends on experience, which is easy to misjudge.

[0003] In recent years, with the rapid development of electronic circuit technology, image processing foundation, and optical vision technology, the relevant research of machine vision has made rapid progress. Machine vision has the advantages of high intelligence, good real-time performance, and multiple perception methods. In view of this situation, this paper proposes a rock wool electric furnace liquid level detection system and method based on machine vision. Summary of the invention

[0004] The present invention provides a rock wool electric melting furnace liquid level detection system and method based on machine vision, which uses an industrial camera to take pictures in real time, and then processes the images in real time to calculate the liquid level height.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] The method for detecting the liquid level of a rock wool electric melting furnace based on machine vision is characterized by comprising the following steps:

[0007] S1. Install an industrial camera above the liquid level and ensure that the camera can observe the liquid surface and the edge of the liquid surface; collect the edge contour image of the rock wool liquid through the industrial camera;

[0008] S2, using median filtering algorithm to eliminate Gaussian noise and salt and pepper noise in the image under high temperature environment and protect the liquid level edge information;

[0009] S3, extracting the liquid surface through image pixel processing technology, and dilating and corroding the image to achieve image enhancement;

[0010] S4, using edge detection based on the Canny operator to extract edge contours, and taking the distance from the farthest point of the extracted liquid level edge to the edge of the image as the liquid level feature value;

[0011] S5. Calculate the extracted liquid level characteristic value to obtain the real liquid level value.

[0012] As a further preferred embodiment of the present invention, S2 adopts a median filtering algorithm to eliminate noise. Median filtering is a nonlinear signal processing technology based on sorting statistics theory that can effectively suppress noise. Its basic principle is to replace the value of a point in a digital image or digital sequence with the median value of each point in a neighborhood of the point, so that the surrounding pixel values ​​are close, thereby eliminating isolated noise points; the specific steps are:

[0013] S21, using a template and sorting the pixels in the template according to the size of the pixel values, generating a monotonically increasing or decreasing two-dimensional data sequence, and outputting it using the following formula:

[0014] g(x,y)=med{f(xm,yn),(m,n∈W)}

[0015] In the formula, f(x, y) represents the original image, g(x, y) represents the processed image, and W is the two-dimensional template;

[0016] S22. An odd number of data is selected through a sampling window in the image for sorting, and the data to be processed is replaced by the sorted median value, so that the pixel value of the data is close to the surrounding pixel values, thereby eliminating isolated noise points.

[0017] As a further preference of the present invention, the template used in the median filtering algorithm is a two-dimensional template, and the filtering window is a 3*3, 5*5, or 7*7 area; in actual use, the window length can be enlarged and the most appropriate one can be selected until the filtering effect is satisfactory. For objects with slowly changing long contours, square and circular windows are generally used, and for sharp-angled objects, a cross-shaped window is generally used; the subsequent program uses a 3*3 rectangular area.

[0018] As a further preferred embodiment of the present invention, the step S3 introduces morphological processing to highlight the liquid surface, the steps are:

[0019] S31, traverse all pixel points of the image, observe the three-channel color information of the liquid surface, and extract the binary image of the liquid surface;

[0020] S32, performing opening and closing operations on the binary image;

[0021] The opening operation eliminates small noise in the image by first corroding and then dilating, while retaining larger objects and features;

[0022] The closing operation maintains the integrity of the object by filling the small holes inside the object through the method of dilation followed by corrosion, and can smooth the boundary of the object. In particular, it can effectively remove small holes and gullies without significantly changing the area of ​​the object, making the contour of the object smoother.

[0023] As a further preferred embodiment of the present invention, the edge detection algorithm of S4 adopts the Canny detection algorithm; the steps are as follows:

[0024] S41, use Gaussian filtering to reduce the noise of the image, the formula is:

[0025]

[0026] S42, calculate the gradient magnitude and direction of each pixel in the image, and use the Sobel operator to calculate the gradient direction and magnitude of the image. The formula is:

[0027]

[0028] In the formula, θ represents the gradient direction of the image, and G represents the amplitude of the image;

[0029] The gradient direction of each point is approximately quantized to one of the four angles of 0, 45, 90, and 135;

[0030] S43. Use the maximum suppression algorithm to eliminate the stray responses caused by edge detection; compare the gradient strength of the current pixel with the gradient interpolation points in the positive and negative gradient directions. If the current pixel is greater than the maximum value of the two interpolation points, then use this point as the edge point to complete non-maximum suppression; then perform double-threshold edge connection, select the minimum threshold as 50 and the maximum threshold as 200, and the optimal processing of edge detection in this environment can be achieved; finally, extract the liquid level characteristic values ​​in the area near the benchmark to improve the accuracy of subsequent measurements.

[0031] As a further preferred embodiment of the present invention, the S5 takes the farthest point of the liquid level edge as the liquid level intersection point, and according to the relationship that the liquid level height is proportional to the liquid level intersection point, the ordinate of the liquid level intersection point is obtained as the effective distance; the steps are:

[0032] S51, traverse all edge contour points, for(…) to get the point set vector <point>;

[0033] S52, sorting the points according to their ordinates, sort(…);

[0034] S53, extracting the ordinate of the farthest point, which is the characteristic value of the liquid level at that time.

[0035] A rock wool electric furnace liquid level detection system based on machine vision, the detection system comprising: a hardware part and a software part;

[0036] The hardware part includes: an industrial computer, an industrial camera, a relay alarm device and a display; the industrial computer is used to host the main environment for the operation of the software part and interact with the industrial camera, the relay alarm device and the display; the industrial camera is used to collect liquid level images and send them to the software part for liquid level height calculation; the display is used to display the liquid level value; the relay alarm device includes a relay and an alarm light, which is used to issue an alarm when the liquid level reaches the warning position;

[0037] The software part includes: an image acquisition module, an image processing module, a height calculation module, a data storage module, an alarm module and a height module; the image acquisition module is used to obtain the edge contour image of the rock wool liquid; the image processing module is used to eliminate the noise in the image under the high temperature environment, extract the liquid surface through the image pixel processing technology, and enhance the image; the height calculation module obtains the liquid level characteristic value through the algorithm, and calculates the real liquid level value according to the relationship between the liquid level height and the liquid level intersection point; the data storage module is used to store the calculated liquid level value; the height module is used to transmit the liquid level value to the display; when the liquid level value exceeds the liquid level warning position, the alarm module is used to transmit the alarm information to the relay alarm device.

[0038] As a further preferred embodiment of the present invention, the industrial camera is a 6-megapixel @ 60fps JZY-3010FS camera equipped with an optical lens.

[0039] Compared with the prior art, the present invention has the following advantages or beneficial effects:

[0040] Aiming at the problem of liquid level detection of rock wool electric furnace under high temperature environment, this application extracts liquid level features through image preprocessing technology. First, the median filtering algorithm is used to eliminate Gaussian noise and salt and pepper noise in the image under high temperature environment, and the liquid level edge information is protected; then the liquid surface is extracted through image pixel processing technology, and the image is expanded and eroded multiple times to achieve image enhancement; then the edge detection based on Canny operator is used to extract the edge contour, and the distance from the farthest point of the extracted liquid level edge to the edge of the image is used as the liquid level characteristic value (the liquid level characteristic value is calculated by the algorithm); finally, the extracted liquid level characteristic value is calculated to obtain the true liquid level value. In addition, this application verifies the feasibility and practicality of the model through simulation experiments. By testing rock wool liquid images and videos of different liquid level heights, the results show that a rock wool electric furnace liquid level detection system and method based on machine vision provided by this application can effectively complete non-contact measurement of liquid level and can quickly obtain accurate measurement results. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0042] Figure 1 Schematic diagram of the equivalent of laser triangle measurement for the liquid level edge;

[0043] Figure 2 This is a schematic diagram of the laser triangulation method;

[0044] Figure 3 The liquid surface extracted before and after image morphological processing;

[0045] Figure 4 It is the Canny edge detection map of the image;

[0046] Figure 5 This is a structural diagram of the liquid level detection system for a rock wool electric furnace;

[0047] Figure 6 This is a schematic diagram of the liquid level detection test bench;

[0048] Figure 7 The liquid level of bucket No. 2 and the extracted liquid surface feature map;

[0049] Figure 8 for Figure 7 The image after image morphology processing;

[0050] In the picture, 1. Water bucket No. 1, 2. Water pump, 3. Water bucket No. 2, 4. Lens, 5. Computer that receives and processes data. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0052] The principle of visual inspection is to analyze and process the collected target object image, extract the required feature data for relevant calculation. Since the edge of the liquid level is always located at the furnace wall, when the furnace is vertical, the intersection of the liquid level edge and the furnace wall can be regarded as a vertical laser irradiating the liquid level, and the lens focuses the light reflected by the target object and forms an image. Figure 1 As shown in Figure 1, the furnace measurement model can be approximated as a laser triangulation measurement model. The following is an introduction to the laser triangulation measurement method.

[0053] Laser triangulation, such as Figure 2 As shown in the figure, a semiconductor laser irradiates the target with laser light. The light reflected from the target is collected by the light receiving lens and an image is formed on the light receiving element. Once the distance to the target changes, the angle of the collected reflected light will also change, and the image position on the light receiving element will change accordingly. The change in the image position on the light receiving element is proportional to the movement of the target, so the change in the image position can be measured as the movement of the target.

[0054] When the liquid level changes, the position of the laser on the camera imaging surface changes; assuming that the positions of the laser on the imaging surface before and after the liquid level change are A' and B' respectively, then A'B' is the moving distance of the laser line in the image, which is calculated by the following formula:

[0055]

[0056] Where K = MO / MO', MO' is the focal length of the camera, MO is the distance from the camera to the target; θ is the center point of the camera lens, θ1 is the angle between the laser beam emitted by the laser transmitter and the vertical direction, θ2 is the angle between the camera lens's normal viewing direction and the horizontal direction, X is the height of the liquid level rise or fall; ∠MBD is the angle between the liquid surface and the camera's line of sight, which is calculated by the following formula:

[0057] ∠MBD=arctan(H+XH′-Xtanθ1)∠MBD

[0058] =arctan(H'-Xtanθ1H+X)

[0059] Where H is the distance from the camera to the initial liquid surface, and H′ is the horizontal distance where the camera is installed.

[0060] The above formula is a complex nonlinear expression that involves multiple variables and trigonometric functions. However, under certain conditions (such as when θ1, θ2, H, and H' are fixed, and the change X is small relative to these fixed values), we can simplify or linearize the formula.

[0061] In practical applications, if the change in liquid level X is small relative to other fixed parameters (such as H, H', θ1, θ2), then the change in ∠MBD will also be small, so it can be approximately considered that the angle change in the sin*sin and cos*cos functions is linear. Under this approximation, we can simplify the above complex nonlinear relationship into a linear relationship, namely:

[0062] A'B'≈k*X+bA′B'≈k*X+b

[0063] In the formula, k and b are constants, which depend on the geometric configuration of the system and the camera parameters. From this formula, it can be seen that the liquid level height is proportional to the distance of the liquid surface intersection (the farthest point distance), and has a linear relationship.

[0064] Based on this, the present invention provides a method for detecting the liquid level of a rock wool electric melting furnace based on machine vision, comprising the following steps:

[0065] S1. Install an industrial camera above the liquid level and ensure that the camera can observe the liquid surface and the edge of the liquid surface; collect the edge contour image of the rock wool liquid through the industrial camera.

[0066] S2. Use the median filtering algorithm to eliminate Gaussian noise and salt and pepper noise in the image under high temperature environment and protect the liquid level edge information. The specific steps are:

[0067] S21, using a two-dimensional template and sorting the pixels in the template according to the pixel value, generating a monotonically increasing or decreasing two-dimensional data sequence, and outputting it using the following formula:

[0068] g(x,y)=med{f(xm,yn),(m,n∈W)}

[0069] In the formula, f(x, y) represents the original image, g(x, y) represents the processed image, and W is the two-dimensional template.

[0070] S22. An odd number of data is selected through a sampling window in the image for sorting, and the data to be processed is replaced by the sorted median value, so that the pixel value of the data is close to the surrounding pixel values, thereby eliminating isolated noise points.

[0071] S3. The liquid surface is extracted through image pixel processing technology, and the image is expanded and eroded to achieve image enhancement.

[0072] The specific steps are:

[0073] S31, traverse all pixel points of the image, observe the three-channel color information of the liquid surface, and extract the binary image of the liquid surface.

[0074] S32, perform opening and closing operations on the binary image. The opening operation eliminates small noise in the image by first corroding and then dilating, while retaining larger objects and features; the closing operation fills small holes inside the object by dilating and then corroding, maintaining the integrity of the object and smoothing the boundary of the object. In particular, it can effectively remove small holes and gullies without significantly changing the area of ​​the object, making the outline of the object smoother. Figure 3 shown.

[0075] S4. Use edge detection based on the Canny operator to extract the edge contour, and use the distance from the farthest point of the extracted liquid level edge to the edge of the image as the liquid level feature value. The specific steps are:

[0076] S41, use Gaussian filtering to reduce the noise of the image, the formula is:

[0077]

[0078] S42, calculate the gradient magnitude and direction of each pixel in the image, and use the Sobel operator to calculate the gradient direction and magnitude of the image. The formula is:

[0079]

[0080] In the formula, θ represents the gradient direction of the image, and G represents the amplitude of the image.

[0081] The gradient direction of each point is approximately quantized to one of the four angles: 0, 45, 90, and 135.

[0082] S43, use the maximum suppression algorithm to eliminate the stray response caused by edge detection; compare the gradient strength of the current pixel with the gradient interpolation points in the positive and negative gradient directions. If the current pixel is greater than the maximum value of the two interpolation points, then take the point as the edge point to complete the non-maximum suppression; then perform double threshold edge connection, select the minimum threshold as 50 and the maximum threshold as 200, and the optimal processing of edge detection in this environment can be achieved; finally, extract the liquid level feature value of the area near the benchmark to improve the accuracy of subsequent measurements. Image Canny edge detection diagram as shown Figure 4 shown.

[0083] S5. Calculate the extracted liquid level characteristic value to obtain the real liquid level value. In this step, the farthest point of the liquid level edge is taken as the liquid level intersection point, and the vertical coordinate of the liquid level intersection point is obtained as the effective distance based on the proportional relationship between the liquid level height and the liquid level intersection point. The specific steps are:

[0084] S51, traverse all edge contour points, for(…) to get the point set vector <point>;

[0085] S52, sorting the points according to their ordinates, sort(…);

[0086] S53, extracting the ordinate of the farthest point, which is the characteristic value of the liquid level at that time.

[0087] like Figure 5 As shown, the present invention provides a rock wool electric furnace liquid level detection system based on machine vision, including: a hardware part and a software part.

[0088] The hardware part includes: industrial control computers, industrial cameras, relay alarm devices and displays. The industrial control computers are used to host the main environment for the operation of the software part and interact with the industrial cameras, relay alarm devices and displays; the industrial cameras are used to collect liquid level images and send them to the software part for liquid level height calculation; the displays are used to display the liquid level values; the relay alarm devices include relays and alarm lights, which are used to sound an alarm when the liquid level reaches the warning position.

[0089] The software part includes: image acquisition module, image processing module, height calculation module, data storage module, alarm module and height module. The image acquisition module is used to obtain the edge contour image of the rock wool liquid; the image processing module is used to eliminate the noise in the image under high temperature environment, extract the liquid surface through image pixel processing technology, and enhance the image; the height calculation module obtains the liquid level characteristic value through the algorithm, and calculates the real liquid level value according to the relationship between the liquid level height and the liquid level intersection point; the data storage module is used to store the calculated liquid level value; the height module is used to transmit the liquid level value to the display; when the liquid level value exceeds the liquid level warning position, the alarm module is used to transmit the alarm information to the relay alarm device.

[0090] To further explain, industrial cameras are mainly composed of several basic modules such as image sensors, internal processing circuits, data interfaces, IO interfaces, and optical interfaces. When the camera is shooting, the light signal first reaches the image sensor through the lens, and then is converted into an electrical signal. The internal processing circuit then processes the image signal algorithmically, and finally transmits data to the host computer through the data interface in accordance with the relevant standard protocol. The IO interface provides signal interaction between the camera and upstream and downstream devices, such as using input signals to trigger the camera to take pictures, and the camera outputs strobe signals to control the light source to light up. In the application of rock wool electric furnaces, high temperature is the biggest test for the detection equipment, so reasonable high temperature protection devices must be used to ensure the normal operation of the image acquisition equipment, and long-term work will carry dust, corrosive gases, etc. In view of the on-site environment, this solution uses the 6 million (3072×2048) @60fps JZY-3010FS high temperature wind and water cooling camera. With the liquid level detection algorithm, the detection accuracy can reach 0.1mm, which can meet the needs of accurate liquid level detection.

[0091] To further explain, in order to obtain a complete high-definition image, the industrial camera must also be equipped with a suitable optical lens. Optical lenses have several important parameters such as focal length, resolution, aperture, field of view and depth of field. The focal length is the distance from the front and rear principal planes (image / object principal planes) to the imaging focus, and the calculation formula is:

[0092]

[0093] Wherein, d is the focal length, mm; d is the working distance, mm; v is the rake surface size, mm; V is the field of view size in the detection direction, mm.

[0094] The calculation formula for the harrow surface size v is:

[0095]

[0096] Where m represents the pixel size, mm; F ​​represents the resolution.

[0097] The depth of field includes the foreground depth of field and the background depth of field, and the calculation formula is:

[0098] ΔL=ΔL1+ΔL2

[0099]

[0100] In the formula, σ represents the diameter of the confusion circle, mm; f represents the focal length of the lens, mm; F ​​represents the lens aperture value, mm; L represents the focusing distance of the lens, mm; ΔL1 represents the foreground depth of field, mm; ΔL2 represents the back depth of field, mm; ΔL represents the full depth of field, mm.

[0101] According to the actual situation of the rock wool electric melting furnace, select a lens with appropriate resolution and a large depth of field, use the above formula to calculate the appropriate depth of field, and adjust the focal length.

[0102] To verify the above detection system, the present application was conducted as follows Figure 6 The simulation experiment shown.

[0103] Experimental equipment: Water bucket No. 1 1, water pump 2, water bucket No. 2 3, MV-CS060-10GC industrial camera, 5mm lens 4, computer for receiving and processing data 5. Among them, water bucket No. 1 1 is used to store water, water bucket No. 2 3 is used to observe the liquid level with an industrial camera, and the liquid level scale in water bucket No. 2 2 is marked in advance, ranging from 0 to 15cm. Water pump 2 is used to supply water from water bucket No. 1 1 to water bucket No. 2 2, simulating the actual environment of an electric furnace directly feeding from above. The computer collects, records and calculates the liquid level in real time. The original image of the liquid level and the image after image processing, as shown in Figure 7 and 8 shown.

[0104] The collected images are preprocessed to extract the liquid level arc, and then the liquid level characteristic value algorithm is used to calculate the characteristic value of the image. The liquid level scale value in the image is observed, and the actual liquid level and liquid level characteristic value data are recorded at this time, as shown in Table 1.

[0105] Table 1 Partial data of actual liquid level Y and liquid level characteristic value X

[0106]

[0107]

[0108] According to the proportional relationship between the liquid level height and the liquid level characteristic value, that is, the linear relationship, the least squares algorithm is used to fit the linear relationship. The fitted relationship is as follows:

[0109] Y=0.0253165X-22.6203015

[0110] The extracted liquid level characteristic values ​​are compared with the actual liquid level height to verify the feasibility of the model in this paper.

[0111] The following is a comparison between the actual liquid level and the measured value. The pump liquid flow rate is maintained during the experiment. The actual liquid level, measured value and error data are shown in Table 2.

[0112] Table 2 Partial values ​​of error data between actual liquid level and measured liquid level

[0113] serial number Actual liquid level (cm) Measuring liquid level (cm) Error(cm) 1 4.99 5.000010 0.010010 2 5.29 5.278481 0.011519 3 5.70 5.683455 0.016545 4 6.15 6.164559 0.014559 5 6.55 6.569623 0.019623 6 7.00 7.000035 0.000035 7 6.50 6.518990 0.018990

[0114] It can be seen from Table 2 that the average error between the actual liquid level and the measured liquid level is 0.1311mm, the maximum absolute error is 0.215mm, and the accuracy can reach within 0.2mm. It can be seen that the rock wool electric furnace liquid level detection system provided by this application has a significant effect on rock wool liquid level detection in a non-contact high-temperature environment and has strong industrial application value.

[0115] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structural transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.< / point> < / point>

Claims

1. A method for detecting the liquid level of a rock wool electric furnace based on machine vision, characterized in that: The steps include: S1. Install an industrial camera above the liquid level and ensure that the camera can observe the liquid surface and the edge of the liquid surface; collect the edge contour image of the rock wool liquid through the industrial camera; S2, using median filtering algorithm to eliminate Gaussian noise and salt and pepper noise in the image under high temperature environment and protect the liquid level edge information; S3, extracting the liquid surface through image pixel processing technology, and dilating and corroding the image to achieve image enhancement; S4, using edge detection based on the Canny operator to extract edge contours, and taking the distance from the farthest point of the extracted liquid level edge to the edge of the image as the liquid level feature value; S5. Calculate the extracted liquid level characteristic value to obtain the real liquid level value.

2. The method for detecting liquid level of a rock wool electric furnace based on machine vision according to claim 1 is characterized in that: The specific steps of S2 are: S21, using a template and sorting the pixels in the template according to the size of the pixel values, generating a monotonically increasing or decreasing two-dimensional data sequence, and outputting it using the following formula: g(x,y)=med{f(xm,yn),(m,n∈W)} In the formula, f(x, y) represents the original image, g(x, y) represents the processed image, and W is the two-dimensional template; S22, selecting an odd number of data for sorting, and replacing the data to be processed with the sorted median value, so that the pixel value of the data is close to the surrounding pixel values, thereby eliminating isolated noise points.

3. The method for detecting liquid level of a rock wool electric furnace based on machine vision according to claim 2 is characterized in that: The template used by the median filtering algorithm is a two-dimensional template, and the filtering window is a 3*3, 5*5, or 7*7 area; in actual use, the filtering effect is improved by enlarging the window length.

4. The method for detecting liquid level of a rock wool electric furnace based on machine vision according to claim 1 is characterized in that: The specific steps of S3 are: S31, traverse all pixel points of the image, observe the three-channel color information of the liquid surface, and extract the binary image of the liquid surface; S32, performing opening and closing operations on the binary image; The opening operation uses a method of first erosion and then expansion to eliminate small noise in the image while retaining larger objects and features; The closing operation adopts a method of first dilation and then erosion to fill small holes inside the object, thereby maintaining the integrity of the object and smoothing the boundary of the object.

5. The method for detecting liquid level of a rock wool electric furnace based on machine vision according to claim 1, characterized in that: The specific steps of S4 are: S41, use Gaussian filtering to reduce the noise of the image, the formula is: S42, calculate the gradient magnitude and direction of each pixel in the image, and use the Sobel operator to calculate the gradient direction and magnitude of the image. The formula is: In the formula, θ represents the gradient direction of the image, and G represents the amplitude of the image; The gradient direction of each point is approximately quantized to one of the four angles of 0, 45, 90, and 135; S43. Use the maximum suppression algorithm to eliminate the stray responses caused by edge detection; compare the gradient strength of the current pixel with the gradient interpolation points in the positive and negative gradient directions. If the current pixel is greater than the maximum value of the two interpolation points, then take this point as the edge point to complete non-maximum suppression; then perform double threshold edge connection to achieve the optimal processing of edge detection in this environment; finally, extract the liquid level feature value of the area near the benchmark.

6. The method for detecting liquid level of a rock wool electric furnace based on machine vision according to claim 5 is characterized in that: In S4, the minimum threshold for selecting the dual-threshold edge connection is 50, and the maximum threshold is 200.

7. The method for detecting liquid level of a rock wool electric furnace based on machine vision according to claim 1 is characterized in that: The S5 takes the farthest point from the edge of the liquid level as the liquid level intersection point, and obtains the ordinate of the liquid level intersection point as the effective distance based on the relationship that the liquid level height is proportional to the liquid level intersection point; the steps are: S51, traverse all edge contour points to obtain a point set; S52, sorting the points according to their ordinates; S53, extracting the ordinate of the farthest point, which is the characteristic value of the liquid level at that time.

8. A rock wool electric furnace liquid level detection system based on machine vision, by using the detection system, the detection method according to any one of claims 1 to 7 can be implemented, characterized in that: The detection system comprises: a hardware part and a software part; The hardware part includes: an industrial computer, an industrial camera, a relay alarm device and a display; the industrial computer is used to host the main environment for the operation of the software part and interact with the industrial camera, the relay alarm device and the display; the industrial camera is used to collect liquid level images and send them to the software part for liquid level height calculation; the display is used to display the liquid level value; the relay alarm device is used to issue an alarm when the liquid level reaches the warning position; The software part includes: an image acquisition module, an image processing module, a height calculation module, a data storage module, an alarm module and a height module; the image acquisition module is used to obtain the edge contour image of the rock wool liquid; the image processing module is used to eliminate the noise in the image under the high temperature environment, extract the liquid surface through the image pixel processing technology, and enhance the image; the height calculation module obtains the liquid level characteristic value through the algorithm, and calculates the real liquid level value according to the relationship between the liquid level height and the liquid level intersection point; the data storage module is used to store the calculated liquid level value; the height module is used to transmit the liquid level value to the display; when the liquid level value exceeds the liquid level warning position, the alarm module is used to transmit the alarm information to the relay alarm device.

9. The machine vision-based rock wool electric furnace liquid level detection system according to claim 8 is characterized in that: The industrial camera is a 6-megapixel @ 60fps JZY-3010FS camera equipped with an optical lens.

10. The machine vision-based rock wool electric furnace liquid level detection system according to claim 8, characterized in that: The relay alarm device comprises a relay and an alarm lamp.

Images