A visual perception-based ice thickness measurement method and device for an ice storage coil

Abstract

The application discloses a kind of ice thickness measurement method and device based on visual perception of ice storage coil, and the method comprises: setting annular or fan ring-shaped calibration disc on ice storage coil.The calibration disc is provided with multiple different color stripe regions arranged concentrically and along radial direction.Judge whether each stripe region is completely covered by ice layer according to the brightness change of calibration disc covered by ice layer in image, and each stripe region is divided into completely covered interval, partially covered interval and uncovered interval.For partially covered interval, the ice layer covering ratio of partially covered interval is obtained, and thus the ice layer thickness is obtained.The application provides the accuracy of ice thickness detection by stage ice thickness analysis, specifically, the ice layer completely covered by ice layer is determined by using the segmentation and brightness of different color stripe regions, and then for the partially covered stripe region, the comprehensive edge strength of integrated gradient edge strength and phase consistency edge strength is used for coverage ratio calculation.

Description

Technical Field

[0001] This invention belongs to the field of ice thickness measurement technology for ice storage devices, specifically relating to a method and device for measuring ice thickness in ice storage coils based on visual perception. Background Technology

[0002] In applications of ice storage equipment such as building energy systems and district cooling stations, accurate measurement of the ice layer thickness on the outer wall of the ice storage coil directly affects system energy efficiency optimization and safe operation control. Current mainstream technologies have the following significant limitations:

[0003] 1. Inherent defects of contact sensing technology: Contact measuring devices, such as temperature sensors and capacitance sensors, need to be directly attached to the surface of the coil. In low-temperature environments (below -30℃), they are easily affected by heat conduction interference from the ice interface, resulting in drift of the measured values. At the same time, the complex wiring structure and installation process significantly increase the deployment cost and maintenance difficulty, making it difficult to meet the multi-point monitoring needs of large energy stations.

[0004] 2. Insufficient environmental adaptability of image measurement technology: (1) Turbid underwater medium leads to increased incident light scattering, resulting in severe attenuation of image contrast and difficulty in effectively extracting ice layer edge features; (2) The superposition of ice layer transmission effect and water surface reflection causes color information distortion and reduces the reliability of color segmentation-based recognition; (3) Low-temperature frost forms irregular attachments on the surface of the coil, interfering with the discrimination of the actual contour of the ice layer. The above factors cause the false detection rate of traditional edge detection algorithms (such as the Canny operator) to increase in complex underwater environments, making it difficult to achieve sub-millimeter level precision measurement.

[0005] 3. System stability challenges in low-temperature environments: Existing technologies lack specific compensation mechanisms for low-temperature operating conditions. Geometric distortion caused by lens material shrinkage due to cold, color recognition deviation caused by light source color temperature drift, and parameter drift during long-term operation all reduce the long-term stability of the measurement system. Traditional linear calibration models are difficult to effectively fit actual nonlinear distortions, further limiting measurement accuracy.

[0006] In summary, existing technologies are limited by their weak ability to suppress environmental interference, poor dynamic adaptability, and insufficient calibration robustness. There is an urgent need to develop a high-precision, interference-resistant, non-contact ice thickness monitoring method. Summary of the Invention

[0007] The purpose of this invention is to provide a method and device for measuring ice thickness in ice storage coils based on visual perception.

[0008] In a first aspect, the present invention provides a method for measuring ice thickness in ice storage coils based on visual perception, comprising:

[0009] A circular or fan-shaped calibration disk is installed on the ice storage coil. The calibration disk has multiple concentric, radially arranged stripe regions of different colors. An image of the calibration disk is acquired using a camera outside the icing area. Each stripe region is extracted from the image through color segmentation.

[0010] Extracting the brightness gradient magnitude of pixels and brightness gradient direction By using two thresholds of different sizes, two classes of undetermined pixels are extracted from all stripe regions. This is based on the direction of the brightness gradient. Perform orientation consistency verification on two types of undetermined pixels.

[0011] Based on the brightness gradient magnitude and the directional consistency verification results of the two types of undetermined pixels, the gradient edge intensity of each undetermined pixel is obtained. The phase consistency edge intensity is then calculated for each undetermined pixel. The gradient edge intensity and the phase consistency edge intensity are normalized and then weighted and summed to obtain the comprehensive edge intensity.

[0012] Based on the brightness change of the calibration disk in the image after it is covered by ice, we can determine whether each stripe area is completely covered by ice. Each stripe area is divided into the inner completely covered area, the middle partially covered area that occupies one stripe area, and the outermost uncovered area.

[0013] For partially covered areas, the ice coverage ratio of the partially covered area is obtained by combining the comprehensive edge intensity of each pixel within the corresponding stripe area with the comprehensive edge intensity benchmark value. The ice thickness is then determined based on the inner and outer radii of the fully covered area and the ice coverage ratio of the partially covered area.

[0014] Preferably, the image extracts the brightness gradient magnitude. and brightness gradient direction Gaussian filtering is performed beforehand.

[0015] Preferably, the process of extracting Class I and Class II undetermined pixels is as follows: Set different lower and upper threshold values; classify pixels with brightness gradient magnitudes greater than the upper threshold as Class I undetermined pixels; and further classify pixels with brightness gradient magnitudes as Class II undetermined pixels. Pixels between the lower threshold and the upper threshold (inclusive) are classified as second-class undetermined pixels. The ratio of the lower threshold to the upper threshold is 1:3.

[0016] Preferably, the process of verifying the directional consistency is as follows: extract the brightness gradient directions of eight adjacent pixels of the two types of undetermined pixels. Establish a validation relation. If the verification relation is satisfied, it means that the two types of undetermined pixels have passed the orientation consistency verification. If the verification relation is not satisfied, it means that the two types of undetermined pixels have failed the orientation consistency verification.

[0017] Preferably, the gradient edge intensity of the first type of undetermined pixels The gradient edge intensity expression for the two types of undetermined pixels is: ;in, This represents the brightness gradient magnitude.

[0018] Preferably, the expression for the phase coherence edge strength is:

[0019]

[0020] Among them, ϕ n For local frequency phase, It represents the average phase of all pixels in the image.

[0021] As a preferred option, the ice coverage ratio of the partially covered area... The expression is as follows:

[0022]

[0023] in, It represents the sum of the edge intensities of all pixels within the partially covered area; k is the brightness correction coefficient. To obtain the baseline value for comprehensive edge intensity, the comprehensive edge intensity is extracted from the corresponding stripe regions on the unfrozen image and summed. This represents the difference between the average brightness of the corresponding striped area and the average brightness of the background. This represents the maximum brightness value in the image.

[0024] Preferably, the images acquired by the camera undergo preprocessing. The preprocessing steps include underwater scattering correction, color correction, and multi-scale denoising. The underwater scattering correction process involves calculating the transmittance based on the underwater scattering coefficient and the distance from the calibration disk to the camera lens, and then correcting the pixel value of each pixel using the transmittance and background scattered light intensity. The multi-scale denoising includes non-local mean filtering, distortion correction, and contrast enhancement.

[0025] Secondly, the present invention provides an ice thickness measuring device for ice storage coils, used to perform the aforementioned ice thickness measuring method for ice storage coils. The ice thickness measuring device includes a calibration plate, an image acquisition module, and a processor. The calibration plate is concentrically fixed to one straight section of the ice storage coil being measured. The image acquisition module is fixed to the inner wall of the ice storage tank. The image acquisition module includes a camera and a supplementary light. The supplementary light is a ring light, fixed around the outer periphery of the camera and facing the calibration plate. The supplementary light is a polarized supplementary light.

[0026] Preferably, the saturation and brightness of the color in each stripe area are both 100%, and the hue... Where n is the index of the striped region from the inside out, and N is the number of striped regions.

[0027] The present invention has the following beneficial effects.

[0028] 1. This invention fixes a calibration plate with multiple striped areas on an ice storage pan. By recognizing the state of ice coverage in each striped area through image recognition, the ice thickness is dynamically estimated, realizing non-contact stepless and accurate measurement of ice thickness in the ice storage pan. This breaks through the dependence of traditional sensors on contact measurement and improves deployment efficiency and environmental adaptability.

[0029] 2. This invention improves the accuracy of ice thickness detection through phased ice thickness analysis. Specifically, it uses the segmentation and brightness of striped regions of different colors to determine the ice layer that is completely covered by ice. Then, for the partially covered striped regions, based on the characteristic that the ice-covered areas are more blurred than the uncovered areas, it uses the comprehensive edge intensity that integrates gradient edge intensity and phase consistency edge intensity to calculate the coverage ratio.

[0030] 3. This invention innovatively integrates nonlocal mean denoising, wavelet threshold denoising, and block histogram equalization to specifically solve the problems of image blurring and contrast reduction caused by underwater turbidity and low temperature frost, thereby improving the accuracy, robustness, and adaptability of ice thickness detection in ice storage coils. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall structure of Embodiment 1 of the present invention.

[0032] Figure 2 This is a schematic diagram of the calibration disk in Embodiment 1 of the present invention.

[0033] Figure 3 This is a flowchart of Embodiment 2 of the present invention. Detailed Implementation

[0034] The present invention will be further described below with reference to the accompanying drawings.

[0035] Example 1

[0036] like Figure 1 As shown, a vision-based ice thickness measurement device for ice storage coils includes a calibration plate 1, an image acquisition module, and a processor 2.

[0037] The calibration disk 1 is fixed near the end of one of the straight pipe sections of the ice storage coil 5 being tested. The calibration disk 1 is fan-shaped, specifically semi-circular in this embodiment. The inner semi-circular edge of the calibration disk 1 is fixed to the corresponding straight pipe section of the ice storage coil 5. The center of the calibration disk 1 coincides with the axis of the straight pipe section of the ice storage coil 5.

[0038] The image acquisition module is fixed to the inner wall of the ice storage tank. The image acquisition module is positioned between the area where the ice storage coil 5 is located and the inner wall of the ice storage tank. The area where the image acquisition module is located will not freeze during the entire ice storage process. The image acquisition module is positioned directly opposite the calibration disk 1.

[0039] The image acquisition module is used to capture images of the ice ring and calibration disk 1. The processor 2 obtains the thickness of the ice ring by detecting the radial position of the outer edge of the ice ring on the calibration disk 1 in the image.

[0040] In some embodiments, the calibration disk 1 has multiple striped regions on its side. Each striped region is a concentric fan-shaped ring, arranged sequentially from the inside out. Different striped regions have different colors, which helps the image acquisition module determine the area where the outer edge of the ice ring is located. In two adjacent striped regions, the outer diameter of the inner striped region is equal to the inner diameter of the outer striped region. In some further embodiments, the width of the striped region (i.e., the difference between the inner and outer radii) is 5 mm.

[0041] like Figure 2As shown, the colors of each stripe region should be significantly different. In this embodiment, each stripe region is set in the HSV color space. The three primary colors of the HSV color space are red (H∈[0,10]∪[350,360], S>0.8, V>0.7), yellow (H∈[20,120], S>0.7, V>0.6), and cyan (H∈[180,200], S>0.65, V>0.5). The color stability under low temperature conditions is verified by the CIELAB color difference formula (ΔE*<5). In this embodiment, there are a total of six striped areas, and the colors of the six striped areas are pure red with an HSV value of (0.000°, 100%, 100%), pure yellow with an HSV value of (60°, 100%, 100%), pure green with an HSV value of (120°, 100%, 100%), pure cyan with an HSV value of (180°, 100%, 100%), pure blue with an HSV value of (240°, 100%, 100%), and pure purple with an HSV value of (300°, 100%, 100%).

[0042] In some embodiments, the calibration disk 1 is made of a low-temperature resistant material capable of withstanding -30°C, such as high-density polyethylene.

[0043] In some embodiments, the inner edge radius of the calibration disk 1 is equal to the outer radius of the ice storage coil 5. The difference between the inner and outer edge radii of the calibration disk 1 is greater than the maximum ice thickness of the ice storage coil 5. In some embodiments, the maximum ice thickness of the ice storage coil 5 is 50 mm; the difference between the inner and outer diameters of the calibration disk 1 should be the maximum ice thickness of the coil plus some margin, for example: if the maximum ice thickness of the ice storage coil 5 is 50 mm, the difference between the inner and outer diameters of the calibration disk 1 is set to 60 mm.

[0044] In some embodiments, the outer surface of the calibration disk 1 is provided with a hydrophobic coating. In some further embodiments, the hydrophobic coating is made of fluorosilane nanomaterials with a contact angle of 107° to 117° and a sliding angle of <8°. The hydrophobic coating helps to reduce ice crystal adhesion.

[0045] The image acquisition module includes a camera 3 and a fill light 4. The fill light 4 is a ring light, which is fixed around the outer periphery of the camera 3 and faces the calibration disk 1.

[0046] In some embodiments, the camera 3 is an industrial-grade waterproof and low-temperature resistant camera 3, which supports long-term underwater operation at -30℃ and has IP68 protection.

[0047] In some embodiments, the supplementary light 4 is a ring-polarized supplementary light 4, which can reduce water surface reflection and improve underwater imaging clarity through polarization separation technology.

[0048] In some embodiments, the processor 2 is located outside the ice storage tank and is connected to the image acquisition module via a cable.

[0049] Example 2

[0050] like Figure 3 As shown, a method for measuring ice thickness in an ice storage coil 5 based on visual perception is used, employing the ice thickness measuring device for the ice storage coil 5 provided in Example 1. The method includes the following steps:

[0051] Step 1: Image Acquisition and Preprocessing

[0052] The fill light 4 is turned on, and the camera 3 continuously captures images in the RGB space and uploads them to the processor 2. The processor 2 preprocesses the images.

[0053] To avoid severe color distortion and low contrast caused by turbid water, the following pretreatment process is provided:

[0054] (1) Underwater scattering model correction:

[0055] Based on the Jaffe-McGlamery scattering model, the expression for transmittance t(x) is established as follows:

[0056] t(x)=e −βd(x)

[0057] Where β is the underwater scattering coefficient; d(x) is the distance from calibration disk 1 to the lens of camera 3.

[0058] Based on the transmittance t(x), background scattered light interference is removed, and the true pixel value J(x) of each pixel is obtained:

[0059]

[0060] Where I(x) is the pixel value received by camera 3, and A is the background scattered light intensity in RGB space. The value is determined experimentally. In this embodiment, A=(0.20, 0.45, 0.60).

[0061] (2) Color correction:

[0062] For images corrected using an underwater scattering model, an underwater white balance algorithm is introduced to compensate for color cast based on the difference between the red and green channels:

[0063]

[0064] in, The red channel pixel value is the supplemented value; k is the compensation coefficient; ε is the preset parameter used to prevent division by zero; R, G, and B are the red, green, and blue channel pixel values ​​corrected by the underwater scattering model, respectively.

[0065] (3) Multi-scale denoising:

[0066] The color-corrected image is then subjected to nonlocal mean filtering, distortion correction, and contrast enhancement in sequence.

[0067] 1) Nonlocal mean filtering: The parameters are set as follows: search window is 21×21, similarity window is 5×5, smoothing parameter h=1.2σ; σ is the standard value of noise. The smoothing parameter h is used to suppress Gaussian noise.

[0068] 2) Lower-level thresholding denoising: Using the db4 wavelet basis function, 3-level decomposition, and soft thresholding. Suppress underwater scattering noise; This represents the total number of pixels.

[0069] 3) The distortion correction process is as follows: A mapping table is generated based on Zhang Zhengyou's calibration method to compensate for distortion caused by lens material shrinkage at low temperatures, and a lens temperature compensation model is established simultaneously. The lens temperature compensation model is used to compensate for changes in focal length at low temperatures. In this embodiment, the temperature change range ΔT = -30℃, and the focal length change rate Δ(f / f0) = -0.08%.

[0070] 4) Contrast Enhancement: The image is processed using a block histogram equalization method to improve the visibility of ice layer edges. In this block histogram equalization method, the block size is set to 8×8, and the contrast ratio (CL) is set to 3.

[0071] Step 2, Color Segmentation:

[0072] In this step, the preprocessed image is further processed in HSV space. When there is no ice on the disk, a dynamic color threshold is set to segment the stripe regions of each color. The dynamic thresholding algorithm overcomes the limitations of a fixed threshold through a color temperature compensation model, as follows:

[0073]

[0074] Where, ΔH(T) = 0.5° / ℃ × ΔT; , These represent the hue of a pixel before and after color temperature compensation. , These represent the saturation of the pixel before and after color temperature compensation. , These represent the brightness of pixels before and after color temperature compensation; by compensating for color shift caused by low temperature, the regions of different stripe areas are segmented. , where n represents the ordinal number of the striped region. After using color offset correction, the misclassification rate of striped region segmentation is reduced.

[0075] Step 3: Coverage Status Determination

[0076] (1) Constructing an improved Canny operator:

[0077] Use Gaussian filtering function Smooth image noise and enhance edge continuity; parameter: σ=1.5.

[0078]

[0079] in, These are the smoothed pixel values; Coordinates on the image after color segmentation The pixel value at that location; σ is the Gaussian filter function. The parameter is 1.5.

[0080] The Sobel operator is used to calculate the magnitude of the brightness gradient in the smoothed image. and brightness gradient direction :

[0081]

[0082]

[0083] (2) Dual threshold setting:

[0084]

[0085]

[0086] in, , These are the upper threshold and the lower threshold, respectively. This represents the maximum value of the brightness gradient magnitude for all pixels. In this embodiment, the ratio of the lower threshold to the upper threshold is 1:3.

[0087] (3) Preliminary classification of pixel types:

[0088] For each pixel in the smoothed image, the following judgment is performed:

[0089] ①. If the brightness gradient magnitude of a pixel is Greater than the upper threshold If the pixel is a strong edge, it is classified as a type of undetermined pixel.

[0090] ②. If the brightness gradient magnitude of a pixel is In [ , Within the interval, i.e., gradient magnitude Greater than or equal to the lower threshold If the pixel is not found to be a weak edge, it will be classified as a second-class undetermined pixel and will require subsequent orientation consistency verification.

[0091] ③. If the brightness gradient magnitude of a pixel Less than the lower threshold If the pixel is not an edge pixel, it is directly deleted and not considered a boundary pixel.

[0092] (4) Calculation of gradient edge intensity of undetermined pixels:

[0093] Gradient edge strength of a class of undetermined pixels The expression is:

[0094]

[0095] The orientation consistency verification of two types of undetermined pixels is performed as follows:

[0096] For weak edge pixels, an orientation consistency verification is introduced for the eight adjacent pixels. The verification relationship is as follows:

[0097]

[0098] in, Calculated as an average; Let represent the direction of the brightness gradient of the i-th neighboring pixel of the weak edge pixel; i = 1, 2, ..., 8; The direction of the brightness gradient for weak edge pixels; The orientation angle is set to 22.5° in this embodiment.

[0099] If the verification formula is satisfied, it means that the weak edge pixel has passed the orientation consistency verification. If the verification formula is not satisfied, it means that the weak edge pixel has failed the orientation consistency verification.

[0100] After completing the orientation consistency verification, the gradient edge intensity is calculated for all undetermined pixels of type II. as follows:

[0101]

[0102] (5) Phase Congruency (PC) detection of undetermined pixels:

[0103] For two types of undetermined pixels, calculate the phase-consistency edge intensity in the frequency domain. :

[0104]

[0105] Among them, ϕ nFor local frequency phase, It represents the average phase of all pixels in the image.

[0106] Phase coherence edge strength The larger the value, the more prominent the pixel is as an edge.

[0107] (6) Joint test:

[0108] Gradient edge strength for each undetermined pixel and phase consistency edge strength Normalization was performed on each of them:

[0109]

[0110] in, , These are the normalized gradient edge strength and the phase-consistent edge strength, respectively.

[0111] Weighted calculation of the overall edge strength of each undetermined pixel as follows:

[0112]

[0113] Where w1 and w2 are the weights of gradient edge strength and phase consistency edge strength, respectively; they satisfy w1+w2=1, and the specific values ​​are obtained experimentally.

[0114] (7) Multi-feature fusion judgment:

[0115] ①. Brightness Contrast: Calculate the average brightness difference between each stripe area and the background area. ; , These are the average brightness values ​​of the nth stripe region and the background region in HSV space, respectively. The background region refers to the area outside calibration disk 1 in the image. If If , then the nth stripe region is considered not to be completely covered by ice. If the nth stripe region is completely covered by ice, then the nth stripe region is considered to be completely covered by ice. Based on the judgment result, three intervals are formed: the innermost completely covered interval, the middle partially covered interval occupying one stripe region, and the outermost uncovered interval. The scaling factor is the scaling factor corresponding to the nth stripe region; scaling factor Experimental measurements show that the value range is generally 0.5 to 1. This represents the maximum brightness of the image.

[0116] In this embodiment, all stripe regions within the outermost stripe region that is determined to be completely covered by ice constitute the fully covered interval; the adjacent stripe regions outside the outermost stripe region that is determined to be completely covered by ice constitute the covered interval; and the remaining stripe regions constitute the uncovered interval.

[0117] ②. Coverage ratio calculation:

[0118] The striped area that is part of the complete ice cover region has the following ice cover percentage. .

[0119] The striped areas that belong to the uncovered region have the following ice coverage percentage. .

[0120] For striped regions that belong to the partially covered area, calculate the sum of the edge intensities of that striped region. as follows:

[0121]

[0122] in, The striped area is partially covered by ice.

[0123] Calculate the ice coverage percentage of the striped areas partially covered by ice. as follows:

[0124]

[0125] Where k is the brightness correction coefficient, which is determined experimentally and takes a negative value with a small absolute value. The baseline value for comprehensive edge strength was obtained through experiments before freezing. The term is a brightness correction term, whose function is to overcome the limitations of a single feature by fusing multiple features of edge intensity and brightness, thereby improving the accuracy and robustness of ice coverage ratio measurement in complex underwater environments. It not only compensates for the uncertainty in edge intensity measurement but also adapts to different types of ice, especially under conditions of poor underwater image quality, significantly improving the reliability of the system.

[0126] The principle of using comprehensive edge strength to detect the ice coverage ratio in this embodiment is that the area covered by ice in the image is more blurred and has lower edge strength than the area not covered by ice.

[0127] The comprehensive edge strength benchmark value The acquisition process is as follows:

[0128] While the water is not frozen, images of calibration disk 1 are acquired and color segmented, and the comprehensive edge intensity of each pixel is calculated according to steps (2) to (6) of step three. For each stripe area The sum of the comprehensive edge intensities in the uniced state within each stripe region is calculated as the benchmark value of the comprehensive edge intensities for each stripe region. as follows:

[0129]

[0130] Where n represents the ordinal number of the striped region, with values ​​of 1, 2, 3, 4, 5, 6.

[0131] The ice coverage ratio covering different stripes was calculated using the method described above. (0~1), combined with color offset to correct misjudgment.

[0132] In this step, to address the issues of high noise levels in underwater images and the susceptibility of traditional Canny images to isolated noise interference, a directional consistency constraint (gradient direction difference between adjacent pixels ≤ 22.5°) is introduced, and PC results are weighted together. This serves two purposes: first, to suppress noise edges: the directional consistency constraint reduces isolated noise interference; and second, to enhance edge continuity: PC supplements high-frequency details and reduces the ice layer boundary breakage rate.

[0133] Furthermore, this step, through brightness contrast ΔI and edge intensity weighted fusion, combined with dynamic threshold segmentation and morphological optimization, can accurately extract the ice-covered region M_covered and provide a reliable coverage ratio for thickness calculation. This method exhibits strong robustness in low-temperature underwater environments, especially in dealing with non-uniform ice cover and frost interference.

[0134] Step 4: Thickness Calculation

[0135] The radial pixel radius of the covered area is calculated using the coverage ratio and the upper and lower boundary radii of the stripes:

[0136]

[0137] in, The pixel thickness of the ice layer. , The lower and upper boundary pixel radii of the nth stripe are obtained by capturing the image and performing color segmentation before freezing.

[0138] The actual thickness D of the ice layer is calculated as follows:

[0139]

[0140] in, L(·) is the mapping function between pixel radius and actual radius.

[0141] In this embodiment, the calibration and standardization process of the camera 3 is as follows:

[0142] (1) Internal parameter calibration: Use a 9×6 checkerboard calibration plate (grid spacing 20mm) to obtain focal length and distortion coefficient.

[0143] (2) Spatial mapping calibration:

[0144] The calibration disk 1 was photographed in an ice-free state. The linear relationship L(r) between the pixel length r and the actual length L(r) was fitted, and a piecewise cubic spline interpolation model was established:

[0145]

[0146] Where, r t The critical radius is used to determine the segmentation point through cross-validation; a1, b1, c1, and d1 are four fitting coefficients when the pixel length r is less than the critical radius; a2, b2, c2, and d2 are four fitting coefficients when the pixel length r is greater than or equal to the critical radius.

[0147] The reasons for choosing the cubic spline interpolation model are as follows:

[0148] 1) Smoothness: The second derivative is continuous, avoiding abrupt changes in linear interpolation.

[0149] 2) Flexibility: Adaptable to nonlinear distortion of calibration disk 1 (such as larger edge distortion).

[0150] 3) Computational efficiency: Compared to higher-order polynomials, it has lower computational complexity (response time < 5s).

[0151] (3) The effect of the critical radius rt

[0152] Necessity of piecewise fitting: Central region of calibration disk 1 (r) <r t Less affected by lens distortion, using a low curvature (small a1, b1) polynomial can avoid overfitting while maintaining smoothness; edge regions (r ≥ r t The distortion is significant, requiring high curvature (large a2 and b2) to accurately fit the complex bending caused by the distortion.

[0153] The critical radius is determined by selecting r through cross-validation. t This minimizes the overall fitting error (RMSE < 0.3 mm).

[0154] (4) Periodic calibration:

[0155] Initial calibration acquires the camera's three intrinsic parameters and spatial mapping relationship. Periodic online calibration compensates for parameter drift during long-term operation. Triggering conditions: performed monthly or upon detection of abnormal thickness fluctuations. Once the triggering conditions are met, a spatial mapping calibration begins.

[0156] In this embodiment, error optimization is achieved through the following measures:

[0157] (1) Reduce interference from transparent ice layer: polarized light source separates surface reflected light, reducing color distortion caused by transmission.

[0158] (2) Vibration suppression: Hardware vibration reduction design + short exposure time (1 / 2000s) reduces motion blur.

[0159] (3) Robustness enhancement: multi-frame fusion voting judgment, abnormal thickness threshold alarm, and incremental learning to update the classifier.

[0160] (4) Calibration disk 1 design: The transparent acrylic material integrates radial fluorescent stripes (HSV spatial difference is large), and the surface is coated with a hydrophobic coating to reduce ice layer adhesion interference.

[0161] (5) Multi-algorithm image processing framework: integrates non-local mean denoising, wavelet threshold denoising, distortion correction and CLAHE enhancement to improve the quality of underwater images.

[0162] (6) Dynamic calibration mechanism: Combining initial calibration with periodic online calibration, it compensates for the impact of environmental changes on camera parameters and spatial mapping.

[0163] (7) Thickness calculation model: Based on geometric mapping and linear interpolation, the accurate conversion of ice layer coverage to thickness is realized.

Claims

1. A method for measuring ice thickness in ice storage coils based on visual perception, characterized in that, include: A circular or fan-shaped calibration disk is provided on the ice storage coil; the calibration disk has multiple concentric striped areas of different colors arranged radially. The camera is used to acquire images of the calibration disk outside the icing area; each stripe region is extracted from the image by color segmentation. Extracting the brightness gradient magnitude of pixels and brightness gradient direction By using two thresholds of different sizes, first-class and second-class undetermined pixels are extracted from all stripe regions; based on the brightness gradient direction... Perform orientation consistency verification on two types of undetermined pixels; The process of extracting Class I and Class II undetermined pixels is as follows: Set different lower and upper threshold values; classify pixels with brightness gradient magnitudes greater than the upper threshold as Class I undetermined pixels; and then... Pixels between the lower threshold and the upper threshold are classified as second-class undetermined pixels. Based on the brightness gradient magnitude and the orientation consistency verification results of the two types of undetermined pixels, the gradient edge intensity of each undetermined pixel is obtained; the phase consistency edge intensity of the undetermined pixel is calculated; the gradient edge intensity and the phase consistency edge intensity are normalized and then weighted and summed to obtain the comprehensive edge intensity. Based on the brightness change of the calibration disk in the image after it is covered by ice, it is determined whether each stripe area is completely covered by ice. Each stripe area is divided into the inner completely covered area, the middle partially covered area occupying one stripe area, and the outermost uncovered area. For partially covered areas, the ice coverage ratio of the partially covered area is obtained by combining the comprehensive edge intensity of each pixel within the corresponding stripe area with the comprehensive edge intensity benchmark value; the ice thickness is obtained based on the inner and outer radii of the fully covered area and the ice coverage ratio of the partially covered area.

2. The method for measuring ice thickness in an ice storage coil according to claim 1, characterized in that: The image extracts the brightness gradient magnitude. and brightness gradient direction Gaussian filtering is performed beforehand.

3. The method for measuring ice thickness in an ice storage coil according to claim 1, characterized in that: The process of verifying directional consistency is as follows: extract the brightness gradient direction of eight adjacent pixels of the two types of undetermined pixels. Establish verification relation If the verification relation is satisfied, it means that the two types of undetermined pixels have passed the orientation consistency verification; if the verification relation is not satisfied, it means that the two types of undetermined pixels have failed the orientation consistency verification; where, Let represent the brightness gradient direction of the i-th neighboring pixel of the second type of undetermined pixel; i=1,2,...,8; The brightness gradient direction of the second type of undetermined pixel; This is a constraint value for the direction angle.

4. The method for measuring ice thickness in an ice storage coil according to claim 1, characterized in that: The gradient edge intensity of the first type of undetermined pixels The gradient edge intensity expression for the two types of undetermined pixels is: ;in, This represents the brightness gradient magnitude.

5. The method for measuring ice thickness in an ice storage coil according to claim 1, characterized in that: The expression for the phase coherence edge strength is: Among them, ϕ n For local frequency phase, It represents the average phase of all pixels in the image.

6. The method for measuring ice thickness in an ice storage coil according to claim 1, characterized in that: Ice cover percentage in partially covered areas The expression is as follows: in, It is the sum of the overall edge intensities of all pixels within the partially covered area; k is the brightness correction coefficient; To obtain the baseline value for comprehensive edge intensity, the comprehensive edge intensity is extracted from the corresponding stripe regions on the unfrozen image and summed. This represents the difference between the average brightness of the corresponding striped area and the average brightness of the background. This represents the maximum brightness value in the image.

7. The method for measuring ice thickness in an ice storage coil according to claim 1, characterized in that: The images acquired by the camera are preprocessed; the preprocessing process includes underwater scattering correction, color correction and multi-scale denoising; the underwater scattering correction process is as follows: the transmittance is calculated based on the underwater scattering coefficient and the distance from the calibration disk to the camera lens, and the pixel value of each pixel is corrected using the transmittance and the intensity of background scattered light; the multi-scale denoising includes nonlocal mean filtering, distortion correction and contrast enhancement.

8. A device for measuring ice thickness in an ice storage coil, characterized in that: Used to perform the ice thickness measurement method for ice storage coils as described in claim 1; the ice thickness measurement device for ice storage coils includes a calibration plate (1), an image acquisition module and a processor (2); the calibration plate (1) is concentrically fixed on one of the straight pipe sections of the ice storage coil (5) to be measured; the image acquisition module is fixed on the inner side wall of the ice storage tank; the image acquisition module includes a camera (3) and a supplementary light (4); the supplementary light (4) is a ring light, which is fixed around the outer periphery of the camera (3) and faces the calibration plate (1); the supplementary light (4) is a polarized supplementary light.

9. The ice thickness measuring device for an ice storage coil according to claim 8, characterized in that: The saturation and brightness of the colors in each striped area are both 100%, and the hue... Where n is the index of the striped region from the inside out, and N is the number of striped regions.

Images