A solid-liquid gradual interface liquid level visual measurement system and method

Abstract

The application discloses a kind of solid-liquid gradual interface liquid level visual measurement system and method, belong to machine vision measurement and image processing technical field.The application obtains clear glue liquid area, solid-liquid gradual layer area and floating dregs area by image acquisition, polarized illumination, image pre-processing, image fusion and semantic segmentation;Subsequently, based on solid-liquid gradual layer area, the maximum connected domain is extracted, the boundary candidate points of upper interface and lower interface are obtained respectively, and the stable upper interface measurement point and lower interface measurement point are determined by eliminating abnormal points and smoothing constraint;Finally, combined with the zero line and camera calibration result, the interface pixel position is converted into actual liquid level height.The application can determine the main extraction boundary of clear glue liquid, the subsequent separation and recovery range of solid-liquid gradual layer and the filtering boundary of floating dregs simultaneously, improve the stability and accuracy of solid-liquid gradual interface liquid level measurement and material layering in natural macromolecule production process, and help intelligent separation and collection of production material.

Description

Technical Field

[0001] This invention relates to a visual measurement system and method for liquid level at a solid-liquid gradient interface, which is particularly suitable for visual measurement scenarios in the production or extraction of natural polymers where floating residue, solid-liquid gradient layer and clear liquid coexist, and the upper and lower interfaces have low contrast, strong noise interference, and the boundaries have a gradual transition. It belongs to the fields of machine vision measurement, liquid level detection and image processing technology. Background Technology

[0002] In the production or extraction of natural polymers, after sedimentation, stratification, or settling, the material system typically forms zones with different physical states, from top to bottom: floating sludge, a solid-liquid transition layer, and a clear colloid. The clear colloid is the primary extraction target, while the floating sludge usually needs to be filtered or removed. The solid-liquid transition layer, located between the two, still contains some clear colloid and other usable components, and therefore should not be directly treated as waste. Instead, it should be identified and measured as part of the subsequent separation and recovery process. In actual production, the produced material is placed in a container, and the clear colloid and the solid-liquid transition layer are released and collected separately through valves at the bottom of the container. During the release process, manual observation is required through an observation window on the container. Once the clear colloid has been released, the valve must be closed immediately, and a collection container for the solid-liquid transition layer must be prepared before opening the valve again to collect it. This manual separation and collection method increases manpower, and if the valve is not closed promptly, the solid-liquid transition layer or floating sludge may mix into the clear colloid or the solid-liquid transition layer. Therefore, those skilled in the art desire to achieve automatic detection of the three interfaces and thereby control the automatic opening and closing of valves to realize intelligent separation and collection. However, the presence of the solid-liquid gradient layer greatly increases the difficulty of detecting the three interfaces; the reasons are as follows.

[0003] The solid-liquid gradient layer refers to the transition layer between the floating sludge and the clearing liquid, whose grayscale, texture, turbidity, transmission characteristics, and scattering characteristics gradually change with spatial position. The upper interface of the solid-liquid gradient layer is adjacent to the floating sludge, and the lower interface is adjacent to the clearing liquid. Unlike ordinary clear liquid level lines, the upper and lower interfaces are often not single clear edges, but rather transitional boundaries with a certain thickness, local breaks, and unstable morphology. In actual visual inspection, the measurement of natural polymer materials faces the following problems: multiphase mixing state, floating sludge occlusion, turbidity of the solid-liquid gradient layer, and reflections from the measurement window and changes in external lighting. These problems lead to low contrast between the upper and lower interfaces in the image, high noise, discontinuous boundaries, and local missegmentation. If only single image threshold segmentation, simple edge detection, or direct extreme value search methods are used, it is easy to cause instability in the judgment of the clearing liquid extraction boundary, the subsequent separation and recovery range of the solid-liquid gradient layer, and the floating sludge filtration boundary.

[0004] Therefore, a visual measurement system and method is needed that can form a complete chain from image acquisition, imaging enhancement, region segmentation, extraction of upper and lower interfaces to actual height conversion, so as to achieve stable identification of the upper and lower interfaces of the solid-liquid gradient layer and provide positional basis for the extraction of clear liquid, subsequent separation and recovery of solid-liquid gradient layer and filtration of floating sludge. Summary of the Invention

[0005] To address the problems of unclear boundaries, low image contrast, strong noise interference, difficulty in stably distinguishing the upper and lower interfaces, and insufficient accuracy in interface height conversion during the production or extraction of natural polymers in existing technologies, the present invention aims to propose a visual measurement system and method for the liquid level of the solid-liquid gradient interface. By separately determining the upper and lower interfaces of the solid-liquid gradient layer, the present invention can provide stable and accurate visual measurement results for determining the extraction boundary of the clearing liquid, the subsequent separation and recovery range of the solid-liquid gradient layer, and the filtration boundary of the floating sludge.

[0006] The technical solution of this invention is implemented as follows:

[0007] A visual measurement system for liquid level at a solid-liquid gradient interface, wherein the solid-liquid gradient interface comprises the upper interface between the solid-liquid gradient layer and the upper floating scum region, and the lower interface between the solid-liquid gradient layer and the lower clearing liquid region; including...

[0008] The image acquisition unit uses a prism camera to simultaneously acquire visible light and near-infrared images of the area where the observation window is located, so as to ensure the consistency of different modal images in time and space.

[0009] The image fusion unit is used to fuse the visible light image and the near-infrared image acquired by the image acquisition unit to obtain a fused image;

[0010] The image segmentation unit is used to segment the clear colloid region, the solid-liquid gradient layer region, and the floating slag region from the fused image;

[0011] The camera calibration unit is used to establish the mapping relationship between image pixel coordinates and actual physical space;

[0012] The liquid level measurement unit is used to determine the upper and lower interfaces of the solid-liquid gradient layer region based on the solid-liquid gradient layer region segmented by the image segmentation unit, and to calculate the height of the upper interface, the height of the lower interface, and the range of the solid-liquid gradient layer height based on the predetermined measurement zero starting point and the mapping relationship provided by the camera calibration unit.

[0013] Furthermore, it also includes a polarization illumination unit, which is used to provide polarization illumination to the area where the observation window is located, so as to reduce interference from reflections from the measurement window, liquid surface reflections, and scattered light.

[0014] The polarization illumination unit includes an illumination source, a light source polarizer disposed at the light-emitting end of the illumination source, and a camera lens polarizer disposed at the front end of the prism camera lens; the illumination source is used to provide active illumination to the area where the observation window is located, and the illumination source consists of an independent visible light illumination source and a near-infrared illumination source, both of which share the light source polarizer; the light source polarizer and the camera lens polarizer are configured to have a cross-polarization relationship to reduce reflection interference in the image.

[0015] Furthermore, it also includes an image preprocessing unit, which performs noise reduction, brightness and contrast enhancement, and interface sharpening on the acquired image to improve the distinguishability between the clearing liquid area, the solid-liquid gradient layer area, and the floating slag area.

[0016] This invention also provides a visual measurement method for liquid level at a solid-liquid gradient interface, wherein the solid-liquid gradient interface is the upper interface between the solid-liquid gradient layer and the upper floating slag region, and the lower interface between the solid-liquid gradient layer and the lower clearing liquid region; the steps are as follows.

[0017] 1) Image acquisition: A prism camera is used to simultaneously acquire visible light and near-infrared images of the area where the observation window is located, so as to ensure the consistency of different modal images in time and space;

[0018] 2) Image preprocessing: Denoising, brightness and contrast enhancement and interface sharpening are performed on the visible light image and near-infrared image acquired in step 1) to improve the distinguishability between the clearing liquid area, the solid-liquid gradient layer area and the floating slag area.

[0019] 3) Image fusion: The preprocessed visible light image and near-infrared image are fused to obtain a fused image;

[0020] 4) Image segmentation: Segment the clear colloid region, solid-liquid gradient layer region, and floating slag region from the fused image;

[0021] 5) Liquid level measurement: Based on the solid-liquid gradient layer region obtained in step 4), determine the upper and lower interfaces of the solid-liquid gradient layer region respectively.

[0022] Further, in step 1), during image acquisition, active illumination is provided to the area where the observation window is located through an illumination source. The illumination source consists of an independent visible light illumination source and a near-infrared illumination source, and both have a shared light source polarizer at their light-emitting ends. A camera lens polarizer is provided at the front end of the prism camera lens. The light source polarizer and the camera lens polarizer are set to a cross-polarization relationship to reduce reflection interference in the image.

[0023] Furthermore, in step 2), the brightness and contrast enhancement process employs a linear enhancement method, expressed as:

[0024]

[0025] in, This represents the grayscale value of the original image at pixel coordinates. For the enhanced grayscale value, This is the contrast gain coefficient. This is the brightness compensation coefficient; through the... and The adjustment was made to increase the grayscale difference between the solid-liquid gradient interface area and the surrounding background, thereby enhancing the interface visibility;

[0026] In interface sharpening, an unsharpening mask is used, expressed as follows:

[0027]

[0028] in, The image after sharpening. This represents Gaussian blur operation. For Gaussian kernel parameters, This is a sharpening weighting coefficient; through this processing, edge features near the solid-liquid interface can be enhanced while suppressing local noise.

[0029] Furthermore, in step 3), image fusion employs a deep learning-based multimodal image fusion method to jointly represent the complementary features of visible light and near-infrared images, outputting a fused image for subsequent region segmentation and upper / lower interface measurement.

[0030] Further, in step 6), when measuring the liquid level, contour extraction is performed on the binary mask corresponding to the solid-liquid gradient layer region, and the contour with the largest area is selected as the target solid-liquid gradient layer region. Then, the largest connected region mask is formed by filling the largest contour, which is used as the position measurement area of ​​the upper and lower interfaces of the solid-liquid gradient layer.

[0031] On the maximum connected region mask, non-zero pixels are searched column by column. For each column, the topmost pixel is extracted as the upper interface candidate point and the bottommost pixel is extracted as the lower interface candidate point. The upper interface candidate point is used to represent the boundary candidate position adjacent to the solid-liquid gradient layer and the floating slag region, and the lower interface candidate point is used to represent the boundary candidate position adjacent to the solid-liquid gradient layer and the clearing liquid region. Thus, the upper interface candidate point set and the lower interface candidate point set are obtained.

[0032] Then, Hampel outlier removal is performed on the candidate point set on the upper interface and the candidate point set on the lower interface respectively; then, gradient, curvature and residual filtering based on Savitzky-Golay smoothing is performed on the removed candidate point set.

[0033] Finally, from the filtered candidate point sets for the upper and lower interfaces, the point with the lowest vertical coordinate is selected as the lowest point of the corresponding interface, and the position of the corresponding upper and lower interfaces is determined based on the lowest point.

[0034] Furthermore, it also includes a step of calibrating the prism camera described in step 1) to establish a mapping relationship between image pixel coordinates and actual physical space, and to calculate the upper interface height and lower interface height based on a predetermined measurement zero point and the established mapping relationship.

[0035] Further, the calculation process for the upper and lower interface heights is as follows: A horizontal reference line is set as a unified zero line in the acquired image, and this unified zero line is used as the measurement zero starting point; reference points with the same horizontal coordinates as the lowest points of the upper and lower interfaces are constructed on the unified zero line as zero line reference points; the mapping relationship established by camera calibration is used to map the lowest points of the upper and lower interfaces and their respective zero line reference points from pixel coordinates to world coordinates; the Euclidean distance between each lowest point and the corresponding zero line reference point is calculated using the following formula to obtain the upper and lower interface heights H, and the solid-liquid gradient layer height range is determined based on the upper and lower interface heights.

[0036]

[0037] in, The world coordinates corresponding to the zero-line reference point constructed for the lowest point of the upper or lower interface. The world coordinates corresponding to the lowest point of the upper or lower interface. This is the correction value.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. This invention improves the image distinguishability between the clearing liquid region, the solid-liquid gradient layer region, and the floating slag region under low contrast conditions by using polarized illumination, image preprocessing, and fusion of visible light and near-infrared image information, providing a reliable image basis for the stable segmentation of the upper and lower interfaces of the solid-liquid gradient layer.

[0040] 2. This invention does not merely detect a single liquid level line, but rather determines the upper and lower interfaces of the solid-liquid gradient layer separately. The lower interface is used to determine the main extraction boundary of the clearing solution, the upper interface is used to determine the filtration boundary of the floating sludge, and the area between the upper and lower interfaces is used to determine the subsequent separation and recovery range of the solid-liquid gradient layer. This method can simultaneously serve clearing solution extraction, recovery of usable components from the solid-liquid gradient layer, and filtration of floating sludge, improving the stability of liquid level measurement and material utilization in the production process of natural polymers.

[0041] 3. The measurement system of this invention is installed next to the container holding the materials on the production site, and can detect the material interface in real time during the production process. The measurement system is further linked with the valve at the bottom of the container, and automatically controls the opening or closing of the valve based on the detection results. This enables intelligent separation and collection of the clearing liquid, solid-liquid gradient layer and floating slag, which not only improves the stability and accuracy of separation, but also saves labor costs. Furthermore, by calculating the height of the upper and lower interfaces in real time and quantifying the height value, the valve opening can be reduced in advance when the height is less than the set value, and the valve can be completely closed when the height is zero. This avoids the drawbacks of lag and inaccurate separation caused by controlling the opening and closing of the valve solely through the interface line (which cannot be quantified). Attached Figure Description

[0042] Figure 1 - Schematic diagram of the principle of visual measurement of liquid level at solid-liquid gradient interface in this invention.

[0043] Figure 2 - A schematic diagram of the components related to image acquisition in this invention.

[0044] Figure 3 - A schematic diagram of the polarization illumination unit of this invention.

[0045] Figure 4 - A schematic diagram comparing the results of selecting the boundary of the solid-liquid gradient layer in this invention with existing methods. Detailed Implementation

[0046] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0047] This invention provides a visual measurement system and method for solid-liquid gradient interface liquid level, particularly suitable for layered measurement scenarios where floating sludge, a solid-liquid gradient layer, and a clear colloid coexist during the production or extraction of natural polymers. The solid-liquid gradient layer is a transition layer located between the floating sludge region and the clear colloid region; the upper interface of the solid-liquid gradient layer is the boundary adjacent to the floating sludge region, and the lower interface is the boundary adjacent to the clear colloid region. Specifically, the clear colloid region is used to determine the main extraction boundary, the solid-liquid gradient layer region is used to determine the subsequent separation and recovery range, and the floating sludge region is used to determine the filtration or rejection boundary.

[0048] See Figure 1 The visual measurement system of this invention mainly includes an image acquisition unit, a polarized illumination unit, an image preprocessing unit, an image fusion unit, an image segmentation unit, a camera calibration unit, and a liquid level measurement unit. The measurement focus of this invention is to determine the upper and lower interfaces of the solid-liquid gradient layer to obtain the main extraction boundary of the clear liquid, the subsequent separation and recovery range of the solid-liquid gradient layer, and the filtration boundary of the floating sludge.

[0049] The image acquisition unit and the polarization illumination unit together constitute the image acquisition section of this invention. For its specific structure, please refer to [link / reference needed]. Figure 2 In addition to the image acquisition unit 1, the system also includes a camera lens polarizer 2, an illumination source 3, and a light source polarizer 4. The image acquisition unit 1 is used to acquire images of the solid-liquid stratification observation area (the area where the observation window on the container is located). Specifically, a prism camera is used to simultaneously acquire visible light and near-infrared images of the area where the observation window is located. This synchronous acquisition method ensures the temporal consistency and spatial correspondence of images of different modalities, thereby reducing registration errors caused by time shifts, mechanical jitter, or parallax changes. The image acquisition unit 1 is positioned facing the solid-liquid stratification observation area to ensure that the areas containing floating slag, the solid-liquid gradient layer, and the clearing liquid are within the effective field of view of the camera. The camera lens polarizer 2 is located at the front end of the lens of the image acquisition unit 1. The illumination source 3 is located on one side of the solid-liquid stratification observation area to provide active illumination; the light source polarizer 4 is located in front of the light-emitting end of the illumination source 3 to polarize the illumination light. To reduce the impact of reflections from the measurement window, liquid surface reflections, and scattered light on the identification of the upper and lower interfaces, this invention employs a combined visible and near-infrared illumination method. Specifically, the illumination source consists of both a visible light source and a near-infrared source, both sharing a common polarizer. Therefore, the polarizer 4 is preferably a polarizer covering the entire wavelength range of visible and near-infrared light. These components together constitute a polarized imaging structure suitable for measuring the liquid level at solid-liquid transition interfaces. Polarized illumination improves the image resolution between floating scum, the solid-liquid transition layer, and the clearing liquid.

[0050] like Figure 3 As shown, the illumination light emitted by the illumination source 3 is polarized into polarized light with a predetermined polarization direction after passing through the light source polarizer 4, and then illuminates the glass observation window area. The reflected light, transmitted light, and scattered light from the observation area of ​​the glass observation window are polarized and filtered by the camera lens polarizer 2 before entering the image acquisition unit 1. The light source polarizer 4 and the camera lens polarizer 2 are set to a cross-polarization or approximately cross-polarization relationship to reduce the proportion of reflected light entering the lens, thereby reducing the interference of high-brightness reflections and glare on the imaging of the upper and lower interfaces.

[0051] In this embodiment, the image acquisition unit 1, camera lens polarizer 2, illumination source 3, and light source polarizer 4 are all fixedly installed using mounting brackets or optical fasteners to maintain a stable relative positional relationship between the components. Figures 2 to 3 As shown in the diagram, this invention can provide a high-quality image basis for subsequent segmentation of the clearing liquid region, the solid-liquid gradient layer region, and the floating slag region.

[0052] During the measurement process, the image acquisition unit 1 first simultaneously acquires visible light and near-infrared images of the solid-liquid stratification observation area. Let the acquired visible light image be... Near-infrared image is .

[0053] To improve the accuracy of subsequent image fusion and segmentation, an image preprocessing unit is needed to preprocess the original image. The preprocessing includes denoising, brightness / contrast enhancement, and interface sharpening to improve image quality near the upper and lower interfaces. In brightness / contrast enhancement, a linear enhancement method can be used, with the following expression:

[0054]

[0055] in, This represents the grayscale value of the original image at pixel coordinates. For the enhanced grayscale value, This is the contrast gain coefficient. This is the brightness compensation coefficient. (Based on...) and Adjustments can increase the grayscale difference between the solid-liquid gradient interface area and the surrounding background, thereby enhancing interface visibility.

[0056] In interface sharpening, an unsharpening mask can be used, and its expression is:

[0057]

[0058] in, The image after sharpening. This represents Gaussian blur operation. For Gaussian kernel parameters, These are the sharpening weighting coefficients. This processing can enhance edge features near the solid-liquid interface while suppressing local noise.

[0059] The preprocessed visible light image and near-infrared image are denoted as follows: and The image is then input into the image fusion unit. The image fusion unit fuses the two images to obtain a fused image more suitable for interface recognition. The image fusion unit employs a deep learning-based multimodal image fusion model, and its fusion process can be represented as follows:

[0060]

[0061] in, Indicates a fused image. Represents the fusion model. This represents the parameters of the fusion model. Through fusion processing, the detailed texture information of the visible light image and the anti-interference information of the near-infrared image can be comprehensively utilized, thereby improving the characterization ability of the upper and lower interfaces of the solid-liquid gradient layer under low contrast conditions.

[0062] The fused image is input to the image segmentation unit for processing the fused image. Semantic segmentation is performed to extract the clear gel region, the solid-liquid gradient layer region, and the floating slag region. The segmentation process can be represented as follows:

[0063]

[0064] in, Indicates the segmentation result. Represents the segmentation model. The segmentation model parameters are represented. The segmentation results include at least labels for the clear liquid region, the solid-liquid gradient layer region, and the floating slag region. Since this invention mainly focuses on the accurate measurement of the solid-liquid gradient interface, the subsequent focus is on using the solid-liquid gradient layer region to locate the upper and lower interfaces.

[0065] Let the binary mask corresponding to the solid-liquid gradient layer region be... Then we have:

[0066]

[0067] in, This represents the set of pixels in the solid-liquid gradient layer region of the segmentation result.

[0068] The camera calibration unit is used to establish the mapping relationship between image pixel coordinates and actual physical space. It acquires images of the calibration board in multiple different poses and extracts corner information to solve for the camera's intrinsic, extrinsic, and distortion parameters. Let the homogeneous coordinates of a point in the world coordinate system be:

[0069]

[0070] Its corresponding homogeneous pixel coordinates are:

[0071]

[0072] Then the two satisfy the perspective projection relationship:

[0073]

[0074] in, As a scaling factor, The camera intrinsic parameter matrix, Let be a rotation matrix. This is the translation vector. Through calibration, a mapping relationship can be established between pixel coordinates and the actual physical coordinates of the measurement plane, denoted as:

[0075]

[0076] in, This represents the mapping function obtained from the calibration. This mapping relationship provides the basis for subsequent conversion of the actual dimensions of the upper and lower interface heights.

[0077] The liquid level measurement unit is mainly used to determine the interface position and calculate the liquid level height based on the segmentation results of the gradient layer region. Specifically, it first starts from the binary mask of the gradient layer region. Connected components are extracted. Since the segmentation results may contain small pseudo-regions caused by reflections, stains, local shadows, and missegmentation, only the largest connected component is retained as the target solid-liquid gradient layer region.

[0078] Let the set of connected components in the binary mask be . , No. The area of ​​a connected region is defined as:

[0079]

[0080] in, Represents connected components The number of pixels in the target connected region determines the area of ​​the largest target connected region. satisfy:

[0081]

[0082] Construct a target mask using the largest connected component. By extracting the largest connected component, the impact of local reflections, reflection artifacts, bubbles, stains, and small mis-segmented areas on the measurement of the upper and lower interfaces can be effectively reduced, retaining only the main area corresponding to the actual solid-liquid gradient layer, thus providing a basic area for the measurement of the upper and lower interfaces.

[0083] After obtaining the target mask Then, candidate boundary points are extracted column by column along the image column direction. Let the first... The set of ordinates of all pixels belonging to the target mask in the column is:

[0084]

[0085] Because the zero coordinate is located at the top left corner of the image in the image coordinate system design, and the ordinate arrow points downwards, the ordinate of each column has the smallest ordinate at the top boundary and the largest ordinate at the bottom boundary. Therefore, when... ( When referring to the empty set, the candidate points for the upper and lower boundaries of the column are defined as follows:

[0086]

[0087] Based on this, the candidate point set for the upper boundary and the candidate point set for the lower boundary can be constructed:

[0088]

[0089] Since the solid-liquid gradient interface usually appears as a band-shaped region of a certain thickness in the image rather than a single clear edge, the spatial distribution morphology of the solid-liquid gradient layer can be described more completely by extracting candidate points of the upper and lower boundaries column by column. Among them, the upper boundary corresponds to the upper interface of the solid-liquid gradient layer and the floating slag, and the lower boundary corresponds to the lower interface of the solid-liquid gradient layer and the clearing liquid.

[0090] like Figure 4 As shown, when abrupt changes or discontinuities occur in the segmentation of the solid-liquid graded layer, choosing a simple boundary-finding strategy can lead to issues such as… Figure 4 (a) shows the erroneous upper boundary segmentation measurement result. To improve the stability of the upper and lower boundary point extraction, this invention performs outlier removal processing on the boundary candidate point set. The first layer of outlier removal uses the Hampel filtering method. For the first outlier in a certain boundary sequence... There are points, at a length of [number] points. Calculating the local median within a sliding window And median absolute deviation:

[0091]

[0092] in, Further define the robust scaling estimator. for:

[0093]

[0094] in, For variable parameters, adjust the estimator A point is considered an outlier when the following condition is met:

[0095]

[0096] in, This is the threshold coefficient. Hampel filtering can effectively remove isolated outliers caused by reflections, burrs, blemishes, and local missegmentation.

[0097] After the first layer of outlier removal, the boundary point sequence is further filtered based on a smoothing curve. The Savitzky-Golay smoothing method is used to perform local polynomial fitting on the boundary sequence to obtain the smoothed boundary curve. Original boundary With smooth boundary The residual between them is defined as:

[0098]

[0099] Simultaneously calculate the first and second derivatives of the smooth curve:

[0100]

[0101] in, Indicates the local trend of change at the boundary. This represents the local curvature change of the boundary. Then, based on the robustness thresholds of the residuals, gradients, and curvature, points that do not conform to the trend of continuous boundaries are further eliminated to remove local inflection points, abrupt change points, and discontinuous boundary points, so that the remaining boundary point set is more consistent with the geometric characteristics of the real interface, thereby improving the continuity and stability of the upper and lower interface point sets.

[0102] As mentioned earlier regarding the coordinate system setup, a larger ordinate value in the image coordinate system indicates a lower position for the point. After outlier removal, the point with the largest ordinate value is selected from both the upper and lower boundary point sets and defined as the lowest point of that boundary.

[0103]

[0104]

[0105] When multiple points have the same maximum y-coordinate, any one of them can be chosen as the final minimum point. Based on the method described above, the following can be obtained: Figure 4 (b) shows the lowest point of the upper interface and the lowest point of the lower interface of the solid-liquid gradient layer.

[0106] To standardize the benchmark for liquid level measurement, this invention employs an image annotation method to determine a unified zero line. Specifically, the operator can set a horizontal reference line in the image display interface, with its ordinate marked as follows: For the determined lowest point of the upper interface or the lowest point of the lower interface... Construct a zero-line reference point on the unified zero line, whose x-coordinate is consistent with its:

[0107]

[0108] Subsequently, based on the pixel-physical space mapping relationship established by the camera calibration unit, the zero-line reference point was... and the lowest point of the interface These are mapped to the actual physical space and denoted as:

[0109]

[0110] The corresponding liquid level height It can be represented as:

[0111]

[0112] in, Represents Euclidean distance. This represents the system correction value. If the physical space coordinates are further written as... , The liquid level height can then be further expressed as:

[0113]

[0114] from Figure 1 As can be seen, the entire process of visual measurement in this invention can be divided into three stages. The first stage is multimodal imaging and preprocessing, which includes three stages: polarization illumination, image acquisition, and image preprocessing. The second stage is image fusion and segmentation, which mainly includes two stages: multimodal image fusion and liquid region segmentation. The third stage is interface localization and liquid level calculation, which involves maximum connected component extraction, boundary candidate point extraction, outlier point removal and curve smoothing, minimum point determination, and camera calibration, and finally outputs the liquid level height of the upper and lower interfaces.

[0115] In one specific implementation example, the complete operation flow of the present invention is as follows:

[0116] Visible light and near-infrared images of the solid-liquid layered observation area under polarized illumination are acquired simultaneously. The acquired images are then subjected to noise reduction, brightness enhancement, and interface sharpening in sequence. The visible light and near-infrared images are then input into an image fusion unit to obtain a fused image.

[0117] Semantic segmentation is performed on the fused image to extract the clear colloid region, the solid-liquid gradient layer region, and the floating slag region. Then, the maximum connected component is extracted from the solid-liquid gradient layer region, and candidate points for the upper and lower interfaces are extracted column by column along the horizontal direction of the image. Then, outliers are removed by Hampel filtering and smoothing curve constraints, and measurement points for the upper and lower interfaces are selected from the processed boundary points.

[0118] Based on the mapping relationship established by the calibrated unified zero line and camera calibration, the distance between the zero line reference point and the corresponding interface measurement point is converted into the actual height, thereby obtaining the main extraction boundary of the clearing liquid, the subsequent separation and recovery range of the solid-liquid gradient layer, and the filtration boundary of the floating sludge.

[0119] Through the above structure and steps, this invention forms a complete visual measurement chain from image acquisition, polarization suppression of reflection, image fusion, region segmentation, extraction of upper and lower interfaces to actual height calculation. Compared with methods that directly estimate liquid level based on single image threshold segmentation or simple edge detection, this invention can simultaneously distinguish between clear liquid, solid-liquid gradient layer, and floating sludge, and stably measure the upper and lower interfaces of the solid-liquid gradient layer, thereby improving the accuracy of liquid level measurement, material stratification judgment, and usable component recovery in the production process of natural polymers.

[0120] Finally, it should be noted that the above examples of the present invention are merely illustrative and not intended to limit the implementation of the invention. Although the applicant has described the present invention in detail with reference to preferred embodiments, those skilled in the art can make other variations and modifications based on the above description. It is impossible to exhaustively list all possible implementations here. All obvious variations or modifications derived from the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A visual measurement system for liquid level at a solid-liquid gradient interface, wherein the solid-liquid gradient interface is the upper interface between the solid-liquid gradient layer and the upper floating slag region, and the lower interface between the solid-liquid gradient layer and the lower clearing liquid region; characterized in that: include The image acquisition unit uses a prism camera to simultaneously acquire visible light and near-infrared images of the area where the observation window is located, so as to ensure the consistency of different modal images in time and space. The image fusion unit is used to fuse the visible light image and the near-infrared image acquired by the image acquisition unit to obtain a fused image; The image segmentation unit is used to segment the clear colloid region, the solid-liquid gradient layer region, and the floating slag region from the fused image; The camera calibration unit is used to establish the mapping relationship between image pixel coordinates and actual physical space; The liquid level measurement unit is used to determine the upper and lower interfaces of the solid-liquid gradient layer region based on the solid-liquid gradient layer region segmented by the image segmentation unit, and to calculate the height of the upper interface, the height of the lower interface, and the range of the solid-liquid gradient layer height based on the predetermined measurement zero starting point and the mapping relationship provided by the camera calibration unit.

2. The visual measurement system for liquid level at a solid-liquid gradient interface according to claim 1, characterized in that: It also includes a polarization illumination unit, which is used to provide polarization illumination to the area where the observation window is located, so as to reduce interference from reflections from the measurement window, liquid surface reflections, and scattered light. The polarization illumination unit includes an illumination source, a light source polarizer disposed at the light-emitting end of the illumination source, and a camera lens polarizer disposed at the front end of the prism camera lens; the illumination source is used to provide active illumination to the area where the observation window is located, and the illumination source consists of an independent visible light illumination source and a near-infrared illumination source, both of which share the light source polarizer; the light source polarizer and the camera lens polarizer are configured to have a cross-polarization relationship to reduce reflection interference in the image.

3. The visual measurement system for liquid level at a solid-liquid gradient interface according to claim 1, characterized in that: It also includes an image preprocessing unit, which performs noise reduction, brightness and contrast enhancement, and interface sharpening on the acquired image to improve the distinguishability between the clearing liquid area, the solid-liquid gradient layer area, and the floating slag area.

4. A method for visually measuring liquid level at a solid-liquid gradient interface, wherein the solid-liquid gradient interface is the upper interface between the solid-liquid gradient layer and the upper floating slag region, and the lower interface between the solid-liquid gradient layer and the lower clearing liquid region; characterized in that: The steps are as follows: 1) Image acquisition: A prism camera is used to simultaneously acquire visible light and near-infrared images of the area where the observation window is located, so as to ensure the consistency of different modal images in time and space; 2) Image preprocessing: Denoising, brightness and contrast enhancement and interface sharpening are performed on the visible light image and near-infrared image acquired in step 1) to improve the distinguishability between the clearing liquid area, the solid-liquid gradient layer area and the floating slag area. 3) Image fusion: The preprocessed visible light image and near-infrared image are fused to obtain a fused image; 4) Image segmentation: Segment the clear colloid region, solid-liquid gradient layer region, and floating slag region from the fused image; 5) Liquid level measurement: Based on the solid-liquid gradient layer region obtained in step 4), determine the upper and lower interfaces of the solid-liquid gradient layer region respectively.

5. The method for visually measuring liquid level at a solid-liquid gradient interface according to claim 4, characterized in that: Step 1) During image acquisition, active illumination is provided to the area where the observation window is located through an illumination source. The illumination source consists of an independent visible light illumination source and a near-infrared illumination source, and both have a common light source polarizer at their light-emitting ends. A camera lens polarizer is set at the front end of the prism camera lens. The light source polarizer and the camera lens polarizer are set to a cross-polarization relationship to reduce reflection interference in the image.

6. The method for visually measuring liquid level at a solid-liquid gradient interface according to claim 4, characterized in that: Step 2) In the brightness and contrast enhancement process, a linear enhancement method is adopted, and the expression is: in, This represents the grayscale value of the original image at pixel coordinates. For the enhanced grayscale value, This is the contrast gain coefficient. This is the brightness compensation coefficient; through the... and The adjustment was made to increase the grayscale difference between the solid-liquid gradient interface area and the surrounding background, thereby enhancing the interface visibility; In interface sharpening, an unsharpening mask is used, expressed as follows: in, The image after sharpening. This represents Gaussian blur operation. For Gaussian kernel parameters, This is a sharpening weighting coefficient; through this processing, edge features near the solid-liquid interface can be enhanced while suppressing local noise.

7. The method for visually measuring liquid level at a solid-liquid gradient interface according to claim 4, characterized in that: Step 3) Image fusion adopts a deep learning-based multimodal image fusion method to jointly represent the complementary features of visible light images and near-infrared images, and outputs a fused image for subsequent region segmentation and upper and lower interface measurement.

8. The method for visually measuring liquid level at a solid-liquid gradient interface according to claim 4, characterized in that: Step 6) When measuring the liquid level, perform contour extraction on the binary mask corresponding to the solid-liquid gradient layer region, select the contour with the largest area as the target solid-liquid gradient layer region, and then fill the largest connected region mask according to the largest contour to use as the position measurement area of ​​the upper and lower interfaces of the solid-liquid gradient layer. On the maximum connected region mask, non-zero pixels are searched column by column. For each column, the topmost pixel is extracted as the upper interface candidate point and the bottommost pixel is extracted as the lower interface candidate point. The upper interface candidate point is used to represent the boundary candidate position adjacent to the solid-liquid gradient layer and the floating slag region, and the lower interface candidate point is used to represent the boundary candidate position adjacent to the solid-liquid gradient layer and the clearing liquid region. Thus, the upper interface candidate point set and the lower interface candidate point set are obtained. Then, Hampel outlier removal is performed on the candidate point set on the upper interface and the candidate point set on the lower interface respectively; then, gradient, curvature and residual filtering based on Savitzky-Golay smoothing is performed on the removed candidate point set. Finally, from the filtered candidate point sets for the upper and lower interfaces, the point with the lowest vertical coordinate is selected as the lowest point of the corresponding interface, and the position of the corresponding upper and lower interfaces is determined based on the lowest point.

9. The method for visually measuring liquid level at a solid-liquid gradient interface according to claim 8, characterized in that: It also includes a step of calibrating the prism camera described in step 1) to establish a mapping relationship between image pixel coordinates and actual physical space, and to calculate the upper interface height and lower interface height based on a predetermined measurement zero point and the established mapping relationship.

10. The method for visually measuring liquid level at a solid-liquid gradient interface according to claim 9, characterized in that: The calculation process for the upper and lower interface heights is as follows: A horizontal reference line is set as a unified zero line in the acquired image, and this unified zero line is used as the measurement zero starting point; reference points with the same horizontal coordinates as the lowest points of the upper and lower interfaces are constructed on the unified zero line as zero line reference points; the mapping relationship established by camera calibration is used to map the lowest points of the upper and lower interfaces and their respective zero line reference points from pixel coordinates to world coordinates; the Euclidean distance between each lowest point and the corresponding zero line reference point is calculated using the following formula to obtain the upper and lower interface heights H, and the solid-liquid gradient layer height range is determined based on the upper and lower interface heights. in, The world coordinates corresponding to the zero-line reference point constructed for the lowest point of the upper or lower interface. The world coordinates corresponding to the lowest point of the upper or lower interface. This is the correction value.

Images