Real-time image monitoring method and system for a yucca extract separation process

Abstract

The present application belongs to the technical field of image processing, and particularly relates to a real-time image monitoring method and system for a yucca extract separation process, which comprises the following steps: obtaining a separation process image, obtaining uneven illumination according to pixel brightness, determining an adaptive filter size according to the uneven illumination, performing enhancement processing on the image to obtain an enhanced image, obtaining a separation tendency index according to local differences and texture features of the enhanced image, segmenting the image to obtain a connected region according to the separation tendency index, and obtaining a solid phase aggregation degree according to morphological features of the connected region. Finally, the end point of the separation is determined according to the change trend of the solid phase aggregation degree and the area variance in a time window. The uneven illumination problem in an industrial field is overcome by using an adaptive enhancement algorithm, and the end point of the separation is accurately determined in combination with morphological features, thereby improving the accuracy and robustness of the monitoring.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a real-time image monitoring method and system for the separation process of yucca extract. Background Technology

[0002] Solid-liquid separation is a crucial step in the production of yucca extract, as its effectiveness and endpoint determination directly impact product quality and production efficiency. During separation, the endpoint is reached when the sedimentation and aggregation of solid particles essentially cease, and the solid and liquid phases reach dynamic equilibrium. After reaching the endpoint, the product morphology no longer undergoes significant changes. Therefore, accurate determination of the separation endpoint is essential for monitoring the separation process and forms the basis for subsequent optimization of the extract separation process.

[0003] Real-time monitoring of this process using machine vision technology is an important means of replacing traditional manual observation and achieving automated control and optimization. However, visual monitoring faces many challenges in actual industrial production environments. The lighting conditions in industrial sites are usually complex and variable, often resulting in uneven lighting in the acquired images, such as localized overexposure or underexposure. This uneven lighting is the primary obstacle affecting subsequent solid and liquid phase feature extraction and analysis.

[0004] In related technologies, such as algorithms based on Retinex theory, image enhancement is often achieved by separating the illumination component and the reflection component. However, most of these technologies use a fixed set of filter scales. This fixed-scale approach cannot adaptively cope with the dynamic changes in local illumination differences in the image with spatial location. As a result, the contrast between the solid particles and the liquid background of yucca extract remains very low and the boundaries are blurred at certain separation stages. This makes it difficult to accurately distinguish between solid particles and liquid background under low-contrast conditions, which in turn leads to a lack of reliability in subsequent analysis of particle morphology and aggregation state, ultimately affecting the accurate determination of the separation endpoint. Summary of the Invention

[0005] To address the technical problems of poor image quality caused by uneven lighting in industrial settings, and the difficulty in accurately monitoring the separation endpoint due to low contrast and blurred boundaries between the solid and liquid phases, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a real-time image monitoring method for the separation process of yucca extract, comprising: decomposing the separation process image into a brightness image, a green-red component image, and a blue-yellow component image; obtaining the illumination unevenness of each pixel based on the difference between the brightness and the illumination estimate of each pixel in the separation process image, the standard deviation of the brightness of the pixel within a local window of the pixel, and the mean of the gradient magnitude of the pixel within the local window of the pixel; wherein the illumination estimate is obtained by Gaussian filtering the brightness image; obtaining the filter size of each pixel based on the illumination unevenness of each pixel in the separation process image and the mean of the illumination unevenness of all pixels; and adjusting the brightness image based on the filter size of each pixel. A light estimation map is obtained through filtering. An enhanced reflection component map is obtained based on the brightness map and the light estimation map, resulting in an enhanced image. A separation tendency index is obtained based on the mean and standard deviation of the brightness of pixels within a local window of each pixel in the enhanced image, the local gradient entropy, and the chroma. The separation tendency indices of all pixels are combined to form a separation tendency map. The separation tendency map is segmented to obtain several regions. The solid-phase aggregation degree is obtained based on the mean of the roundness of all regions and the variance of the region area. The chroma is the sum of the green-red and blue-yellow components of the pixel. The separation process of yucca extract is monitored based on the solid-phase aggregation degree of the separation process map at each time step and the variance of the area of ​​the segmented regions in the separation process map.

[0007] This invention first calculates the illumination unevenness of each pixel in an image, combining the difference between pixel brightness and the estimated illumination value, the standard deviation of local brightness, and the mean of local gradient, to accurately reflect the complexity of local illumination. Furthermore, based on the illumination unevenness, this invention adaptively determines the filter size for each pixel for subsequent image enhancement, effectively suppressing overexposure or underexposure caused by complex lighting changes in industrial environments, while preserving details of the solid-liquid boundary. Next, this invention calculates a solid-liquid separation tendency index by combining local brightness, texture features, and chroma information of the enhanced image, accurately distinguishing solid particles from the liquid background. Finally, the solid aggregation degree is obtained by calculating the average roundness and area variance of the segmented region, and by monitoring the changing trends of solid aggregation degree and area variance within a time window, reliable judgment of the separation endpoint is achieved, improving the automation level and accuracy of monitoring the yucca extract separation process.

[0008] Preferably, the unevenness of illumination satisfies the following relationship: In the formula, The separation process diagram is shown in Figure 1. Uneven illumination of individual pixels The first step in the separation process diagram The brightness of each pixel The first step in the separation process diagram Illumination estimate for each pixel The separation process diagram is shown in Figure 1. The standard deviation of pixel brightness within a local window of a given pixel The first in the brightness diagram The average gradient magnitude of pixels within a local window of each pixel. To prevent division by zero errors, The separation process diagram is shown in Figure 1. A set of pixel indices within a local window of pixels. It is a function for maximizing the value.

[0009] This invention combines the deviation between pixel brightness and its estimated illumination value, standardizes it using the local brightness standard deviation, and introduces the local gradient mean as a weight to calculate local illumination unevenness. It can effectively distinguish between gradient changes caused by real edges or particle boundaries and changes in brightness in flat areas caused by illumination gradients, providing a reliable basis for selecting appropriate filtering scales in subsequent adaptive enhancement algorithms.

[0010] Preferably, the filter size satisfies the following relationship: In the formula, The first in the brightness diagram The filter size used per pixel and These are the preset maximum and minimum sizes, respectively. The separation process diagram is shown in Figure 1. Uneven illumination of individual pixels This represents the average illumination unevenness of all pixels in the separation process image. It is an S-shaped curve function. This is the floor symbol.

[0011] This invention obtains the filter size by measuring illumination non-uniformity, thus achieving adaptive determination of the filter size. When the illumination non-uniformity is high, a large-size filter is automatically used to fully remove the non-uniform illumination components; when the illumination non-uniformity is low, a small-size filter is used to preserve the details of the solid-liquid boundary. This solves the problem that traditional fixed-scale filtering cannot simultaneously achieve both illumination suppression and detail preservation.

[0012] Preferably, the step of obtaining the separation tendency index based on the mean and standard deviation of the brightness of each pixel within a local window in the enhanced image, the local gradient entropy, and the chroma includes: processing the enhanced reflection component map using the Sobel operator to obtain the brightness gradient magnitude of each pixel in the enhanced image; constructing a histogram of each pixel based on the brightness gradient magnitude of each pixel within a local window to obtain the local gradient entropy of each pixel; obtaining the mean and standard deviation of the brightness of each pixel within a local window in the enhanced image, and standardizing the pixel brightness using the mean and standard deviation; and obtaining the separation tendency index based on the standardized brightness, local gradient entropy, and chroma of each pixel.

[0013] Preferably, the separation tendency index satisfies the following relationship: In the formula, To enhance the image of the first Separation tendency index of individual pixels To enhance the image of the first The brightness of each pixel and The first one in the enhanced image The mean and standard deviation of pixel brightness within a local window of each pixel. To enhance the image of the first Local gradient entropy of each pixel To enhance the maximum local gradient entropy appearing in the image, To enhance the image of the first The color saturation of each pixel To enhance the image of the first The average chroma of pixels within a local window of each pixel. To prevent division by zero errors, It is an exponential function with the natural constant as its base. It is an S-shaped curve function.

[0014] This invention constructs a separation tendency index by combining the local relative brightness, local relative chroma, and local gradient entropy of pixels. It can take advantage of the physical characteristics that solid particles usually have high brightness and specific chroma, and at the same time use local gradient entropy to effectively suppress complex texture areas, so as to accurately distinguish solid particles from liquid background even under low contrast and complex working conditions.

[0015] Preferably, the solid-phase aggregation degree satisfies the following relationship: In the formula, The solid-phase aggregation degree is shown in the separation process diagram. To separate the variance of the area of ​​all regions in the trend line plot, For the separation tendency diagram, the first The roundness of each region To separate the number of regions in the trend map, It is an exponential function with the natural constant as its base. Custom weight parameters.

[0016] This invention obtains the solid-phase aggregation degree by measuring the average roundness of all segmented regions and the variance of the region area, thereby achieving the assessment of the solid-phase aggregation state. The average roundness reflects whether the particle cluster morphology tends to be compact and regular, while the area variance reflects the uniformity of particle cluster size. The combination of the two can accurately reflect whether the separation process has reached a mature state of stable particle morphology and uniform size, providing a key indicator for judging the separation endpoint.

[0017] Preferably, the step of obtaining the enhanced reflection component map based on the luminance map and the illumination estimation map, and obtaining the enhanced image, includes: subtracting the luminance map from the illumination estimation map logarithmically to obtain the logarithmic enhanced reflection component map; performing an exponential operation on the logarithmic enhanced reflection component map to obtain the enhanced reflection component map; and recombinating the enhanced reflection component map with the green-red component map and the blue-yellow component map to obtain the enhanced image.

[0018] Preferably, the image segmentation of the separation tendency map includes: performing image segmentation of the separation tendency map using the Otsu's method.

[0019] Preferably, the step of monitoring the separation process of yucca extract based on the solid-phase aggregation degree of the separation process image at each time point and the variance of the area of ​​the segmented region in the separation process image includes: setting a time window; performing linear regression on each solid-phase aggregation degree within the time window using the least squares method to obtain a first regression line; performing linear regression on the variance of each region area in each image corresponding to the time window using the least squares method to obtain a second regression line; and determining that the separation process has reached the separation endpoint in response to the slope of the first regression line being less than a first threshold and the slope of the second regression line being less than a second threshold.

[0020] Secondly, the present invention provides a real-time image monitoring system for the separation process of yucca extract, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned real-time image monitoring method for the separation process of yucca extract is implemented.

[0021] By adopting the above technical solution, a computer program is generated from the real-time image monitoring method for the separation process of yucca extract, and stored in a memory for loading and execution by a processor. This allows for the creation of a terminal device based on the memory and processor, making it convenient to use.

[0022] The beneficial effects of this invention are as follows: First, by constructing an illumination non-uniformity and combining it with adaptive filtering size, this invention achieves proactive adaptation to the complex and ever-changing illumination conditions in industrial settings, effectively solving the problem of detail loss or insufficient enhancement in overexposed or underexposed areas by traditional fixed-parameter algorithms. Second, based on illumination adaptive enhancement, this invention constructs a solid-liquid separation tendency index by fusing multi-dimensional features such as local brightness, chroma, and local gradient entropy of the enhanced image. This allows the identification of solid particles to move beyond simple brightness or texture and proceed within a richer feature space, enabling accurate extraction of solid particles even under challenging conditions such as blurred solid-liquid boundaries and low contrast. Finally, by performing morphological analysis on the segmented solid regions, this invention introduces two indicators—solid aggregation degree and region area variance—to jointly characterize the separation process. By monitoring whether the trends of these two indicators simultaneously stabilize over time, it provides an objective basis for determining the separation endpoint, solving the problems of strong subjectivity and poor consistency associated with manual observation. This provides effective support for the automated monitoring and optimized control of the yucca extract separation process. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a real-time image monitoring method for the separation process of yucca extract according to the present invention;

[0024] Figure 2 This is a schematic diagram illustrating the separation process in this invention;

[0025] Figure 3 This is a schematic illustration of the enhanced image in this invention;

[0026] Figure 4 This is a schematic illustration of the segmentation result of the separation tendency map in this invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] This invention discloses a real-time image monitoring method for the separation process of yucca extract, referring to... Figure 1 This includes steps S1-S5:

[0030] S1. Real-time image acquisition: Obtain the illumination unevenness based on the brightness of each pixel in the image.

[0031] It should be noted that during the separation of yucca extract, the lighting conditions at the industrial site are complex and variable, often resulting in locally overexposed or underexposed images. This is the primary obstacle affecting the extraction of solid-liquid phase features. To achieve the adaptability of subsequent enhancement algorithms, this invention obtains the illumination unevenness based on the brightness of each pixel in the image.

[0032] Specifically, the separation process of yucca extract is captured in real time using an industrial camera. The separation process image is converted from RGB color space to LAB color space to obtain the brightness, green-red component, and blue-yellow component of each pixel. The brightness of all pixels is combined into a brightness image, the green-red component of all pixels is combined into a green-red component image, and the blue-yellow component of all pixels is combined into a blue-yellow component image. The gradient magnitude of each pixel in the brightness image is obtained using the Sobel operator, and Gaussian filtering is applied to the brightness image to obtain an illumination estimation image. The value of each pixel in the illumination estimation image is the illumination estimation value of the corresponding pixel in the separation process image.

[0033] Furthermore, the standard deviation of the brightness of each pixel within a local window in the separation process image is obtained, and the mean value of the gradient magnitude of each pixel within a local window in the brightness image is obtained. Based on the difference between the brightness and the estimated illumination value of each pixel in the separation process image, the standard deviation of the brightness of each pixel within a local window, and the mean value of the gradient magnitude of each pixel within a local window, the illumination non-uniformity of each pixel is obtained. The side length of the local window is 21, and the implementer can determine the side length of the local window according to the actual situation.

[0034] Specifically, the unevenness of illumination satisfies the following relationship:

[0035] ;

[0036] In the formula, The separation process diagram is shown in Figure 1. Uneven illumination of individual pixels The first step in the separation process diagram The brightness of each pixel The first step in the separation process diagram Illumination estimate for each pixel The separation process diagram is shown in Figure 1. The standard deviation of pixel brightness within a local window of a given pixel The first in the brightness diagram The average gradient magnitude of pixels within a local window of each pixel. To prevent division by zero errors in parameters, this embodiment... The value is 0.01, and the implementers can adjust it according to the actual situation. The separation process diagram is shown in Figure 1. A set of pixel indices within a local window of pixels. It is a function for maximizing the value.

[0037] in, The separation process diagram was measured. The degree to which the brightness of a pixel deviates from the ambient light within a local window; the larger the value, the more significant the difference. The greater the difference between the brightness of the first pixel and the estimated illumination value, the greater the difference between the second and third pixels. Within a local window of a given pixel, there may be brightness anomalies that deviate from the overall brightness level, causing uneven brightness within the local window. The greater the unevenness of illumination at each pixel, the smaller the value indicates that the... The smaller the difference between the brightness of the first pixel and the estimated illumination value, the better. The more uniform the overall brightness level of the pixels within a local window of a given pixel, the better. The smaller the illumination unevenness of each pixel; since different pixels have different local characteristics, the formula is improved by dividing by... The way to Standardize it.

[0038] Because low-frequency illumination gradients may exist in the image, it may be difficult to distinguish between flat areas of brightness gradients and brightness anomalies at actual solid-liquid boundaries or particle edges. Therefore, by... Further adjustments ;when The larger the value, the more likely it is to indicate the number of... The more likely a local window of a given pixel corresponds to a true solid-liquid boundary or particle edge, the better. The greater the unevenness of illumination per pixel; when The smaller the number, the more likely it is to be a problem. The more likely a local window of a pixel is to correspond to a region with flat brightness, such as inside a liquid phase, then the... The smaller the unevenness of illumination per pixel.

[0039] S2. Obtain the enhanced reflection component map based on the illumination unevenness of each pixel in the separation process map, and obtain the enhanced image based on the enhanced reflection component map.

[0040] It should be noted that in the field of image enhancement, Retinex-based algorithms are commonly used to separate illumination and reflection components. However, traditional enhancement algorithms typically apply a fixed set of filter scales across the entire image when estimating ambient illumination. This fixed-scale approach cannot effectively address the problem of dynamic changes in local illumination across spatial location within an image. For example, applying a large-scale filter to a relatively uniformly illuminated area may result in over-smoothing and loss of details at solid-liquid boundaries; conversely, applying a small-scale filter to a region with drastic illumination changes may fail to adequately remove the illumination component, leading to insufficient enhancement. Therefore, this invention obtains the enhanced reflection component image based on the illumination unevenness of each pixel in the separation process image.

[0041] Specifically, the average value of the illumination unevenness of all pixels in the separation process image is obtained. Based on the illumination unevenness of each pixel in the separation process image and the average value of the illumination unevenness of all pixels, as well as the preset maximum and minimum sizes, the filtering size of each pixel is obtained. The maximum size is 11 and the minimum size is 5. The implementer can determine the maximum and minimum sizes according to the actual situation.

[0042] Specifically, the filter size satisfies the following relationship:

[0043] ;

[0044] In the formula, The first in the brightness diagram The filter size used per pixel and These are the preset maximum and minimum sizes, respectively. The separation process diagram is shown in Figure 1. Uneven illumination of individual pixels This represents the average illumination unevenness of all pixels in the separation process image. It is an S-shaped curve function. This is the floor symbol.

[0045] in, The larger the value, the higher the brightness in the graph. The greater the difference between the illumination unevenness of a pixel and the overall level of illumination unevenness in the image, the larger the filter size should be to fully remove the uneven illumination components. The smaller the value, the higher the brightness in the graph. The more uniform the illumination within a local window of a pixel, the smaller the filter size should be to preserve the details of the solid-liquid boundary.

[0046] Furthermore, the brightness map is processed using Gaussian filtering based on the filtering size of each pixel to obtain the illumination estimation map. According to the principle of the Retinex algorithm, the brightness map and the illumination estimation map are logarithmically subtracted to obtain the logarithmically enhanced reflection component map. Then, each value in the logarithmically enhanced reflection component map is exponentially operated to obtain the enhanced reflection component map. The logarithmic subtraction between the two maps is the logarithmic subtraction of the pixel values ​​of pixels at the same position in the two images. This operation is an existing step in the Retinex algorithm and will not be elaborated here.

[0047] Furthermore, the enhanced reflection component map is recombined with the green-red component map and the blue-yellow component map to obtain the enhanced image.

[0048] For example, Figure 2 This is a schematic diagram illustrating the separation process in this invention. Figure 3 The image schematically illustrates the enhanced image obtained by the present invention. As can be seen from the image, the illumination in the enhanced image obtained by the present invention is more uniform, which can be better used for monitoring the yucca separation process.

[0049] S3. Obtain the separation tendency index based on the local differences and texture features of the enhanced image.

[0050] It should be noted that although uneven illumination is suppressed in the enhanced image, the contrast between the solid particles and the liquid background of the yucca extract remains low at certain stages (especially in the early or late stages of separation), and the boundaries are blurred. To accurately distinguish between the solid particles and the liquid background under low contrast, this invention obtains a separation tendency index based on the local differences and texture features of the enhanced image.

[0051] Specifically, the enhanced reflection component map is processed by the Sobel operator to obtain the brightness gradient magnitude of each pixel in the enhanced image. Based on the brightness gradient magnitude of all pixels in the local window of each pixel in the enhanced image, a histogram of brightness gradient magnitude of pixels in the local window of the pixel is constructed. The local gradient entropy of each pixel is obtained based on the brightness gradient magnitude histogram.

[0052] Furthermore, the mean and standard deviation of the brightness of each pixel within a local window in the enhanced image are obtained. The sum of the blue-yellow and green-red components of each pixel is taken as the chroma of that pixel. The mean of the chroma of each pixel within a local window is obtained. Based on the mean and standard deviation of the brightness of each pixel within a local window, the local gradient entropy, and the chroma, the separation tendency index is obtained.

[0053] Specifically, the segregation propensity index satisfies the following relationship:

[0054] ;

[0055] In the formula, To enhance the image of the first Separation tendency index of individual pixels To enhance the image of the first The brightness of each pixel and The first one in the enhanced image The mean and standard deviation of pixel brightness within a local window of each pixel. To enhance the image of the first Local gradient entropy of each pixel To enhance the maximum local gradient entropy appearing in the image, To enhance the image of the first The color saturation of each pixel To enhance the image of the first The average chroma of pixels within a local window of each pixel. To prevent division by zero errors, It is an exponential function with the natural constant as its base. It is an S-shaped curve function.

[0056] in, The larger the number, the more likely it is to be the first. The brighter a pixel is, the more likely it is to be significantly brighter than the overall brightness of the pixels within its local window. This pixel is a local bright spot. Because yucca extract solid particles have a brighter physical property than liquid particles, therefore the... The greater the separation tendency index of each pixel, the better; conversely, the smaller the separation tendency index, the less likely the pixel is to be separated. The smaller, the more... The smaller the separation tendency index of each pixel, the better.

[0057] when The larger the value, the more likely it is to be the first. The more significantly higher the chroma of a pixel is than the average chroma of its local neighborhood, the more likely that pixel is a locally prominent color. Since the solid particles of yucca extract have chroma characteristics distinct from the liquid background, therefore... The greater the separation tendency index of each pixel, the better; conversely, the smaller the separation tendency index, the less likely the pixel is to be separated. The smaller, the more... The smaller the separation tendency index of each pixel, the better.

[0058] when The larger the value, the more likely it is to be the first. The more widespread and chaotic the gradient distribution of pixels within a local window of a given pixel, the more complex the texture within that local window, and the more likely it is caused by image noise, bubbles in the separation process, or highly blurred dynamic boundaries. The more likely a pixel is not located at a solid particle, the more it should suppress the separation tendency index, making the first pixel... The smaller the separation tendency index of each pixel, the better; when The smaller the number, the more likely it is to be the first. The more concentrated and uniform the gradient distribution of pixels within a local window of a pixel, the smoother the texture of that region, and the more likely it corresponds to the interior of stable solid particles or a smooth liquid background. In this case, there is no need to excessively suppress the separation tendency index.

[0059] S4. Obtain the connected regions based on the separation tendency index, and obtain the solid-phase aggregation degree of the separation process diagram based on the morphology of the connected regions.

[0060] It should be noted that during the separation of yucca extract, when the sedimentation and aggregation of solid particles have basically stopped, the separation of the solid and liquid phases has reached a dynamic equilibrium, and the separation endpoint has been reached. After reaching the separation endpoint, the morphology of the product no longer changes significantly. Therefore, the determination of the separation endpoint is an important part of the monitoring of the yucca extract separation process. In order to better monitor the yucca extract separation process, determine the separation endpoint, and optimize the extract separation process, this invention obtains the connected regions based on the separation tendency index, and obtains the solid phase aggregation degree of the separation process diagram based on the morphology of the connected regions.

[0061] Specifically, the separation tendency index of all pixels is combined into a separation tendency map, and the separation tendency map is segmented to obtain several regions; the roundness and area of ​​each region are obtained, and the solid-phase aggregation degree is obtained based on the mean of the roundness of each region and the variance of the area of ​​the region.

[0062] In one embodiment, the algorithm used for image segmentation of the separation tendency map is the Otsu's method.

[0063] Specifically, the solid-phase aggregation degree satisfies the following relationship:

[0064] ;

[0065] In the formula, The solid-phase aggregation degree is shown in the separation process diagram. To separate the variance of the area of ​​all regions in the trend line plot, For the separation tendency diagram, the first The roundness of each region To separate the number of regions in the trend map, It is an exponential function with the natural constant as its base. In this embodiment, to customize the weight parameters, The value is 5, and the implementers can adjust it according to the actual situation. Size.

[0066] in, The mean of the roundness of all regions segmented in the separation process diagram represents the aggregation of solid particles during the separation of yucca extract. This aggregation refers not only to the increase in the number of particles but also to the process of their morphology becoming more stable and compact. When the separation becomes stable and mature, the solid particles or particle clusters transform from an initial irregular, loose, flocculated state into a denser, more regular shape. Ideally, their projected shape on the two-dimensional image is closest to a circle. Therefore, the degree of solid aggregation is calculated by taking the mean of the roundness of all regions. The smaller the particle size, the more likely the solid particles or particle clusters are to be in an irregular, loose, flocculated state, and the lower the degree of solid aggregation. The larger the particle or particle cluster, the more likely it is to be in a dense and regular form, and the greater the solid phase aggregation degree.

[0067] In the separation process of yucca extract, the ideal aggregation state is not only characterized by high roundness of the region in the image, but also by high uniformity of the region area. If the particle clusters are highly round but vary greatly in size, it indicates that the separation process is unstable and has not yet reached the separation endpoint. Therefore, when A larger value indicates a more uniform area in the separation tendency diagram, and thus a greater degree of solid-phase aggregation in the separation process diagram; when The smaller the value, the more uneven the area of ​​the region in the separation tendency diagram, and the smaller the solid-phase aggregation degree in the separation process diagram.

[0068] It should be noted that if no region is detected in the image segmentation, the solid-phase aggregation degree of the separation process map is 0, indicating that no particles have been precipitated yet.

[0069] For example, Figure 4 This is the segmentation result of the separation tendency map in this invention. The upper left corner of the map shows the monitoring time and the solid phase aggregation degree, while the upper right corner shows that the current separation process is still in progress and has not yet reached the separation endpoint.

[0070] S5. Determine the separation endpoint based on the solid-phase aggregation degree of the separation process diagram at each time point.

[0071] Specifically, during the separation of yucca extract, the separation process images of yucca extract are acquired in real time. The solid-phase aggregation degree of the yucca extract separation process image at each moment and the variance of the segmented area in the image are obtained. A time window is set, and the data within the time window length before each moment is taken as the data within the time window at that moment. The variance of all solid-phase aggregation degrees and segmented areas within each time window are obtained. The least squares method is used to perform linear regression on each solid-phase aggregation degree within the time window to obtain the first regression line. The least squares method is used to perform linear regression on the variance of each area within the time window to obtain the second regression line. If the slope of the first regression line is less than the first threshold and the slope of the second regression line is less than the second threshold, the separation process reaches the end point. The length of the time window is 5 seconds, the first threshold is 0.005, and the second threshold is 0.01. The implementer can determine the length of the time window and the size of the threshold according to the actual situation.

[0072] This invention also discloses a real-time image monitoring system for the separation process of yucca extract, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a real-time image monitoring method for the separation process of yucca extract according to the present invention is implemented.

[0073] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A real-time image monitoring method for the separation process of yucca extract, characterized in that, include: The separation process image is decomposed into a brightness image, a green-red component image, and a blue-yellow component image. The illumination non-uniformity of each pixel is obtained based on the difference between the brightness of each pixel and the illumination estimate, the standard deviation of the brightness of the pixel within the local window of the pixel, and the mean value of the gradient magnitude of the pixel within the local window of the pixel. The illumination estimate is obtained by Gaussian filtering the brightness image. The filtering size of each pixel is obtained based on the illumination non-uniformity of each pixel in the separation process diagram and the average value of the illumination non-uniformity of all pixels. The brightness map is filtered by the filtering size of each pixel to obtain the illumination estimation map. The enhanced reflection component map is obtained based on the brightness map and the illumination estimation map, and the enhanced image is obtained. Based on the mean and standard deviation of pixel brightness within the local window of each pixel in the enhanced image, the local gradient entropy, and the chroma acquisition separation tendency index, the following conditions are met: , To enhance the image of the first Separation tendency index of individual pixels and The first one in the enhanced image The mean and standard deviation of pixel brightness within a local window of each pixel. , , The first one in the enhanced image The brightness, local gradient entropy, and chroma of each pixel. To enhance the maximum local gradient entropy appearing in the image, To enhance the image of the first The average chroma of pixels within a local window of each pixel. To prevent division by zero errors, It is an exponential function with the natural constant as its base. It is an S-shaped curve function; The separation tendency indices of all pixels are combined to form a separation tendency map. Image segmentation is then performed on the separation tendency map to obtain several regions. The solid-phase aggregation degree is obtained based on the mean of the roundness of all regions and the variance of the region area, satisfying the following: , The solid-phase aggregation degree is shown in the separation process diagram. To separate the variance of the area of ​​all regions in the trend line plot, For the separation tendency diagram, the first The roundness of each region To separate the number of regions in the trend map, Custom weight parameters; Chroma is the sum of the green and red components and the blue and yellow components of a pixel; The separation process of yucca extract was monitored based on the solid-phase aggregation degree of the separation process diagram at each time point and the variance of the area of ​​the region segmented by the separation process diagram.

2. The real-time image monitoring method for the separation process of yucca extract according to claim 1, characterized in that, The illumination unevenness satisfies the following relationship: ； In the formula, The separation process diagram is shown in Figure 1. Uneven illumination of individual pixels The first step in the separation process diagram The brightness of each pixel The first step in the separation process diagram Illumination estimate for each pixel The separation process diagram is shown in Figure 1. The standard deviation of pixel brightness within a local window of a given pixel The first in the brightness diagram The average gradient magnitude of pixels within a local window of each pixel. To prevent division by zero errors, The separation process diagram is shown in Figure 1. A set of pixel indices within a local window of pixels. It is a function for maximizing the value.

3. The real-time image monitoring method for the separation process of yucca extract according to claim 1, characterized in that, The filter size satisfies the following relationship: ； In the formula, The first in the brightness diagram The filter size used per pixel and These are the preset maximum and minimum sizes, respectively. The separation process diagram is shown in Figure 1. Uneven illumination of individual pixels This represents the average illumination unevenness of all pixels in the separation process image. It is an S-shaped curve function. This is the floor symbol.

4. The real-time image monitoring method for the separation process of yucca extract according to claim 1, characterized in that, The step of obtaining a separation tendency index based on the mean and standard deviation of pixel brightness within a local window of each pixel in the enhanced image, the local gradient entropy, and chroma includes: The enhanced reflection component map is processed using the Sobel operator to obtain the brightness gradient magnitude of each pixel in the enhanced image. A histogram of each pixel is constructed based on the brightness gradient magnitude of each pixel within its local window to obtain the local gradient entropy of each pixel. The mean and standard deviation of the brightness of each pixel within its local window in the enhanced image are obtained, and the brightness of the pixels is standardized using the mean and standard deviation. The separation tendency index is obtained based on the standardized brightness, local gradient entropy, and chroma of each pixel.

5. The real-time image monitoring method for the separation process of yucca extract according to claim 1, characterized in that, The step of obtaining the enhanced reflection component map based on the luminance map and the illumination estimation map, and obtaining the enhanced image, includes: subtracting the luminance map from the illumination estimation map logarithmically to obtain the logarithmic enhanced reflection component map; performing an exponential operation on the logarithmic enhanced reflection component map to obtain the enhanced reflection component map; and recombining the enhanced reflection component map with the green-red component map and the blue-yellow component map to obtain the enhanced image.

6. The real-time image monitoring method for the separation process of yucca extract according to claim 1, characterized in that, The image segmentation of the separation tendency map includes: performing image segmentation of the separation tendency map using the Otsu's method.

7. The real-time image monitoring method for the separation process of yucca extract according to claim 1, characterized in that, The monitoring of the yucca extract separation process based on the solid-phase aggregation degree of the separation process map at each time point and the variance of the area of ​​the segmented region in the separation process map includes: Set a time window; perform linear regression on the solid-phase aggregation degree within the time window using the least squares method to obtain the first regression line; perform linear regression on the variance of the area of ​​each region in each image corresponding to the time window using the least squares method to obtain the second regression line; if the slope of the first regression line is less than the first threshold and the slope of the second regression line is less than the second threshold, it is determined that the separation process has reached the separation endpoint.

8. A real-time image monitoring system for the separation process of yucca extract, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a real-time image monitoring method for the separation process of yucca extract according to any one of claims 1-7.

Images