Permanganate titration end point automatic identification method and system

Abstract

The invention relates to the technical field of titration reaction, and discloses a permanganate titration end point automatic identification method, which comprises the following steps: S1, collecting a video stream of a titration experiment in real time, and framing the video stream to obtain image data; s2, preprocessing the image data; s3, performing preliminary detection and classification on the color change of the preprocessed image by using a KNN model; s4, background noise is eliminated through cooperative work of KNN and GMM Gaussian mixture models, and an image of a color change area is extracted; s5, performing feature analysis on the image of the color change area; calculating the area and the average chromaticity of the color change region; and S6, determining a titration end point according to the analysis result, and stopping the titration process when a predetermined condition is met. By monitoring the color change of the to-be-detected liquid in the titration process in real time and combining an image processing technology and a machine learning algorithm, the titration end point can be accurately identified, so that errors possibly caused by naked eye observation are avoided.

Description

Technical Field

[0001] The invention relates to the technical field of titration reaction, and in particular to a method and system for automatically identifying a permanganate titration endpoint. Background Art

[0002] Titration, as a classic quantitative analysis method, is widely used in water quality testing to determine key parameters such as pH, hardness, chloride ion content, dissolved oxygen and organic matter content in water samples. Through these tests, the quality of water bodies can be accurately assessed to ensure water safety and compliance. For example, pH reflects the chemical properties of water bodies, hardness affects the quality of industrial and domestic water, and chloride ions and dissolved oxygen are used to determine the quality of drinking water and sewage.

[0003] During the titration experiment, the change in color is an important basis for judging the progress of the titration reaction and determining the endpoint. Potassium permanganate (KMnO4) is often used as a titrant for redox titration. When its deep purple solution is added to the reducing test solution, it will quickly react with the reducing agent therein, causing the color of KMnO4 to fade quickly and the solution to become colorless. Only when the titration endpoint is approaching and the reducing substances in the solution are nearly exhausted, the purple color of KMnO4 will begin to appear and gradually deepen until the reaction stops completely and the solution color remains a stable light purple, which indicates that the titration reaction has reached its endpoint.

[0004] Traditional titration methods mostly rely on visual observation or simple color sensors to judge color changes to infer whether the reaction has reached the equivalence point or titration endpoint. These methods are easily disturbed by factors such as ambient light and subtle color changes, affecting the accuracy of the test results. Summary of the invention

[0005] The object of the present invention is to provide a method for automatically identifying the endpoint of permanganate titration, which can more accurately identify the titration endpoint by real-time monitoring the color change of the test liquid during the titration process and combining image processing technology and machine learning algorithm, thereby avoiding the errors that may be caused by naked eye observation or simple color sensors.

[0006] The technical solution adopted by the present invention to solve the technical problem is: to provide a method for automatically identifying the end point of permanganate titration, comprising the following steps: S1, real-time acquisition of the video stream of the titration experiment, and frame-dividing the video stream to obtain image data; S2, preprocessing the image data; S3, use the KNN model to perform preliminary detection and classification of the color changes of the preprocessed images; S4, the background noise is eliminated through the collaborative work of KNN and GMM mixed Gaussian model, and the image of the color change area is extracted; S5, performing feature analysis on the image of the color change region; analyzing and calculating the area and average chromaticity of the color change region; S6. Determine the titration endpoint according to the above analysis results, and stop the titration process when a predetermined condition is met.

[0007] Furthermore, S2 includes: Convert each frame of image from RGB color space to Lab color space; Gaussian filtering is used to denoise and smooth the L channel in the Lab color space. The formula is: ; Apply contrast-limited adaptive histogram equalization to adjust the brightness distribution. The formula is: ; The L channel is processed in layers through the layered image algorithm, including: using Gaussian filtering to smooth the background layer, compressing and smoothing the exposure area of ​​the background layer. The formula is: ; The edge information of the image detail layer is extracted through the edge detection algorithm. The formula is: ; Sobel x and Sobel y They represent the gradient values ​​of the image in the horizontal direction (x-axis) and the vertical direction (y-axis), respectively, and are calculated using the Sobel operator. They are the core parameters of edge detection and are used to measure the intensity of brightness changes in the image in the horizontal and vertical directions. and They are the partial derivatives of the brightness function L(x,y) in the x and y directions, that is, the local brightness change rate of the image in the horizontal and vertical directions. The derivatives are calculated using the Sobel convolution kernel and reflect the strength of the image edge (the larger the gradient, the more obvious the edge).

[0008] Use a sharpening filter to enhance the details. The formula is: ; The processed brightness layer is resynthesized, and the background layer and detail layer are combined to generate a resynthesized L channel; Merge the resynthesized L channel with the original a and b channels to obtain the processed Lab image; Convert the processed Lab image back to RGB color space; in, is the filter window size, ( i , j) is the offset relative to the current pixel, i is the horizontal offset, and both are vertical offsets; k represents the radius of the Gaussian kernel, that is, the range of the convolution kernel extending from the central pixel to the surrounding area; k controls the spatial range of the Gaussian filter and directly affects the size and smoothness of the filter; and are the minimum and maximum pixel values ​​in the original brightness image, respectively. is the standard deviation of the Gaussian distribution, is the original image, is the denoised brightness image, is the image after contrast limited adaptive histogram equalization, is the brightness image after smoothing, For the image after detail enhancement, is a coefficient that controls detail enhancement.

[0009] Furthermore, S3 includes: During the titration process, each frame of the image is acquired in real time; Extract the chromaticity information of each frame of image as a feature vector; Compare the feature vector of the current frame image with the feature vector in the known sample set, and classify the current frame into a colorless state, a color gradient state, or a titration endpoint state by calculating the Euclidean distance between the two; Calculate the distance between the feature vector of the current frame image and all labeled samples, and select the K nearest neighbors; According to the color change categories of the K neighbors, the final classification result of the current frame image is determined by majority voting.

[0010] Furthermore, the Euclidean distance calculation formula is: ; in, is the Euclidean distance between the current frame image and the sample feature; n represents the dimension of the feature vector, that is, the total number of features; i is an index variable used to traverse each component in the feature vector; is the weight coefficient of different features; and are the first A quantity.

[0011] Furthermore, S4 includes: During the colorless state stage of titration, each frame of the image is continuously monitored by the KNN algorithm and classified as colorless; Use the GMM model to extract the noise of each frame image and separate the foreground and background information; Process multiple frames of images and add the noise of each frame to generate a comprehensive noise image; In the color gradient and titration endpoint stages, the GMM model is used to process each frame of the image, delete the noise extracted in the colorless state stage, and extract the color change area.

[0012] Furthermore, the GMM model is used to extract the noise of each frame image, and the formula for separating foreground and background information is: ; in, Indicates the probability that a pixel belongs to the foreground or background; Represents the number of Gaussian components in the Gaussian mixture model; i is the index variable; is the weight of each Gaussian component; and are the mean and covariance matrices of the Gaussian components respectively; Mean and the covariance matrix is a Gaussian distribution with parameters.

[0013] Furthermore, S5 includes: Calculate the area S and average chromaticity C of the color change region; The calculation formula for the area S of the color change region is: ; The calculation formula of average chromaticity C is: ; in, is the pixel set of the extracted color change area, is the value of each pixel in the area, Indicates the chromaticity value of each pixel in the color change area.

[0014] Furthermore, S6 determines that the titration endpoint needs to meet the following conditions at the same time: a) KNN algorithm determines that the current frame is the titration endpoint state; b) The change in the area S of the color change region per unit time is less than the set area change threshold; c) The change in average chromaticity C per unit time is less than the set chromaticity change threshold.

[0015] A permanganate titration endpoint automatic recognition system is also provided, which is characterized by comprising: A camera is used to collect the video stream of the liquid to be tested during the titration process in real time and transmit the video data to the image processing module; An image processing module is used to pre-process each frame of the image from the RGB color space to the Lab color space; The KNN classification module is used to perform dynamic color monitoring by calculating the Euclidean distance between the feature vector of the current frame image and the feature vector of the known sample, and classify the current frame into a colorless state, a color gradient state, or a titration endpoint state; The GMM analysis module is used to extract the noise part of each frame image in the colorless stage of titration, and continue to process and remove the previously extracted colorless noise in the color gradient and titration endpoint stages, thereby improving the accuracy of color change detection; The feature analysis module is used to calculate the area S of the color change region and the change of the average chromaticity C, which are important indicators for determining the titration endpoint; An endpoint determination module is used to determine the titration endpoint according to the area S of the color change region and the stability of the average chromaticity C, and to stop the titration process when a predetermined condition is met; The control module is used to control the entire titration experiment process and automatically stop the experiment operation when the titration endpoint is reached; The display module is used to enable the experimenter to observe the processed images and prompt information in real time.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention can more accurately identify the titration endpoint by real-time monitoring of the color change of the test liquid during the titration process and combining image processing technology and machine learning algorithms, thereby avoiding errors that may be caused by naked eye observation or simple color sensors.

[0017] 2. Compared with traditional color recognition methods, this invention can capture more subtle color changes, which makes the judgment of the reaction progress more accurate and helps to improve the accuracy and reliability of titration experiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of a method for automatically identifying a permanganate titration endpoint according to Example 1 of the present invention; Figure 2 This is a rendering of the effect of the S1 image preprocessing of the method for automatic identification of the permanganate titration endpoint in Example 1 of the present invention; Figure 3 These are three state diagrams of a titration process classified by a KNN algorithm of a method for automatic identification of a permanganate titration endpoint in Example 1 of the present invention; Figure 4 This is an effect diagram of extracting the noise part of an image using a GMM model in a method for automatically identifying the endpoint of permanganate titration in Example 1 of the present invention; Figure 5 It is a foreground image, a noise image and a denoised effect image after being processed using a GMM model in a method for automatically identifying a permanganate titration endpoint in Example 1 of the present invention; Figure 6 It is a structural schematic diagram of an automatic permanganate titration endpoint recognition system according to Example 2 of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0020] like Figure 1 As shown, S1: real-time acquisition of the video stream of the titration experiment, and frame division of the video stream to obtain image data; The titration experiment video stream is collected in real time by a camera and input into a computer, and the video is divided into frames to obtain image data.

[0021] S2: pre-process the image data to reduce the exposure interference caused by the cold light source and the white background reflection; First, each frame of the image is converted from RGB space to Lab color space. In Lab space, Gaussian filtering is used to denoise the L channel to smooth the noise in the image, as shown in formula (2-1). For areas with uneven brightness, contrast-limited adaptive histogram equalization is used to adjust the brightness distribution, enhance the local contrast of the image, improve the uneven brightness, effectively reduce overexposure, and balance the overall brightness. The formula is shown in (2-2). Then, the L channel is layered using a layered image algorithm. The background layer is smoothed using a Gaussian filter to ensure that the exposed area of ​​the background layer is compressed and smoothed. The formula is shown in (2-3). For the detail layer, the edge information of the image is extracted using an edge detection algorithm, and then a sharpening filter is used to enhance the details. The Sobel algorithm is used to detect areas with significant brightness changes without affecting the color part, and ultimately a detail-enhanced image is generated. The formulas are shown in formulas (2-4) and (2-5).

[0022] (2-1); (2-2); (2-3); (2-4); (2-5); in, is the filter window size, ( i , j) is the offset relative to the current pixel, and are the minimum and maximum pixel values ​​in the original brightness image, respectively. is the standard deviation of the Gaussian distribution, is the original image, is the denoised brightness image, is the image after contrast limited adaptive histogram equalization, is the brightness image after smoothing, For the image after detail enhancement, is a coefficient that controls detail enhancement.

[0023] The processed brightness layer is resynthesized, combined with the background layer and detail layer to generate the final L channel. This L channel is then merged with the original a and b channels to obtain a Lab image with uniform exposure and clear details. Finally, the Lab image is converted back to RGB space, which not only maintains the brightness consistency of the image, but also retains the integrity of the color. The effect of the image after processing is shown in the figure below. Figure 2 shown.

[0024] S3: Use the KNN model to perform preliminary detection and classification of color changes in the preprocessed images; During the titration process, the color changes gradually. In order to monitor this process in real time, the K-Nearest Neighbors (KNN) algorithm is used to perform dynamic color monitoring throughout the process. The image features of each frame are compared with the features of the known sample, and the Euclidean distance between the two is calculated to classify the current frame into three states: colorless, color gradient, and titration endpoint. The chromaticity information of each frame image is extracted as a feature, and the feature vector of each frame image is converted into Calculate the characteristic vector of known samples (colorless, gradient, titration endpoint) The Euclidean distance of is shown in formula (2-6).

[0025] (2-6) in, is the Euclidean distance between the current frame image and the sample feature, is the weight coefficient of different features, and are the first A quantity.

[0026] By calculating the distance between the current frame image and all labeled samples, and selecting the K nearest neighbors (that is, the K samples with the most similar color changes). After finding the K nearest neighbors, the KNN algorithm will use a majority vote to determine the classification result of the current frame image based on the categories of these K samples. In the early stage of titration, the KNN algorithm matches the feature vector of each frame with the "colorless" sample set. Since the solution has not yet changed color, the image feature value is close to the colorless state sample, and KNN will judge it as a colorless state. When the titration reaction approaches the end point, the color of the solution gradually changes, and the color features in the image begin to deviate from the colorless sample. At this time, the distance calculated by KNN begins to increase, and gradually decreases from the distance to the sample in the color gradient state. KNN will gradually classify the current frame as a color gradient state based on the change in distance. When the titration reaction reaches the end point, the color changes significantly. The KNN algorithm will calculate that the distance between the current frame and the titration endpoint state sample is the smallest, thereby classifying the image as the titration endpoint state. The three states in the titration process are as follows: Figure 3 shown.

[0027] S4: The background noise is eliminated through the collaborative work of KNN and GMM mixed Gaussian model, and the image of the color change area is extracted; During the colorless phase of the titration: The KNN algorithm continuously monitors each frame of the image and classifies it as "colorless". Since the solution has not yet started to change color, the information in the image is mainly composed of noise. Therefore, KNN can only recognize the characteristics of some noise. At this stage, the noise part of each frame of the image is extracted through the Gaussian mixture model (GMM). The effect is shown in the figure below. Figure 4 (a). The GMM model can effectively model the image and separate the foreground (possible color changes) and background (noise) information. The formula of GMM is shown in formula (2-7).

[0028] (2-7) in, Indicates the probability that a pixel belongs to the foreground or background. represents the number of Gaussian components in the Gaussian mixture model, is the weight of each Gaussian component, and are the mean and covariance matrices of the Gaussian components, respectively. Mean and the covariance matrix is a Gaussian distribution with parameters.

[0029] By processing multiple frames of images, the GMM model continuously extracts the noise area, superimposes the noise of each frame to generate a comprehensive noise image, and performs morphological processing on the noise image to remove smaller non-noise points and further optimize the noise image. The effect is as follows: Figure 4 As shown in (b). Since what is identified in the colorless stage is mainly noise, the noise extraction process provides a key basis for the denoising process in the subsequent stage.

[0030] During the color gradient and titration endpoint stage: When the titration reaction begins, the color of the solution changes slightly, and KNN begins to recognize the color change and classifies it as a color gradient state. At this point, GMM continues to process each frame of the image, but the color change is usually very subtle and is accompanied by a lot of noise. At this stage, GMM is used to extract the color change part of the image. Since the color change is relatively small, GMM may extract many noise signals at the same time. In order to solve the problem of noise interference in color change extraction, the system deletes the noise extracted in the colorless stage. This step is critical because by eliminating the known colorless noise, the actual color change area can be more clearly identified. This means that the system can distinguish color changes from previously extracted noise, thereby improving the accuracy of detection. After the colorless stage noise is deleted, the part of the GMM image that is finally left is the color change area. This enables the system to accurately extract and monitor the color changes of the solution in the color development state. The foreground image, noise image, and denoising effect image after GMM processing are shown in the figure below. Figure 5 shown.

[0031] S5: Perform feature analysis on the image of the color change region; analyze and calculate the area and average chromaticity of the color change region; In order to further analyze the significance of color change, it is necessary to calculate the area S of the color change area and the average chromaticity C of the color change area as important indicators for judging the titration endpoint. The size of the area can reflect the degree of color change. The calculation of the average chromaticity reflects the depth and nature of the color change. By quantitatively analyzing the chromaticity of the change area, it is possible to more accurately identify the slight color changes during the titration process. The formulas for the area S of the color change area and the average chromaticity C are shown in formulas (2-8) and (2-9).

[0032] (2-8) (2-9) in, is the pixel set of the extracted color change area, is the value of each pixel in the region, usually 1 (foreground) or 0 (background) in a binary mask image. Indicates the chromaticity value of each pixel in the color change area.

[0033] S6. Determine the titration endpoint according to the above analysis results, and stop the titration process when a predetermined condition is met.

[0034] By extracting the color change area and average chromaticity The system can significantly reflect the subtle color changes during the titration process. While continuously monitoring the area and average chromaticity of the color change area, the system will set a threshold to determine whether the changes in these indicators have reached stability. The judgment of the titration endpoint needs to meet the following three conditions: 1. KNN judgment: The KNN algorithm continuously monitors the state of each frame of the image during the entire titration process. When KNN determines that the current frame is at the titration endpoint, it indicates that the color change has basically stopped.

[0035] 2. Area Stability: Monitor the area of ​​the extracted color change region If the area change is less than the set threshold of area change per unit time within a certain period of time, it indicates that the color change amplitude tends to be stable, which means that the titration reaction is close to completion.

[0036] 3. Average chromaticity Stability: Average Chromaticity It is the main and accurate reference condition for determining the titration endpoint. The system continuously monitors the change in average chromaticity. If the change in average chromaticity is less than the set threshold of chromaticity change per unit time within a certain period of time, it means that the color change has stabilized.

[0037] When the above three conditions are met at the same time, the system will prompt that the titration endpoint has been reached and the experiment will end automatically. This process ensures the accuracy and effectiveness of the experimental results and makes the monitoring of the titration process more reliable.

[0038] Embodiment 2: like Figure 6 As shown, a permanganate titration endpoint automatic recognition system comprises: A camera is used to collect the video stream of the liquid to be tested during the titration process in real time and transmit the video data to the image processing module; An image processing module is used to pre-process each frame of the image from the RGB color space to the Lab color space; The KNN classification module is used to perform dynamic color monitoring by calculating the Euclidean distance between the feature vector of the current frame image and the feature vector of the known sample, and classify the current frame into a colorless state, a color gradient state, or a titration endpoint state; The GMM analysis module is used to extract the noise part of each frame image in the colorless stage of titration, and continue to process and remove the previously extracted colorless noise in the color gradient and titration endpoint stages, thereby improving the accuracy of color change detection; The feature analysis module is used to calculate the area S of the color change region and the change of the average chromaticity C, which are important indicators for determining the titration endpoint; An endpoint determination module is used to determine the titration endpoint according to the area S of the color change region and the stability of the average chromaticity C, and to stop the titration process when a predetermined condition is met; The control module is used to control the entire titration experiment process and automatically stop the experiment operation when the titration endpoint is reached; The display module is used to enable the experimenter to observe the processed images and prompt information in real time.

[0039] This article uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for automatically identifying the endpoint of permanganate titration, characterized in that: The following steps are involved: S1, real-time acquisition of the video stream of the titration experiment, and frame-dividing the video stream to obtain image data; S2, preprocessing the image data; S3, use the KNN model to perform preliminary detection and classification of the color changes of the preprocessed images; S4, the background noise is eliminated through the collaborative work of KNN and GMM mixed Gaussian model, and the image of the color change area is extracted; S5, performing feature analysis on the image of the color change region; analyzing and calculating the area and average chromaticity of the color change region; S6. Determine the titration endpoint according to the above analysis results, and stop the titration process when a predetermined condition is met.

2. The method for automatically identifying the endpoint of permanganate titration according to claim 1, wherein S2 include: Convert each frame of image from RGB color space to Lab color space; Gaussian filtering is used to denoise and smooth the L channel in the Lab color space. The formula is: ； Apply contrast-limited adaptive histogram equalization to adjust the brightness distribution. The formula is: ； The L channel is processed in layers through the layered image algorithm, including: using Gaussian filtering to smooth the background layer, compressing and smoothing the exposure area of ​​the background layer. The formula is: ； The edge information of the image detail layer is extracted through the edge detection algorithm. The formula is: ； Use a sharpening filter to enhance the details. The formula is: ； The processed brightness layer is resynthesized, and the background layer and detail layer are combined to generate a resynthesized L channel; Merge the resynthesized L channel with the original a and b channels to obtain the processed Lab image; Convert the processed Lab image back to RGB color space; in, is the filter window size, is the offset relative to the current pixel; Sobel x and Sobel y They represent the gradient values ​​of the image in the horizontal x-axis and vertical y-axis, respectively, which are calculated by the Sobel operator; k represents the radius of the Gaussian kernel, that is, the range of the convolution kernel extending from the central pixel to the surrounding areas; and are the minimum and maximum pixel values ​​in the original brightness image, respectively. is the standard deviation of the Gaussian distribution, is the original image, is the denoised brightness image, is the image after contrast limited adaptive histogram equalization, is the brightness image after smoothing, For the image after detail enhancement, is a coefficient that controls detail enhancement.

3. The method for automatically identifying the endpoint of permanganate titration according to claim 2, wherein S3 include: During the titration process, each frame of the image is acquired in real time; Extract the chromaticity information of each frame of image as a feature vector; Compare the feature vector of the current frame image with the feature vector in the known sample set, and classify the current frame into a colorless state, a color gradient state, or a titration endpoint state by calculating the Euclidean distance between the two; Calculate the distance between the feature vector of the current frame image and all labeled samples, and select the K nearest neighbors; According to the color change categories of the K neighbors, the final classification result of the current frame image is determined by majority voting.

4. A method for automatically identifying a permanganate titration endpoint according to claim 3, characterized in that: The Euclidean distance calculation formula is: ； in, is the Euclidean distance between the current frame image and the sample feature; n represents the dimension of the feature vector, that is, the total number of features; i is the index variable; is the weight coefficient of different features; and are the first A quantity.

5. The method for automatically identifying the endpoint of permanganate titration according to claim 3, wherein S4 include: During the colorless state stage of titration, each frame of the image is continuously monitored by the KNN algorithm and classified as colorless; Use the GMM model to extract the noise of each frame image and separate the foreground and background information; Process multiple frames of images and add the noise of each frame to generate a comprehensive noise image; In the color gradient and titration endpoint stages, the GMM model is used to process each frame of the image, delete the noise extracted in the colorless state stage, and extract the color change area.

6. A method for automatically identifying a permanganate titration endpoint according to claim 5, characterized in that: The GMM model is used to extract the noise of each frame image and the formula for separating foreground and background information is: ； in, Indicates the probability that a pixel belongs to the foreground or background; Represents the number of Gaussian components in the Gaussian mixture model; i is the index variable; is the weight of each Gaussian component; and are the mean and covariance matrices of the Gaussian components respectively; Mean and the covariance matrix is a Gaussian distribution with parameters.

7. The method for automatically identifying the endpoint of permanganate titration according to claim 5, wherein S5 include: Calculate the area S and average chromaticity C of the color change region; The calculation formula for the area S of the color change region is: ； The calculation formula of average chromaticity C is: ； in, is the pixel set of the extracted color change area, is the value of each pixel in the area, Indicates the chromaticity value of each pixel in the color change area.

8. A method for automatically identifying a permanganate titration endpoint according to claim 7, characterized in that: S6 determines the titration endpoint when the following conditions are met simultaneously: a) KNN algorithm determines that the current frame is the titration endpoint state; b) The change in the area S of the color change region per unit time is less than the set area change threshold; c) The change in average chromaticity C per unit time is less than the set chromaticity change threshold.

9. A permanganate titration endpoint automatic recognition system, characterized in that: include: A camera is used to collect the video stream of the liquid to be tested during the titration process in real time and transmit the video data to the image processing module; An image processing module is used to pre-process each frame of the image from the RGB color space to the Lab color space; The KNN classification module is used to perform dynamic color monitoring by calculating the Euclidean distance between the feature vector of the current frame image and the feature vector of the known sample, and classify the current frame into a colorless state, a color gradient state, or a titration endpoint state; The GMM analysis module is used to extract the noise part of each frame image in the colorless stage of titration, and continue to process and remove the previously extracted colorless noise in the color gradient and titration endpoint stages, thereby improving the accuracy of color change detection; The feature analysis module is used to calculate the area S of the color change region and the change of the average chromaticity C, which are important indicators for determining the titration endpoint; An endpoint determination module is used to determine the titration endpoint according to the area S of the color change region and the stability of the average chromaticity C, and to stop the titration process when a predetermined condition is met; The control module is used to control the entire titration experiment process and automatically stop the experiment operation when the titration endpoint is reached; The display module is used to enable the experimenter to observe the processed images and prompt information in real time.

Images