Color titration identification method, system and equipment based on combined prediction model

Abstract

The invention relates to a color titration identification method, system and equipment based on a combined prediction model, and belongs to the technical field of solution titration identification, and the method comprises the following steps: training an ARIMA model by using hue measured values; judging whether the difference value between the hue measured value after each titration and the hue predicted value of the ARIMA model is greater than a first threshold value or not; if yes, determining that the current dropping point is a color mutation inflection point, and collecting at least three continuous hue measured values after the color mutation inflection point to train a linear regression model; and based on the trained linear regression model, iteratively executing three times of hue value prediction and titration operation until an average difference value between three continuous predicted values and measured values is smaller than a second threshold value, and judging that the first dropwise adding point in the three continuous dropwise adding points is a stable color inflection point. According to the method, different models are used for performing color identification on different reaction stages, the method can meet the requirements of a multi-color titration formula, and meanwhile, the accuracy and the anti-interference capability of color titration are improved.

Description

Technical Field

[0001] This invention relates to the field of titration reaction technology, and in particular to a color titration identification method, system, and device based on a combined prediction model. Background Technology

[0002] Color titration is one of the most widely used analytical methods in chemical analysis. By analyzing the degree of color change of the reaction solution during the titration process, the chemical composition of the reaction solution can be determined. It is suitable for determining chemical components in fields such as petroleum, food, medicine, and environment.

[0003] Currently, traditional manual colorimetric titration involves adding water and an indicator to a solution of unknown concentration, then titrating it with a solution of known concentration. During titration, once the titrant and analyte have completely reacted, the solution color will change significantly, and the endpoint is reached if the color does not return to its original state within 30 seconds. This traditional titration relies on the operator's visual judgment, and varying sensitivity to color can lead to inaccuracies. This repetitive titration process is time-consuming and can introduce uncertainty into the experimental results. Furthermore, corrosive titrants may pose health risks, increasing the overall risk of the titration process.

[0004] Existing automated color titration identification systems lack versatility, requiring modifications to the titration formula or adjustments to equipment parameters for each chemical composition analysis. After formula modifications or changes, the equipment often fails to adapt, necessitating operator resetting of parameters. Furthermore, the titration process is subject to stringent conditions, and the accuracy of color recognition is primarily influenced by the sensor, resulting in high maintenance, operating, and manufacturing costs for automated titration equipment. Summary of the Invention

[0005] (a) Technical problems to be solved In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a color titration identification method, system and device based on a combined prediction model, which solves the technical problems of poor universality, low anti-interference ability and high operation and maintenance cost of existing automated color titration identification.

[0006] (II) Technical Solution To achieve the above objectives, the main technical solutions adopted by the present invention include: In a first aspect, embodiments of the present invention provide a color titration recognition method based on a combined prediction model, comprising: During the titration reaction stage, the measured color values ​​of the solution were collected after each addition of titration reagent, and when the number of additions reached the target sample size, the obtained measured color values ​​were used to train the preset ARIMA model. The measured hue value after each addition of titrant during the cyclic titration process is compared with the predicted hue value obtained by the trained ARIMA model, and it is determined whether the difference is greater than the set first threshold. When the difference is greater than the set first threshold, the current titration point is determined to be the color change inflection point, and at least three consecutive measured hue values ​​after the color change inflection point are collected as a dynamic training set to train the preset linear regression model. Based on a trained linear regression model, the hue value prediction and titration reagent addition operations are performed iteratively three times until the average difference between the three consecutive hue prediction values ​​and the measured hue values ​​is less than the set second threshold. At this point, the first of the three consecutive addition points is determined to be the color stability inflection point.

[0007] Optionally, during the titration reaction stage, the measured hue values ​​of the solution are collected after each addition of titrant. Before training the pre-defined ARIMA model using the obtained measured hue values ​​when the target sample size is reached after the number of additions, the method further includes: In response to the titration operation command, the titration system is initialized, including tubing rinsing, background light source adjustment, and titration identification and positioning. After initialization, the titration system acquires image frames of the solution to be titrated and converts the image frames into an HSV color model. After Gaussian blur denoising of the HSV color model, the average hue value of the pixels in the region to be titrated in the denoised HSV color model is determined as the original hue value.

[0008] Optionally, during the titration reaction stage, the measured hue values ​​of the solution are collected after each addition of titrant, and when the target sample size is reached after the number of additions, the pre-defined ARIMA model is trained using the obtained measured hue values, including: During the titration reaction stage, image frames are acquired after each addition of titration reagent, and when the target sample size is reached after the number of additions, the acquired image frames are converted into the HSV color model. After Gaussian blur denoising of all HSV color models, the average hue value of the pixels in the region of the solution to be titrated in the denoised HSV color model is determined as the measured hue value. Determine whether the difference between the measured hue value and the original hue value of the solution to be titrated is greater than the set hue difference threshold; If the difference between the measured hue value and the original hue value of the solution to be titrated is greater than the set hue difference threshold, a prompt message indicating that the solution to be titrated is contaminated will be output. If the difference between the measured hue value and the original hue value of the solution to be titrated is not greater than the set hue difference threshold, then the original hue value and all measured hue values ​​are used to train the preset ARIMA model.

[0009] Optionally, the measured hue value after each addition of titrant during the cyclic titration process is compared with the predicted hue value obtained by the trained ARIMA model, and the determination of whether the difference is greater than a set first threshold includes: During the cyclic titration process, the color prediction value of the solution after the first addition of titrant is obtained by using a trained ARIMA model; The predicted hue value is compared with the corresponding measured hue value, and it is determined whether the difference is greater than the set first threshold. If the difference is greater than the set first threshold, the current titration point is determined to be a color change inflection point; If the difference is not greater than the set first threshold, the hue value of the solution at the current titration point is added to the training set of the ARIMA model. The ARIMA model is then iteratively trained using the updated training set so that the iteratively trained ARIMA model can predict the hue value after the next titration reagent is added.

[0010] Optionally, when the difference is greater than a set first threshold, the current titration point is determined as the color change inflection point, and at least three consecutive measured hue values ​​after the color change inflection point are collected as a dynamic training set to train the preset linear regression model, including: If the difference between the predicted hue value and the measured hue value is greater than the set first threshold, the current titration point is determined as the color change inflection point, and the number of cyclic titrations to reach the color change inflection point is recorded. Determine whether the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point exceeds a set threshold. If the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point does not exceed the set amplitude threshold, a control command to end the titration will be output. If the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point exceeds a set threshold, the measured hue values ​​obtained after the color abrupt change inflection point will be used as a dynamic training set to train the preset linear regression model.

[0011] Optionally, based on a trained linear regression model, three iterations of hue value prediction and titration reagent addition are performed until the average difference between the three consecutive predicted hue values ​​and the measured hue values ​​is less than a set second threshold. The first addition point among the three consecutive addition points is then determined as the color stabilization inflection point, including: The predicted hue values ​​of the solution to be titrated after three consecutive titration operations are predicted using a trained linear regression model, and the measured hue values ​​of the solution to be titrated are obtained simultaneously. Determine whether the average difference between three consecutive predicted hue values ​​and measured hue values ​​is less than a set second threshold. When the average difference between the predicted hue value and the measured hue value for three consecutive times is less than the set second threshold, the first of the three consecutive titration points is determined to be the color stability inflection point, and the number of cyclic titrations to reach the color stability inflection point is recorded and the control command to end the titration is output. If the average difference between three consecutive predicted hue values ​​and the measured hue values ​​is not less than the set second threshold, then the three consecutive measured hue values ​​will replace the earliest three consecutive measured hue values ​​in the dynamic training set. The linear regression model will then be iteratively trained using the updated training set so that the iteratively trained linear regression model can predict the hue values ​​after the next three additions of titration reagent.

[0012] Secondly, embodiments of the present invention provide a color titration recognition system based on a combined prediction model, comprising: The ARIMA model training module is used to collect the measured color values ​​of the solution after each addition of titration reagent during the titration reaction stage, and to train the preset ARIMA model using the obtained measured color values ​​when the number of additions reaches the target sample size. The first titration color judgment module is used to compare the measured hue after each addition of titrant during the cyclic titration process with the predicted hue value predicted by the trained ARIMA model, and to determine whether the difference is greater than the set first threshold. The color mutation inflection point determination module is used to determine the current titration point as the color mutation inflection point when the comparison difference is greater than the set first threshold, and to collect at least three consecutive measured hue values ​​after the color mutation inflection point as a dynamic training set to train the preset linear regression model. The second titration color judgment module is used to iteratively perform three hue value prediction and titration reagent addition operations based on a trained linear regression model, and to determine whether the average difference between the three consecutive hue prediction values ​​and the measured hue values ​​is less than the set second threshold. The color stability inflection point determination module is used to determine the first of three consecutive dripping points as the color stability inflection point when the average difference between the predicted hue value and the measured hue value is less than a set second threshold.

[0013] Thirdly, embodiments of the present invention provide a titration apparatus, comprising: Quantitative liquid dispensing device; Industrial cameras; A motion component, connected to an industrial camera, is used to move the industrial camera to a designated position so that the industrial camera can acquire image frames of the solution to be titrated; The controller is connected to the quantitative liquid addition device, the industrial camera, and the motion components, and is used to execute the color titration recognition method based on the combined prediction model described above based on the acquired image frames.

[0014] Optionally, the quantitative liquid addition device includes: a peristaltic pump, a plunger pump, and a titration solution chamber; The peristaltic pump is connected to the titration solution tank pipeline and is used to add titration reagent to the titration solution tank; The plunger pump is connected to the titration solution tank pipeline and electrically connected to the color titration identification device. It is used to control the addition of the corresponding amount of titration reagent to the solution to be titrated according to the control command output by the color titration identification device.

[0015] Optionally, it also includes: a background light source and a light shield; The background light source is connected to the controller and is used to dynamically adjust the background brightness of the color recognition area based on the original hue value of the solution to be titrated output by the controller.

[0016] (III) Beneficial Effects The beneficial effects of this invention are as follows: By acquiring hue values ​​in real time and dynamically training an ARIMA model to predict trends, this method can keenly perceive and capture minute color change inflection points, avoiding the lag and misjudgment risks of traditional subjective visual inspection or fixed threshold methods. After detecting a change, a linear regression model is further dynamically trained using measured values ​​near the change point, and prediction and titration verification are performed iteratively. This can quickly and accurately locate the subsequent stable color inflection point. Through a dual-model collaborative detection and iterative verification mechanism, this invention achieves high-precision closed-loop control for titration endpoint determination, significantly reducing excessive reagent consumption and human judgment errors. It is particularly suitable for complex titration systems with gradual color changes or insignificant abrupt changes, improving the automation level and reliability of color titration analysis. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a color titration identification method based on a combined prediction model, provided in an embodiment of the present invention. Figure 2 This is a flowchart of the pre-titration preparation stage provided in an embodiment of the present invention; Figure 3 This is a flowchart of the learning sample acquisition stage provided in an embodiment of the present invention; Figure 4 This is a flowchart of the solution color change inflection point identification stage provided in an embodiment of the present invention; Figure 5 This is a prediction and measured curve of the inflection point of solution color change provided in an embodiment of the present invention; Figure 6 This is a flowchart of the solution color smoothing inflection point identification stage provided in an embodiment of the present invention; Figure 7 The image shows the predicted and measured curves of the solution color smoothing inflection point provided in an embodiment of the present invention. Detailed Implementation

[0018] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figures 1 to 7 As shown in the embodiment of the present invention, a color titration identification method based on a combined prediction model is proposed, which includes: collecting the measured hue values ​​of the solution after each addition of titrant during the titration reaction stage; and training a preset ARIMA model (autoregressive moving average model) using the acquired measured hue values ​​when the number of additions reaches the target sample size; comparing the measured hue values ​​after each addition of titrant during the cyclic titration process with the predicted hue values ​​obtained by the trained ARIMA model, and determining whether the difference is greater than a set first threshold; if the difference is greater than the set first threshold, determining the current titration point as a color change inflection point, and collecting at least three consecutive measured hue values ​​after the color change inflection point as a dynamic training set to train a preset linear regression model; and iteratively executing three hue value prediction and titrant addition operations based on the trained linear regression model until the average difference between the three consecutive predicted hue values ​​and the measured hue values ​​is less than a set second threshold, and determining the first addition point among the three consecutive addition points as a color stabilization inflection point.

[0020] This embodiment predicts trends by acquiring hue values ​​in real time and dynamically training an ARIMA model. This method can keenly sense and capture subtle color change inflection points, avoiding the lag and misjudgment risks of traditional subjective visual inspection or fixed threshold methods. After detecting a change, a linear regression model is further dynamically trained using measured values ​​near the change point, and prediction and titration verification are performed iteratively. This can quickly and accurately locate the subsequent stable color inflection point. This embodiment achieves high-precision closed-loop control for titration endpoint determination through a dual-model collaborative detection and iterative verification mechanism, significantly reducing excessive reagent consumption and human judgment errors. It is particularly suitable for complex titration systems with gradual color changes or insignificant abrupt changes, improving the automation level and reliability of color titration analysis.

[0021] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0022] Specifically, refer to Figure 1 As shown, this embodiment proposes a color titration recognition method based on a combined prediction model, which includes: S100. During the titration reaction stage, the measured color values ​​of the solution are collected after each addition of titration reagent, and when the number of additions reaches the target sample size, the obtained measured color values ​​are used to train the preset ARIMA model.

[0023] In this embodiment, sub-steps G100 to G300 are included before step S100: G100, in response to titration operation commands, initializes the titration system, including tubing rinsing, background light source adjustment, and titration identification and positioning.

[0024] refer to Figure 2 As shown, in the pre-titration preparation stage, the titration system is initialized in response to the titration operation command to ensure that the overall environment meets the standard requirements for titration. First, the tubing of the titration system is rinsed to fill it with the titrant liquid, ensuring that only one drop of liquid can be added at a time, thus ensuring the accuracy of subsequent liquid additions. Next, the background light source is adjusted based on the original color value of the solution to be titrated. Then, the industrial camera is adjusted so that its shooting angle can capture the identification area of ​​the solution to be titrated. Finally, it is checked whether the identification area is entirely composed of the solution to be titrated. If so, the titration operation is deemed feasible; otherwise, more solution to be titrated needs to be added until the entire identification area is composed of the solution to be titrated.

[0025] G200 controls the titration system after initialization to acquire image frames of the solution to be titrated and convert the image frames into an HSV color model.

[0026] G300: After Gaussian blur denoising of the HSV color model, the average hue value of the pixels in the region of the solution to be titrated in the denoised HSV color model is determined as the original hue value.

[0027] In this embodiment, reference Figure 3 As shown, step S100 may further include sub-steps S110 to S140: S110. During the titration reaction stage, acquire image frames after each addition of titration reagent, and when the number of additions reaches the target sample size, convert the acquired image frames into the HSV color model.

[0028] After each addition of titration reagent, an image frame with stabilized color is captured using an industrial camera, and then the image frame is converted into an HSV color model. In the HSV color model, H represents hue, S represents saturation, and V represents value.

[0029] S120. After Gaussian blur denoising of all HSV color models, the average hue value of the pixels in the region to be titrated in the denoised HSV color model is determined as the measured hue value.

[0030] Gaussian blurring of an image can preserve color details while reducing noise. After averaging the hue values ​​of the pixels in the recognition area, the average value is used as a training sample. The entire HSV color model is traversed to perform noise reduction and hue value calculation to obtain the training samples of the target sample size.

[0031] S130. Determine whether the difference between the measured hue value and the original hue value of the solution to be titrated is greater than the set hue difference threshold.

[0032] S140a. If the difference between the measured hue value and the original hue value of the solution to be titrated is greater than the set hue difference threshold, a prompt message indicating that the solution to be titrated is contaminated will be output.

[0033] S140b If the difference between the measured hue value and the original hue value of the solution to be titrated is not greater than the set hue difference threshold, then the original hue value and all measured hue values ​​are used to train the preset ARIMA model.

[0034] In this embodiment, a pre-analysis of the components of the solution to be titrated is performed before titration to determine whether the solution is contaminated, thus ensuring the purity and stability of the solution. By comparing the color difference between the solution without titrant and the solution with titrant but without color change, the contamination status of the solution to be titrated can be quickly determined. Furthermore, in the absence of contamination, the hue value of the solution with titrant but without color change can be collected as ARIMA model training data.

[0035] S200. Compare the measured hue value after each addition of titration reagent during the cyclic titration process with the predicted hue value obtained by the trained ARIMA model, and determine whether the difference is greater than the set first threshold.

[0036] In this embodiment, reference Figure 4As shown, step S200 may further include sub-steps S210 to S230: S210. During the cyclic titration process, the predicted color value of the solution after the first addition of titration reagent is obtained by using a trained ARIMA model.

[0037] S220. Compare the predicted hue value with the corresponding measured hue value, and determine whether the difference is greater than the set first threshold.

[0038] S230a. When the difference is greater than the set first threshold, the current titration point is determined to be the color change inflection point.

[0039] S230b: When the difference is not greater than the set first threshold, the solution hue value at the current titration point is added to the training set of the ARIMA model. The ARIMA model is then iteratively trained using the updated training set so that the iteratively trained ARIMA model can predict the hue value after the next titration reagent is added.

[0040] In this embodiment, during the cyclic titration process, hue recognition and ARIMA model prediction are used to determine the color abrupt change points of the solution. First, based on n training samples from the ARIMA model's training phase, the ARIMA model is trained to predict the hue value after the (n+1)th titration. Next, after the (n+1)th addition of titrant, the actual hue value is obtained by processing the image using an industrial camera. If the difference between the predicted and actual values ​​is not greater than a set first threshold, it indicates that no abrupt change in the titration color has occurred, and the hue value continues to exhibit stable linearity. The actual hue value after the (n+1)th titration is then added to the original training set, and the first training sample is removed to maintain the sample size at n. Then, the ARIMA model is iteratively trained based on the updated training samples to predict the hue value after the (n+2)th titration reagent is added, until the difference between the predicted and actual values ​​exceeds the set first threshold. Finally, when the difference is greater than the first threshold, due to the previously stable linear training set, the prediction value based on the ARIMA model will continue to predict a stable linear hue value. However, the actual solution color change will cause a large deviation between the actual hue value and the predicted value, which verifies that the solution to be titrated has undergone a color change at the current moment, and thus determines the current liquid addition volume as the solution titration color change point.

[0041] In one specific embodiment, if the instant of the first color change is the inflection point, marking the point of complete reaction, the concentration of the unknown stock solution can be determined accordingly. Figure 5As shown, the ARIMA model is initially trained using n training samples in the initial stage of titration to predict the hue value at the (n+1)th point. After the (n+1)th addition of titration reagent, the actual hue value is obtained by processing the image using an industrial camera, and the difference between the predicted and actual values ​​is calculated in real time. If hue value fluctuations occur during titration but do not change continuously, the difference will not exceed the set threshold, and it will not be identified as a color abrupt change point. When the solution color undergoes a sudden change, the real-time hue value will change accordingly, and the measured hue value curve will show an upward or downward trend. At the same time, since the training value before the abrupt change is still a smooth curve, the predicted value will continue to produce a smooth value prediction during the abrupt change, resulting in a difference greater than the set threshold. The current detection point is considered the instantaneous point of color abrupt change, i.e., the required titration inflection point (inflection point 1). The concentration of the unknown stock solution can be calculated by using the current titration volume, the concentration of the blank sample solution, the concentration of the titration reagent, and the volume of the unknown concentration stock solution.

[0042] S300. When the difference is greater than the set first threshold, the current titration point is determined as the color change inflection point, and at least three consecutive measured hue values ​​after the color change inflection point are collected as a dynamic training set to train the preset linear regression model.

[0043] In this embodiment, step S300 may further include sub-steps S310 to S330: S310. If the difference between the predicted hue value and the measured hue value is greater than the set first threshold, the current titration point is determined as the color change inflection point, and the number of cyclic titrations to reach the color change inflection point is recorded.

[0044] S320. Determine whether the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point exceeds the set amplitude threshold.

[0045] S330a: If the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point does not exceed the set amplitude threshold, then output a control command to end the titration.

[0046] S330b: When the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point exceeds a set amplitude threshold, the measured hue values ​​obtained after the color abrupt change inflection point will be used as a dynamic training set to train the preset linear regression model.

[0047] In this embodiment, during the titration reaction, some solutions continue to react after reaching the color inflection point, with further addition of titrants. The color change becomes more pronounced with each titration. During this process, the solution color varies to different degrees depending on the formulation, resulting in fluctuations in the hue value. Eventually, the hue value smooths out, and this smoothing point is defined as the titration endpoint. Therefore, this embodiment uses at least three consecutive measured hue values ​​after the color inflection point as a training set to train the linear regression model. The trained linear regression model then predicts the hue value. As long as the difference between the predicted and measured values ​​is not less than a second threshold, the prediction and titration process is iteratively executed. Specifically, before reaching the titration smoothing point, the measured hue value after each titration is added to the original training set, replacing the earliest measured hue value, forming a dynamic training set. This ensures that the linear regression model is trained before each prediction.

[0048] S400: Based on the trained linear regression model, iteratively execute three hue value prediction and titration reagent addition operations until the average difference between the three consecutive hue prediction values ​​and the measured hue values ​​is less than the set second threshold. Then, determine the first of the three consecutive addition points as the color stability inflection point.

[0049] In this embodiment, reference Figure 6 As shown, step S400 may further include sub-steps S410 to S430: S410. Predict the hue of the solution to be titrated after three consecutive titration operations using a trained linear regression model, and simultaneously obtain the measured hue of the solution to be titrated.

[0050] S420. Determine whether the average difference between three consecutive predicted hue values ​​and measured hue values ​​is less than a set second threshold.

[0051] S430a: When the average difference between the predicted hue value and the measured hue value for three consecutive times is less than the set second threshold, the first of the three consecutive titration points is determined to be the color stability inflection point, and the number of cyclic titrations to reach the color stability inflection point is recorded and the control command to end the titration is output.

[0052] S430b: When the average difference between three consecutive hue prediction values ​​and the measured hue values ​​is not less than the set second threshold, the three consecutive measured hue values ​​replace the earliest three consecutive measured hue values ​​in the dynamic training set. The linear regression model is then iteratively trained using the updated training set so that the iteratively trained linear regression model can predict the hue prediction values ​​after the next three additions of titration reagent.

[0053] In this embodiment, the linear regression model, due to its inherent characteristics and compatibility with the trend of the change curve, is suitable as a predictive model for the latter half of the titration prediction. Using m (at least 3) actual hue values ​​as the training set, the linear regression model is trained to predict the hue value of the next drop, and the actual hue value after adding reagent is obtained. Since hue becomes unstable, the process is repeated 3 times, ensuring the training set is approximately m times to guarantee prediction accuracy. If the average difference between the 3 predictions and the actual value is greater than a second threshold, it indicates that the color continues to fluctuate significantly, and the iterative training and prediction continue. If the average difference between the 3 predictions and the actual value is less than the second threshold, it indicates that the predicted curve closely matches the actual curve, and the color has stabilized. The first drop point out of the three consecutive drop points is determined as the color stabilization inflection point.

[0054] In one specific embodiment, such as Figure 7 As shown, after identifying the color abrupt change point, the actual hue values ​​from the last m times are selected as the training set, and the hue value after the next titration reagent addition is predicted based on a linear regression model. Subsequently, the actual hue values ​​are obtained and compared with the predicted values. Due to the characteristics of the linear regression model, the predicted hue values ​​will show a trend of increasing or decreasing depending on the rate of change. To ensure the accuracy of the prediction, the entire process is repeated 3 times, always maintaining the training set as the data from the last m times. The average of the differences between the last 3 predictions and the actual values ​​is used as the judgment criterion. When the difference is greater than a second threshold, it is determined that the color is still fluctuating; when the difference is less than the second threshold, the predicted curve closely matches the actual hue value, and it is determined that the color has stabilized. The current point at this time is the color stabilization inflection point, i.e. Figure 7 Inflection point 2. The concentration of the original solution of the solution to be titrated can be calculated by using the current titration volume, the concentration of the blank sample solution, the concentration of the titrating reagent, and the volume of the original solution of unknown concentration.

[0055] Furthermore, this embodiment also proposes a color titration recognition system based on a combined prediction model, including: The ARIMA model training module is used to collect the measured hue values ​​of the solution after each addition of titration reagent during the titration reaction stage, and to train a preset ARIMA model using the obtained measured hue values ​​when the number of additions reaches the target sample size.

[0056] The first titration color judgment module is used to compare the measured hue after each addition of titrant during the cyclic titration process with the predicted hue value predicted by the trained ARIMA model, and to determine whether the difference is greater than the set first threshold.

[0057] The color mutation inflection point determination module is used to determine the current titration point as the color mutation inflection point when the comparison difference is greater than the set first threshold, and to collect at least three consecutive measured hue values ​​after the color mutation inflection point as a dynamic training set to train the preset linear regression model. The second titration color judgment module is used to iteratively perform three hue value prediction and titration reagent addition operations based on a trained linear regression model, and to determine whether the average difference between the three consecutive hue prediction values ​​and the measured hue values ​​is less than a set second threshold.

[0058] The color stability inflection point determination module is used to determine the first of three consecutive dripping points as the color stability inflection point when the average difference between the predicted hue value and the measured hue value is less than a set second threshold.

[0059] Finally, this embodiment also proposes a titration device, including: a quantitative liquid addition device; an industrial camera; a motion component connected to the industrial camera for moving the industrial camera to a designated position so that the industrial camera can acquire image frames of the solution to be titrated; and a controller connected to the quantitative liquid addition device, the industrial camera, and the motion component for executing the color titration recognition method based on the combined prediction model described above based on the acquired image frames.

[0060] In this embodiment, the quantitative liquid addition device includes: a peristaltic pump, a plunger pump, and a titration solution chamber; the peristaltic pump is connected to the titration solution chamber pipeline and is used to add titration reagent to the titration solution chamber; the plunger pump is connected to the titration solution chamber pipeline and electrically connected to the color titration recognition device, and is used to control the corresponding amount of titration reagent to be added to the solution to be titrated according to the control command output by the color titration recognition device.

[0061] In this embodiment, the titration device further includes: a background light source and a light shield; the background light source is connected to the controller and is used to dynamically adjust the background brightness of the color recognition area according to the original hue value of the solution to be titrated output by the controller.

[0062] In summary, embodiments of the present invention propose a color titration identification method, system, and device based on a combined prediction model. First, the color of the first stage of titration is identified using an autoregressive moving average (ARIMA) model to find the inflection point of color abrupt change. Then, the color of the second stage of titration is identified using a linear regression model to find the inflection point of color smoothing. Finally, automated color titration is performed on a solution of unknown concentration, and the actual concentration of the original solution is calculated by back-calculating the inflection point obtained from the analysis of the formulation blank sample and the combined prediction model. This invention, by rationally utilizing the inherent characteristics of the ARIMA and linear regression models, can quickly identify two inflection points of color change, thus meeting the needs of various color titration formulations. Furthermore, using different models for color titration identification at different reaction stages can further improve the accuracy of color titration and the anti-interference capability of the identification analysis.

[0063] Since the systems / devices described in the above embodiments of the present invention are systems / devices used to implement the methods of the above embodiments of the present invention, those skilled in the art can understand the specific structure and modifications of the systems / devices based on the methods described in the above embodiments of the present invention, and therefore will not be repeated here. All systems / devices used in the methods of the above embodiments of the present invention fall within the scope of protection of the present invention.

[0064] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0065] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0066] It should be noted that in the description of this invention, the word "a" or "an" preceding a component does not exclude the existence of multiple such components. This invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. The use of terms such as first, second, third, etc., is merely for convenience and does not indicate any order. These terms can be understood as part of the component names.

[0067] Furthermore, it should be noted that in the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0068] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning of the basic inventive concept, can make other changes and modifications to these embodiments.

[0069] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from the spirit and scope of the invention.

Claims

1. A color titration recognition method based on a combined prediction model, characterized in that, include: During the titration reaction stage, the measured color values ​​of the solution were collected after each addition of titration reagent, and when the number of additions reached the target sample size, the obtained measured color values ​​were used to train the preset ARIMA model. The measured hue value after each addition of titrant during the cyclic titration process is compared with the predicted hue value obtained by the trained ARIMA model, and it is determined whether the difference is greater than the set first threshold. When the difference is greater than the set first threshold, the current titration point is determined to be the color change inflection point, and at least three consecutive measured hue values ​​after the color change inflection point are collected as a dynamic training set to train the preset linear regression model. Based on a trained linear regression model, the hue value prediction and titration reagent addition operations are performed iteratively three times until the average difference between the three consecutive hue prediction values ​​and the measured hue values ​​is less than the set second threshold. At this point, the first of the three consecutive addition points is determined to be the color stability inflection point.

2. The method as described in claim 1, characterized in that, The process includes collecting measured hue values ​​of the solution after each addition of titrant during the titration reaction stage, and before training the pre-defined ARIMA model using the obtained measured hue values ​​when the target sample size is reached after the number of additions, the following steps are also included: In response to the titration operation command, the titration system is initialized, including tubing rinsing, background light source adjustment, and titration identification and positioning. After initialization, the titration system acquires image frames of the solution to be titrated and converts the image frames into an HSV color model. After Gaussian blur denoising of the HSV color model, the average hue value of the pixels in the region to be titrated in the denoised HSV color model is determined as the original hue value.

3. The method as described in claim 1, characterized in that, During the titration reaction stage, the measured color values ​​of the solution were collected after each addition of titrant. When the target sample size was reached after the number of additions, the obtained measured color values ​​were used to train a pre-defined ARIMA model, including: During the titration reaction stage, image frames are acquired after each addition of titration reagent, and when the target sample size is reached after the number of additions, the acquired image frames are converted into the HSV color model. After Gaussian blur denoising of all HSV color models, the average hue value of the pixels in the region of the solution to be titrated in the denoised HSV color model is determined as the measured hue value. Determine whether the difference between the measured hue value and the original hue value of the solution to be titrated is greater than the set hue difference threshold; If the difference between the measured hue value and the original hue value of the solution to be titrated is greater than the set hue difference threshold, a prompt message indicating that the solution to be titrated is contaminated will be output. If the difference between the measured hue value and the original hue value of the solution to be titrated is not greater than the set hue difference threshold, then the original hue value and all measured hue values ​​are used to train the preset ARIMA model.

4. The method as described in claim 1, characterized in that, The measured hue value after each addition of titrant during the cyclic titration process is compared with the predicted hue value obtained by the trained ARIMA model. The determination of whether the difference exceeds a set first threshold includes: During the cyclic titration process, the color prediction value of the solution after the first addition of titrant is obtained by using a trained ARIMA model; The predicted hue value is compared with the corresponding measured hue value, and it is determined whether the difference is greater than the set first threshold. If the difference is greater than the set first threshold, the current titration point is determined to be a color change inflection point; If the difference is not greater than the set first threshold, the hue value of the solution at the current titration point is added to the training set of the ARIMA model. The ARIMA model is then iteratively trained using the updated training set so that the iteratively trained ARIMA model can predict the hue value after the next titration reagent is added.

5. The method as described in claim 1, characterized in that, When the difference exceeds a set first threshold, the current titration point is determined as the color abrupt change inflection point. At least three consecutive measured hue values ​​after the color abrupt change inflection point are then collected as a dynamic training set to train the preset linear regression model, including: If the difference between the predicted hue value and the measured hue value is greater than the set first threshold, the current titration point is determined as the color change inflection point, and the number of cyclic titrations to reach the color change inflection point is recorded. Determine whether the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point exceeds a set threshold. If the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point does not exceed the set amplitude threshold, a control command to end the titration will be output. If the difference between at least three consecutive measured hue values ​​after the color abrupt change inflection point exceeds a set threshold, the measured hue values ​​obtained after the color abrupt change inflection point will be used as a dynamic training set to train the preset linear regression model.

6. The method as described in claim 1, characterized in that, Based on a trained linear regression model, three iterations of hue value prediction and titration reagent addition are performed until the average difference between the three consecutive predicted hue values ​​and the measured hue values ​​is less than a set second threshold. The first of the three consecutive addition points is then determined as the color stabilization inflection point, including: The predicted hue values ​​of the solution to be titrated after three consecutive titration operations are predicted using a trained linear regression model, and the measured hue values ​​of the solution to be titrated are obtained simultaneously. Determine whether the average difference between three consecutive predicted hue values ​​and measured hue values ​​is less than a set second threshold. When the average difference between the predicted hue value and the measured hue value for three consecutive times is less than the set second threshold, the first of the three consecutive titration points is determined to be the color stability inflection point, and the number of cyclic titrations to reach the color stability inflection point is recorded and the control command to end the titration is output. If the average difference between three consecutive predicted hue values ​​and the measured hue values ​​is not less than the set second threshold, then the three consecutive measured hue values ​​will replace the earliest three consecutive measured hue values ​​in the dynamic training set. The linear regression model will then be iteratively trained using the updated training set so that the iteratively trained linear regression model can predict the hue values ​​after the next three additions of titration reagent.

7. A color titration recognition system based on a combined prediction model, characterized in that, include: The ARIMA model training module is used to collect the measured color values ​​of the solution after each addition of titration reagent during the titration reaction stage, and to train the preset ARIMA model using the obtained measured color values ​​when the number of additions reaches the target sample size. The first titration color judgment module is used to compare the measured hue after each addition of titrant during the cyclic titration process with the predicted hue value predicted by the trained ARIMA model, and to determine whether the difference is greater than the set first threshold. The color mutation inflection point determination module is used to determine the current titration point as the color mutation inflection point when the comparison difference is greater than the set first threshold, and to collect at least three consecutive measured hue values ​​after the color mutation inflection point as a dynamic training set to train the preset linear regression model. The second titration color judgment module is used to iteratively perform three hue value prediction and titration reagent addition operations based on a trained linear regression model, and to determine whether the average difference between the three consecutive hue prediction values ​​and the measured hue values ​​is less than the set second threshold. The color stability inflection point determination module is used to determine the first of three consecutive dripping points as the color stability inflection point when the average difference between the predicted hue value and the measured hue value is less than a set second threshold.

8. A titration apparatus, characterized in that, include: Quantitative liquid dispensing device; Industrial cameras; A motion component, connected to an industrial camera, is used to move the industrial camera to a designated position so that the industrial camera can acquire image frames of the solution to be titrated; The controller is connected to the quantitative liquid addition device, the industrial camera, and the motion component, respectively, and is used to execute the color titration recognition method based on the combined prediction model as described in any one of claims 1-6 based on the acquired image frames.

9. The device as described in claim 8, characterized in that, The quantitative liquid addition device includes: a peristaltic pump, a plunger pump, and a titration solution chamber; The peristaltic pump is connected to the titration solution tank pipeline and is used to add titration reagent to the titration solution tank; The plunger pump is connected to the titration solution tank pipeline and electrically connected to the color titration identification device. It is used to control the addition of the corresponding amount of titration reagent to the solution to be titrated according to the control command output by the color titration identification device.

10. The device as claimed in claim 8, characterized in that, Also includes: Background light source and light shield; The background light source is connected to the controller and is used to dynamically adjust the background brightness of the color recognition area based on the original hue value of the solution to be titrated output by the controller.

Images