A method for calculating the water absorption rate of superabsorbent polymers based on an image classification algorithm
An image classification algorithm, utilizing a CNN, automates the calculation of water absorption rates by analyzing video frames, improving accuracy and efficiency over conventional manual methods.
Patent Information
- Application Number
- JP2024531482
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-10-23
- Filing Date
- 2023-10-24
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2043-10-24
AI Technical Summary
Conventional methods for calculating the water absorption rate of water-absorbing materials are inefficient and require manual measurement, limiting their applicability and accuracy.
A method and system using an image classification algorithm, specifically a convolutional neural network (CNN), to analyze frame-by-frame video data of water absorption processes to automatically determine the water absorption rate by distinguishing between images of water absorption in progress and completion.
The method provides accurate and automated calculation of water absorption rates, reducing human error and enabling standardized measurement across various samples.
Smart Images

Figure 2024541549000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to a method for calculating the water absorption rate of a highly water-absorbent material. [Background technology]
[0002] Conventionally, the water absorption rate of a water absorbent material has been calculated, and the material has been used for applications that match the water absorption rate. Conventionally, in order to calculate such water absorption rate, a method has been used in which a sample water absorbent material is placed in the place where it is to be used, and then the water absorption time is measured manually using a timer.
[0003] However, such a method was inefficient because it forced researchers to measure the water absorption time manually one by one.
[0004] Patent Document 1 discloses a method for measuring the change in mass of a water-absorbent material to roughly grasp the time required for the amount of absorbed liquid of a sample to reach the maximum amount of absorbed water. However, this method has a problem in that it can only be applied once the maximum amount of absorbed water and the permeation rate coefficient of the sample are known.
[0005] Relevant prior art includes the following documents: [Prior art documents] [Patent documents]
[0006] [Patent Document 1] JP 2020-051997 A Summary of the Invention [Problem to be solved by the invention]
[0007] The present invention differs from the conventional methods in that it is an object of the present invention to provide a method and system for calculating the degree of water absorption of a water absorbent material from images (frame images) of water absorption video data of the water absorbent material. [Means for solving the problem]
[0008] The present invention provides a method for generating an artificial neural network model for calculating the water absorption rate of a water-absorbent material, the method comprising: a learning data preparation step including image data of a predetermined water-absorbent material and classification data labeled to distinguish whether the image is an image of water-absorbed material or an image that has already absorbed water; a construction step of an artificial neural network in which the predetermined image data is input and the input image is classified as a predetermined label; and a machine learning step of training the artificial neural network by inputting the learning data into the artificial neural network and updating parameters constituting the artificial neural network.
[0009] In this case, the image data of the water absorbent material is acquired by a video data acquisition step of acquiring water absorption video data for a large number of water absorbent materials, and a frame image data acquisition step of acquiring each frame image data constituting the video data. The water absorption video data is characterized in that a predetermined amount of water is put into a container and stirred while a predetermined amount of the water absorbent material is put in, and the process of the water being absorbed and solidified is captured.
[0010] The present invention also provides a method for calculating the water absorption rate of an absorbent material, comprising: an absorbent material image data acquisition step of adding an absorbent material and acquiring image data of the absorbent material over a predetermined period of time; a frame image data acquisition step of separating the image data into frame image data to acquire frame image data; a frame image classification step of inputting the frame image data into an artificial neural network model to classify the frame images; and an absorption rate calculation step of extracting absorption end frames and absorption start frames from the classified frame images and using these to calculate the water absorption rate.
[0011] In this case, the water absorption rate in the water absorption rate calculation step may be calculated by the following Equation 1. [Formula 1] Water absorption speed = (water absorption end frame number - water absorption start frame number) / fps of video data
[0012] Furthermore, the present invention provides a water absorption rate calculation system for absorbent materials, comprising a video data acquisition unit that acquires water absorption video data of the absorbent material, a frame image data acquisition unit that separates the water absorption video data into frame image data that constitute the video data and acquires the frame image data, and a water absorption rate calculation unit that calculates the water absorption start point and the water absorption end point from the frame image data to calculate the water absorption rate of the absorbent material.
[0013] The present invention further comprises a frame image discrimination unit which inputs each frame image data acquired by the frame image data acquisition unit in frame order and distinguishes whether it is an image during absorption, in which the absorbent material is absorbing water, or an image that has already absorbed water, and is characterized in that the water absorption rate calculation unit calculates the water absorption start time from the frame number (start frame number) of the image that the frame image discrimination unit first discriminated as an image during absorption, and calculates the water absorption end time from the frame number (end frame number) of the image that the frame image discrimination unit first discriminated as an image that has already absorbed water.
[0014] The frame image discrimination unit includes an artificial neural network unit that is pre-trained to receive each input frame image data, distinguish it into an image in water absorption or an image after water absorption, and output the discrimination result. The artificial neural network unit receives sample frame image data obtained by separating water absorption sample video data of a large number of sample absorbent materials into each sample frame image data, and data in which each sample frame image data is labeled as in water absorption or after water absorption, as training data, and is trained to match the labeled data with each sample frame image data. Effect of the Invention
[0015] According to the present invention, compared to the conventional method of manually measuring the water absorption rate of an absorbent material using a timer, it is possible to take images of water absorption of absorbent materials with a wide range of water absorption rates, and by applying an image classification algorithm, it is possible to measure the water absorption rate more accurately.
[0016] The following drawings attached to this specification are intended to illustrate preferred embodiments of the present invention and serve to facilitate a further understanding of the technical ideas of the present invention as well as the contents of the invention. Therefore, the present invention should not be interpreted as being limited to only the matters described in the drawings. [Brief description of the drawings]
[0017] [Figure 1] FIG. 1 is a diagram showing an overview of a system for calculating the water absorption rate of a water absorbent material according to the present invention. [Diagram 2] 1A to 1C are diagrams showing examples of images for each frame of a water absorption video according to an embodiment of the present invention. [Diagram 3] FIG. 2 is a diagram showing actual frame images of video data captured to prepare learning data. [Figure 4] FIG. 13 is a diagram showing the calculation output of the score per frame when video data is input to a predictive model trained according to the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0018] The present invention relates to a method and system for calculating the water absorption rate of a superabsorbent polymer (SAP) by utilizing an image classification algorithm. As the image classification algorithm used in the present invention, a normal convolutional neural network (CNN) can be used, and other image classification models that have been publicly known before the filing date of the present invention can also be applied. It should be noted that the technical feature of the present invention does not lie in the image classification model itself, but in the use of the image classification model to calculate the water absorption rate from the water absorption video of the superabsorbent polymer.
[0019] 1. Water absorption rate calculation system for water absorbent material (super absorbent polymer) according to the present invention
[0020] 1.1. Water absorption video data acquisition unit 10
[0021] The present invention includes a video data acquisition unit that acquires video data of the water absorption of the water absorbent material. The video data acquisition unit performs the video data acquisition step in the method for calculating the water absorption rate of the water absorbent material according to the present invention.
[0022] The water absorption video data of a water absorbent material refers to video data (video data) that is captured at a predetermined frame rate (fps) (number of frames per second) of the process of a water absorbent material absorbing water under a predetermined condition. The video data acquisition unit is a component that receives the video data and may include a normal video camera.
[0023] <Example of acquiring water absorption image data of water absorbent material>
[0024] An example of acquiring water absorption video data of a water absorbent material is shown. In the present invention, a 100 mL beaker was filled with 50 mL of 0.9 wt% salt water, and then magnetic stirring was performed at 600 rpm. While the salt water was rotating, 2 g of highly water-absorbent resin (water absorbent material) was added, and the entire process of the salt water being absorbed and solidified was photographed to acquire video data. At this time, the video was photographed with a video camera so that the angle of the surface inside the beaker in the video was 30° to 60°.
[0025] 1.2. Frame image data acquisition unit 20
[0026] The frame image data acquisition unit is a component that separates the image data acquired by the image data acquisition unit (water-absorption image data) or the input image data (water-absorption image data) into image data on a frame basis. It is well known to ordinary engineers that the task of separating predetermined image data into each frame image data can be easily realized by using ordinary image processing software or image processing algorithms.
[0027] The frame image data acquisition unit performs a frame image data acquisition step in the method for generating an artificial neural network model and the method for calculating the water absorption rate of a water absorbent material of the present invention.
[0028] 1.3. Water absorption rate calculation section 30
[0029] The water absorption rate calculation unit includes an artificial neural network model 31 which receives the acquired frame image data, classifies the frame images, and outputs their scores, an absorption progress determination unit 32 which calculates the water absorption start time and the water absorption end time from the output of the artificial neural network model, and a water absorption rate calculation unit 33 which calculates the water absorption rate from the water absorption start time and the water absorption end time.
[0030] (1) Image classification: Artificial neural network model 31
[0031] The present invention includes a frame image discrimination unit which inputs each frame image data acquired by the frame image data acquisition unit in frame order and distinguishes and discriminates whether it is an image during absorption, in which the absorbent material is absorbing water, or an image that has already absorbed water, in which the absorption of water has been completed.
[0032] The image discrimination unit 31 may be composed of an artificial neural network model, and the artificial neural network model of the present invention is an example of an image classification algorithm used to calculate the water absorption rate of the superabsorbent polymer in the present invention described above. As the artificial neural network model, a normal convolutional neural network (CNN) can be used, and other image classification models that have been publicly known before the filing date of the present invention can also be applied.
[0033] The artificial neural network constituting the artificial neural network model of the present invention is machine-learned according to a learning procedure described later, and receives each frame image data constituting the water absorption video data of the above-mentioned water-absorbent material, distinguishes whether the image is an image in the process of absorbing water or an image that has already absorbed water, and outputs the probability that the image is an image in the process of absorbing water or the probability that the image has already absorbed water as a score. The score output for each frame image is a value between 0 and 1. An example of the score for each frame image output by the artificial neural network model of the present invention trained according to the learning procedure described later is shown in Figure 4.
[0034] (2) Learning artificial neural networks
[0035] The artificial neural network model 31 of the present invention includes an artificial neural network that is machine-learned according to the following procedure. As the artificial neural network, a CNN is basically used.
[0036] First, as learning data, a 100mL beaker was filled with 50mL of 0.9wt% salt water, and then magnetic stirring was performed at 600rpm. While the salt water was rotating, 2g of superabsorbent resin (water absorbent) was added, and the entire process of the salt water absorbing and solidifying was filmed to obtain video data. At this time, the video was filmed with a video camera so that the angle of the surface inside the beaker in the video was 30° to 60°.
[0037] Next, the video data is separated into frame image data to obtain the frame image data.
[0038] Such video data and frame image data for learning data can be acquired using the video data acquisition unit 10 and frame image data acquisition unit 20 described above.
[0039] Next, the frame image data is merged with labeling data that labels the frame image as being an image in which water has been absorbed or an image in which water has already been absorbed. The labeling data can also be handled independently for each frame number without being merged with the frame image data.
[0040] FIG. 2 shows an example of how each video data set is separated into frame image data and labeled for preparing learning data.
[0041] Then, an artificial neural network is constructed that receives image data and classifies the input image as a predetermined label. The artificial neural network may have a typical CNN structure, and the data structure of CNN as a data processing structure realized in a computer is well known to those skilled in the art.
[0042] Next, the above-mentioned learning data is input to the artificial neural network constructed as above to proceed with machine learning and complete the learning of the artificial neural network. In such a machine learning step, the parameters of each layer constituting the artificial neural network are updated, and the artificial neural network model 31 of the present invention equipped with the trained artificial neural network receives frame image data obtained from any water-absorbent material video data and outputs a score, which is a discrimination value for whether the image corresponds to an image of water absorption or an image of water absorption.
[0043] Examples of score output values for the water absorption video data and frame image data of the specified water absorbent materials A (sample 10) and B (sample 16) are shown in Figure 4. The closer the score is to 0, the more water absorption has progressed, and the closer the score is to 1, the more water absorption has finished.
[0044] (3) Water absorption progress judgment section 32
[0045] The water absorption progress determination unit 32 calculates the water absorption start time and water absorption end time based on the score of each frame image data output by the artificial neural network model 31.
[0046] In the example in Figure 4, the frame where the score stabilizes near 0 is selected as the frame where water absorption begins, and the frame where the score rises to 0.5 or higher is selected as the frame where water absorption ends.
[0047] (4) Water absorption rate calculation section 33
[0048] The water absorption rate calculation unit 30 includes a water absorption rate calculation unit 33 that calculates the water absorption rate from the water absorption start time and the water absorption end time. The water absorption rate calculation unit 33 calculates the water absorption rate from the calculated water absorption start time and water absorption end time according to the following mathematical formula 2.
[0049] [Formula 2] Water absorption rate = water absorption end point - water absorption start point
[0050] The water absorption rate can also be calculated from the following formula 1.
[0051] [Formula 1] Water absorption speed = (water absorption end frame number - water absorption start frame number) / fps of video data
[0052] 2. Method for calculating the water absorption rate of the water absorbent material (super absorbent polymer) according to the present invention
[0053] The method for calculating the water absorption rate of a water absorbent material according to the present invention comprises the following steps.
[0054] First, as explained in the above embodiment section, the method includes a water-absorbent material image data acquisition step of putting a predetermined water-absorbent material into water and acquiring image data of the water-absorbent material for a predetermined period of time.
[0055] Next, the method includes a frame image data acquisition step of separating the video data into frame image data and acquiring the frame image data, and a frame image classification step of inputting the acquired frame image data into an artificial neural network model that has been machine-learned using the above-mentioned method to classify the frame images.
[0056] The method also includes a water absorption rate calculation step of extracting the water absorption end frame and the water absorption start frame from the classified frame images and calculating the water absorption rate using these. The water absorption rate can be calculated using the above-mentioned formula 1 or formula 2.
[0057] Example 1
[0058] The method for measuring the water absorption rate of Samples 1 to 16 according to the water absorption rate calculation method of the present invention is as follows.
[0059] (i) Training the neural network model
[0060] After filling a 100 ml beaker with 50 ml of 0.9 wt% salt water, magnetic stirring was performed at 600 rpm, and while the water was rotating, 2 g of superabsorbent polymer sample was added and video data was obtained showing the entire process of the polymer absorbing water and solidifying. At this time, video data was obtained for 16 types of superabsorbent polymer sample.
[0061] Next, the frames of the video data were separated to generate frame images, and the generated frame images were then classified into two types, pre-freezing and post-freezing, and labeled to obtain training data.
[0062] Next, a neural network model was trained using the labeled training data by applying CNN as an image classification algorithm.
[0063] (ii) Calculation of water absorption rate for sample (water-absorbent material)
[0064] For each of the superabsorbent resin samples 1 to 16, a 100 mL beaker was filled with 50 mL of 0.9 wt% salt water, and then magnetic stirring was performed at 600 rpm. While the water was rotating, 2 g of each of the samples 1 to 16 was added. Video data was obtained that captured the entire process of the samples absorbing water and solidifying, and the data was separated into frame images.
[0065] The frame images of each sample were sequentially input into the trained neural network model, the frame numbers of the images showing the start and end of water absorption were extracted, and the water absorption rate of each sample was calculated by performing a calculation according to Equation 2.
[0066] The calculated water absorption rates for Samples 1 to 16 are summarized in Table 1 below.
[0067] Comparative Example 1
[0068] In the comparative examples, similar to the above-mentioned examples, 50 mL of 0.9 wt % salt water was filled into a 100 mL beaker for each of the superabsorbent resin samples 1 to 16, and then magnetic stirring was performed at 600 rpm. While the water was rotating, 2 g of samples 1 to 16 were added and the mixture was stirred.
[0069] However, in the comparative examples, the researcher himself used a timer to measure the time during the process of water absorption and solidification. The timer was started when samples 1 to 16 were put into a beaker that was being stirred, and the timer was stopped when the water was absorbed and completely solidified and no further stirring was performed. The water absorption speed was measured based on the time measured by the timer.
[0070] The difference between the measured values of the water absorption rate according to the Examples and the Comparative Examples is shown in Table 1. The unit of the measured values for each Example and Comparative Example is seconds.
[0071] [Table 1]
[0072] As is clear from the above examples and comparative examples, the water absorption rate measured according to the present invention shows an average difference of 0.38 seconds when compared with the conventional method of measuring the water absorption rate of absorbent materials using a timer according to a manual. Considering that the conventional method is subject to operational errors depending on the researcher, it is recognized that the method according to the present invention sufficiently guarantees accuracy.
[0073] In addition, the method of the present invention has the advantage that the process of calculating the water absorption rate of the water absorbent material can be automated, and the water absorption rate can be accurately calculated for various samples using a standardized method.
[0074] (Example 2 / Comparative Example 2)
[0075] In Example 2, Sample 1 in Example 1 above was measured 10 times each by the method according to the present invention and the conventional manual measurement method.
[0076] The table below shows the results of measurements taken using each method.
[0077] [Table 2]
[0078] As is clear from Table 2 above, the method according to the present invention has a much smaller difference between the maximum and minimum values than the conventional method, and also has a smaller standard deviation, so it can be seen that the method according to the present invention is superior in terms of repeatability accuracy. [Explanation of symbols]
[0079] The reference numerals and names of the components used in the drawings and description of the present invention are as follows. 10…Water absorption video data acquisition section 20...Frame image data acquisition unit 30…Water absorption rate calculation section 31...Image discrimination unit (artificial neural network model) 32…Water absorption progress judgment part 33...Water absorption rate calculation section
Claims
1. A step of preparing learning data including image data of a predetermined water-absorbent material and classification data in which the image is labeled according to whether the image is an image in a water-absorbent state or an image that has already absorbed water; A step of constructing an artificial neural network that receives a predetermined image data and classifies the input image as a predetermined label; a machine learning step of learning the artificial neural network by inputting the learning data into the artificial neural network and updating parameters constituting the artificial neural network; 1. A method for generating an artificial neural network model, comprising:
2. The image data of the water absorbent material is An image data acquisition step of acquiring water absorption image data for a large number of water absorbent materials; a frame image data acquisition step of acquiring each frame image data constituting the video data; 2. The method for generating an artificial neural network model according to claim 1, wherein each of the frame image data is acquired using a
3. The water absorption image data is 3. The method for generating an artificial neural network model according to claim 2, wherein a predetermined amount of water is poured into a container, a predetermined amount of the water-absorbing material is added while stirring, and the process of the material absorbing water and solidifying is recorded as video data.
4. a water-absorbent material image data acquisition step of introducing the water-absorbent material and acquiring image data of the water-absorbent material for a predetermined time; a frame image data acquisition step of acquiring frame image data by separating the video data into frame image data; a frame image classification step of inputting the frame image data into an artificial neural network model to classify the frame images; a water absorption rate calculation step of extracting a water absorption end frame and a water absorption start frame from the classified frame images and calculating a water absorption rate using the extracted water absorption end frame and the water absorption start frame; A method for calculating the water absorption rate of a water absorbent material, comprising:
5. The artificial neural network model comprises:
5. A method for calculating the water absorption rate of a water-absorbent material according to claim 4, wherein the artificial neural network model is generated according to the method for generating an artificial neural network model according to any one of claims 1 to 3.
6. The method for calculating the water absorption rate of a water absorbent material according to claim 4, wherein the water absorption rate in the water absorption rate calculation step is calculated by the following mathematical formula 1: [Formula 1] Water absorption speed = (water absorption end frame number - water absorption start frame number) / fps of video data
7. In a system for calculating the water absorption rate of a water absorbent material, A video data acquisition unit for acquiring video data of the water absorbent material; a frame image data acquisition unit that separates the water absorption video data into frame image data constituting the video data and acquires the frame image data; a water absorption rate calculation unit that calculates a water absorption start time and a water absorption end time from the frame image data to calculate a water absorption rate of the water absorbent material; A water absorption rate calculation system for a water absorbent material comprising:
8. A frame image discrimination unit is further provided for discriminating whether each frame image data acquired by the frame image data acquisition unit is an image during absorption, in which the water absorbent material is absorbing water, or an image that has completed absorbing water, by inputting the frame image data acquired by the frame image data acquisition unit in frame order, The water absorption rate calculation unit is The water absorption rate calculation system for absorbent material as described in claim 7, wherein the frame image discrimination unit calculates the water absorption start time from the frame number (start frame number) of the image that is first discriminated as an image in which water has been absorbed, and calculates the water absorption end time from the frame number (end frame number) of the image that is first discriminated as an image that has already absorbed water.
9. The water absorption rate calculation system of claim 7, wherein the water absorption rate is calculated from the following formula 1. [Formula 1] Water absorption speed = (water absorption end frame number - water absorption start frame number) / fps of video data
10. The frame image discrimination unit The water absorption rate calculation system for absorbent materials as described in claim 8, further comprising an artificial neural network section that is pre-trained to receive input frame image data, classify the input frame image data into an image in the state of water absorption or an image that has already absorbed water, and output the result of the discrimination.
11. The artificial neural network unit includes: A water absorption rate calculation system for absorbent materials as described in claim 10, in which sample frame image data obtained by separating water absorption sample video data of a large number of sample absorbent materials into each sample frame image data and data labeled with the water-absorbed and absorbed states for each of the sample frame image data are input as learning data, and the system is trained to match the labeled data with each of the sample frame image data.
Citation Information
Patent Citations
Testing method and device for absorption characteristic of body fluid absorbent article
JP2004212060A
Measurement evaluation device and evaluation method using the same
JP2020051997A
