Chemical experiment precipitation state recognition algorithm

Abstract

The invention discloses a chemical experiment precipitation state recognition algorithm. The algorithm comprises the following steps: acquiring an original video of a solution; dividing the original video into video frame data which are not started in the experiment and video frame data which are obtained after the stirring and standing experiment; generating a dissolving and cleaning template according to the video frame data which is not started in the experiment; stirring and standing the solution to obtain video frame data after a stirring and standing experiment; comparing the video frame data after the stirring and standing experiment with the dissolving and cleaning template to obtain a crystal precipitation state; calculating the area of crystals separated from the solution through a semantic segmentation network; and identifying a crystal precipitation process through the area queue. According to the invention, through comparison of the real-time image and the predefined template, different stages of crystal precipitation are accurately identified. Dependence on artificial experience is reduced, non-contact continuous monitoring is realized, semantic segmentation is performed on the crystal precipitation area, and whether the precipitation process is completed or not is judged by comparing the area of the segmented area with the preset threshold value.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a chemical experiment precipitation state recognition algorithm. Background Art

[0002] In the field of chemical experiments, crystal precipitation is a physical and chemical process that occurs when the solubility of a substance changes with conditions such as temperature and solvent composition. When a solution reaches a supersaturated state, the solute molecules or ions will be arranged in order due to intermolecular forces, forming a crystal structure with a regular geometric shape. This process not only depends on the physical and chemical properties of the substance itself, such as the solubility curve and lattice energy, but is also closely related to external environmental parameters, such as the cooling rate, solvent evaporation rate, and stirring intensity. By controlling these variables, experimenters can promote the separation of solutes from the solution in the form of crystals, thereby achieving the separation and purification of the product.

[0003] Accurately monitoring the crystal precipitation process is of multi-dimensional critical significance to chemical experiments. From the perspective of determining the reaction endpoint, the beginning and complete precipitation of crystals often mark the process nodes of the chemical reaction. Accurately capturing these nodes can avoid over- or under-reaction and ensure the yield of the target product. In terms of product purity control, slight changes in crystallization conditions may lead to the mixing of different crystal forms or impurities. Real-time monitoring of the crystallization state helps to adjust experimental parameters in a timely manner, inhibit the formation of by-products, and improve crystal purity. In addition, for experimental projects that require repeated verification, a standardized crystallization monitoring process is the basis for ensuring the repeatability of experimental results and helps to build a reliable experimental data system.

[0004] Since crystal precipitation usually requires a long kinetic process, experimenters need to continuously observe changes in the solution state. This long-term manual monitoring not only takes up a lot of human resources, but may also lead to monitoring omissions due to fatigue. Secondly, there is the risk of error caused by subjective judgment. Different experimenters have individual differences in the visual recognition of states such as "beginning of precipitation" and "complete crystallization". The lack of quantitative judgment standards may cause deviations in experimental results between different batches, affecting the reliability and comparability of the data. In addition, existing monitoring methods mostly rely on manual observation, and it is difficult to record the dynamic changes of key parameters such as temperature and concentration in real time. It is impossible to provide complete data support for the kinetic analysis of the crystallization process, which restricts the in-depth study of the crystallization mechanism and process optimization. Summary of the Invention

[0005] The present invention provides an algorithm for identifying precipitation states in chemical experiments. This algorithm addresses the problem that the crystal precipitation process in existing techniques is often time-consuming and requires close observation by experimenters to accurately capture changes in crystallization conditions, increasing labor costs and complexity. Due to individual differences, different experimenters may use inconsistent criteria for determining the crystallization state, thus affecting the repeatability and reliability of experimental results.

[0006] On the one hand, an embodiment of the present invention provides a chemical experiment precipitation state recognition algorithm, including:

[0007] Collect original video of the solution;

[0008] The original video is divided into video frame data before the experiment starts and video frame data after the stirring and standing experiment;

[0009] Generate a dissolution template based on the video frame data before the experiment starts;

[0010] Stirring and allowing the solution to stand to obtain video frame data after the stirring and standing experiment;

[0011] Comparing the video frame data after the stirring and standing experiment with the dissolution template to obtain the crystal precipitation state;

[0012] Calculating the area of ​​crystals precipitated from the solution through a semantic segmentation network;

[0013] The precipitation progress of the crystals was identified by area alignment.

[0014] In a possible implementation, generating a dissolution template based on the video frame data before the experiment starts includes:

[0015] Mark the ROI area of ​​the video frame data before the experiment starts;

[0016] A lysis template is generated according to the ROI area.

[0017] In a possible implementation, comparing the video frame data after the stirring and standing experiment with the dissolution template to obtain the crystal precipitation state includes:

[0018] Mark the ROI area of ​​the video frame data after the stirring and standing experiment;

[0019] The crystal precipitation state is obtained by comparing the ROI area of ​​the video frame data after the stirring and standing experiment with the dissolution template.

[0020] In a possible implementation, when the crystal precipitation state is determined to be a precipitation state, the solution is stirred for T seconds;

[0021] After the stirring, the solution is allowed to stand for T seconds to obtain crystal precipitation.

[0022] In a possible implementation, calculating the area of ​​crystals precipitated from the solution using a semantic segmentation network includes:

[0023] Performing semantic segmentation on the video frame data after the stirring and standing experiment in which crystals are present according to the semantic segmentation network to obtain a target crystal region;

[0024] The area of ​​the crystal is calculated based on the segmented target crystal region.

[0025] In a possible implementation, identifying the precipitation process of the crystal by using the area queue includes:

[0026] Set the length of the area queue;

[0027] Adding the area value of the crystal to the end of the area queue and removing the head of the area queue;

[0028] The precipitation process is identified based on the continuity comparison of the values ​​of the area queue.

[0029] In a possible implementation, identifying the precipitation process based on continuity comparison of values ​​of the area queue includes:

[0030] The existence status of the area queue is confirmed according to the difference between the N consecutive values ​​of the area and a preset threshold.

[0031] The chemical experiment precipitation state recognition algorithm of the present invention has the following advantages:

[0032] (1) By comparing real-time images with predefined templates, the different stages of crystal precipitation can be accurately identified. This method reduces the reliance on manual experience and achieves contactless continuous monitoring.

[0033] (2) Semantically segment the region where crystals are precipitated, and then determine whether the precipitation process is complete by comparing the area of ​​the segmented region with a preset threshold. This method provides an accurate and automated means to determine the precipitation endpoint. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1A flowchart of a chemical experiment precipitation state identification algorithm provided in an embodiment of the present application. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0037] Figure 1 A schematic diagram of a flow chart of a chemical experiment precipitation state recognition algorithm provided by an embodiment of the present invention; an embodiment of the present invention provides a chemical experiment precipitation state recognition algorithm, including:

[0038] Collect original video of the solution;

[0039] The original video is divided into video frame data before the experiment starts and video frame data after the stirring and standing experiment;

[0040] Generate a dissolution template based on the video frame data before the experiment starts;

[0041] Stirring and allowing the solution to stand to obtain video frame data after the stirring and standing experiment;

[0042] Comparing the video frame data after the stirring and standing experiment with the dissolution template to obtain the crystal precipitation state;

[0043] Calculating the area of ​​crystals precipitated from the solution through a semantic segmentation network;

[0044] The precipitation progress of the crystals was identified by area alignment.

[0045] Generating a dissolution template according to the video frame data before the experiment begins includes:

[0046] Mark the ROI area of ​​the video frame data before the experiment starts;

[0047] A lysis template is generated according to the ROI area.

[0048] The comparing the video frame data after the stirring and standing experiment with the dissolution template to obtain the crystal precipitation state includes:

[0049] Mark the ROI area of ​​the video frame data after the stirring and standing experiment;

[0050] The crystal precipitation state is obtained by comparing the ROI area of ​​the video frame data after the stirring and standing experiment with the dissolution template.

[0051] When the crystal precipitation state is determined to be a precipitation state, stirring the solution for T seconds;

[0052] After the stirring, the solution is allowed to stand for T seconds to obtain crystal precipitation.

[0053] Calculating the area of ​​crystals precipitated from the solution by using a semantic segmentation network includes:

[0054] Performing semantic segmentation on the video frame data after the stirring and standing experiment in which crystals are present according to the semantic segmentation network to obtain a target crystal region;

[0055] The area of ​​the crystal is calculated based on the segmented target crystal region.

[0056] The step of identifying the precipitation process of the crystals by using an area queue comprises:

[0057] Set the length of the area queue;

[0058] Adding the area value of the crystal to the end of the area queue and removing the head of the area queue;

[0059] The precipitation process is identified based on the continuity comparison of the values ​​of the area queue.

[0060] The identifying the precipitation process according to the continuity comparison of the values ​​of the area queue includes:

[0061] The existence status of the area queue is confirmed according to the difference between the N consecutive values ​​of the area and a preset threshold.

[0062] Exemplarily, the solution in the reaction container is captured and stored in real time by a high-definition camera (resolution ≥ 1080p, frame rate ≥ 30fps), and then the original video is extracted frame by frame by a video streaming algorithm.

[0063] The video frame data before the experiment started is the unprecipitated video frame data. For this part of the data, the region of interest (ROI) is determined by manual annotation. In the specific operation, the main area of ​​the solution in the reaction vessel is outlined through a graphical interface, where the marked area is the ROI, and the unmarked part is the background information. Subsequently, the ROI area in the video frame is extracted by removing the background data. Based on the extracted unprecipitated video frame ROI image data, a template for dissolving and clearing recognition is constructed.

[0064] When the experiment begins, a fixed amount of stirring and standing is performed. This is the video frame data for the precipitation process. For this portion of the data, a region of interest (ROI) is manually labeled. Specifically, a graphical interface is used to outline the main area of ​​the solution in the reaction vessel. The labeled area is the ROI, and the unlabeled area is considered background information. Subsequently, the background data is removed and the ROI area in the video frame is extracted.

[0065] The extracted ROI image data from the precipitation process video frame is compared with the previously constructed dissolution template. The comparison method is to perform a pixel-by-pixel comparison between the ROI area of ​​the current frame and the dissolution template and calculate the pixel difference matrix. The mean square error (MSE) or structural similarity index (SSIM) can be used as a quantitative evaluation indicator. When the difference exceeds a preset threshold, it is determined that crystal precipitation has occurred. The calculation formula for MSE is as follows.

[0066]

[0067] Where I(x, y) is the pixel value of the ROI in the current frame, T(x, y) is the pixel value corresponding to the dissolving template, W and H are the width and height of the ROI, and N = W × H.

[0068] Then, perform stirring and standing control. When the crystal precipitation result is "crystals present", start the stirring device. The stirring time T seconds can be set according to the specific characteristics of the solution (e.g., 10-30 seconds). The stirring speed should be uniform to avoid violent shaking that may cause crystal breakage.

[0069] After stirring, stop stirring and let the solution stand for T seconds to allow the crystals to fully settle. During the standing period, keep the environment stable and avoid vibration.

[0070] Crystal area calculation is based on semantic segmentation, using a DeepLabv3+ semantic segmentation network with an improved network architecture. Training is performed on a pre-annotated crystal image dataset (including annotations of crystalline and non-crystalline regions). The dataset should include crystal images at different concentrations and temperatures to improve model generalization.

[0071] Calculate the area of ​​the precipitated crystals obtained by semantic segmentation using the following formula:

[0072]

[0073] Among them, I c (x,y) is the segmented crystal region. C is the ROI region of the video frame image data, W and H are the width and height of the video frame image data, respectively. R is the area ratio.

[0074] Then, based on the calculation results, if the area ratio of the precipitated crystals is greater than the preset minimum area ratio threshold, the area of ​​the precipitated crystals in the current frame will be added to the area queue; if the area of ​​the precipitated crystals is less than the preset minimum threshold, the process will be restarted by returning to obtain the video frame data after the stirring and standing experiment.

[0075] The total length of the area queue is N. When the area queue is full, a new area is added and the first old area is discarded according to the first-in-first-out principle.

[0076] When the error between the areas of N consecutive precipitated crystals in the area queue is less than a specified threshold, it is determined that crystal precipitation is complete and the precipitation end status is sent. If the error between the areas of N consecutive precipitated crystals is greater than the threshold or the entire area queue is less than N, the process returns to obtaining the video frame data after the stirring and standing experiment and restarts the process.

[0077] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including all changes and modifications that fall within the scope of the present invention and the preferred embodiments.

[0078] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A chemical experiment precipitation state recognition algorithm, characterized in that: include: Collect original video of the solution; The original video is divided into video frame data before the experiment starts and video frame data after the stirring and standing experiment; Generate a dissolution template based on the video frame data before the experiment starts; Stirring and allowing the solution to stand to obtain video frame data after the stirring and standing experiment; Comparing the video frame data after the stirring and standing experiment with the dissolution template to obtain the crystal precipitation state; Calculating the area of ​​crystals precipitated from the solution through a semantic segmentation network; The precipitation progress of the crystals was identified by area alignment.

2. A chemical experiment precipitation state recognition algorithm according to claim 1, characterized in that: The generating of the dissolving template according to the video frame data before the experiment starts includes: Mark the ROI area of ​​the video frame data before the experiment starts; A lysis template is generated according to the ROI area.

3. A chemical experiment precipitation state recognition algorithm according to claim 1, characterized in that: The comparing the video frame data after the stirring and standing experiment with the dissolution template to obtain the crystal precipitation state includes: Mark the ROI area of ​​the video frame data after the stirring and standing experiment; The crystal precipitation state is obtained by comparing the ROI area of ​​the video frame data after the stirring and standing experiment with the dissolution template.

4. A chemical experiment precipitation state recognition algorithm according to claim 1, characterized in that: When the crystal precipitation state is determined to be a precipitation state, stirring the solution for T seconds; After the stirring, the solution is allowed to stand for T seconds to obtain crystal precipitation.

5. A chemical experiment precipitation state recognition algorithm according to claim 1, characterized in that: Calculating the area of ​​crystals precipitated from the solution by using a semantic segmentation network includes: Performing semantic segmentation on the video frame data after the stirring and standing experiment in which crystals are present according to the semantic segmentation network to obtain a target crystal region; The area of ​​the crystal is calculated based on the segmented target crystal region.

6. A chemical experiment precipitation state recognition algorithm according to claim 5, characterized in that: The step of identifying the crystal precipitation process by area queue includes: Set the length of the area queue; Adding the area value of the crystal to the end of the area queue and removing the head of the area queue; The precipitation process is identified based on the continuous comparison of the values ​​of the area queue.

7. A chemical experiment precipitation state recognition algorithm according to claim 6, characterized in that: The identifying the precipitation process according to the continuity comparison of the values ​​of the area queue includes: The existence status of the area queue is confirmed according to the difference between the N consecutive values ​​of the area and a preset threshold.

Images