Liquid level following method and system for pipettor

Abstract

The invention relates to the technical field of pipettors, and provides a pipettor liquid level following method and system, and the method comprises the steps: controlling a pipettor to move along a vertical shaft to a liquid container; controlling the optical sensor to emit light to the surface of liquid in the liquid container, and generating first detection data based on a received light signal; generating second detection data on the basis of capacitance change generated when the dielectric property of liquid in the liquid container is different from that of surrounding air measured by the capacitance sensor; measuring the pressure change of the suction head of the pipettor at different depths in the liquid of the liquid container based on a pressure sensor, and generating third detection data; inputting the first detection data, the second detection data and the third detection data into a multi-modal fusion model to obtain the liquid type and the liquid level position of the liquid in the liquid container; and executing the pipetting operation of the pipettor based on the liquid type and the liquid level position. According to the invention, the liquid level position can be determined more accurately, and the pipetting volume error caused by liquid property difference is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of pipettes, and in particular to a pipette liquid surface following method and a system thereof. Background Art

[0002] In modern biological and chemical laboratory operations, pipettes are widely used as precise pipetting tools. The liquid level following function of pipettes is essential for accurate liquid transfer, especially when dealing with liquids of different densities and viscosities and liquids in containers of different shapes.

[0003] At present, the main method for following the liquid level of the pipette is based on the fixed distance detection method, which is to set a fixed distance sensor near the pipette tip to detect the liquid level. When the tip drops close to the liquid level, the sensor senses the signal change reflected or refracted by the liquid surface, thereby triggering the stop of the liquid suction operation. However, the above method has poor accuracy for liquids with different liquid densities and surface tensions. Due to the different optical properties (such as refractive index) and surface characteristics (such as the shape of the meniscus caused by surface tension) of different liquids, the setting of a fixed distance may cause the liquid to be stopped too early or too late when sucking certain liquids, resulting in inaccurate pipetting volume. For example, for high-viscosity liquids, the surface tension is large and the meniscus is high, which may trigger the sensor before the tip is fully immersed in the appropriate depth, resulting in insufficient volume of liquid sucked; while for low-viscosity, easy-flowing liquids, they may be over-immersed and suck too much liquid. Summary of the invention

[0004] The present invention provides a liquid surface following method and system of a pipette, which are used to accurately determine the liquid surface position and reduce the error of the liquid transfer volume caused by the difference of liquid properties.

[0005] In a first aspect, the present invention provides a pipette liquid level following method, which is applied to a pipette, wherein the pipette tip includes an optical sensor, a capacitive sensor and a pressure sensor, and the method includes:

[0006] In response to a liquid level detection instruction of the pipette, controlling the pipette to move along the vertical axis toward the liquid container;

[0007] Controlling the optical sensor to emit light to the surface of the liquid in the liquid container, and generating first detection data based on the received light signal; the first detection data represents the change of the optical property of the liquid surface;

[0008] Generate second detection data based on the capacitance change caused by the capacitance sensor measuring the difference between the dielectric properties of the liquid in the liquid container and the surrounding air;

[0009] Generate third detection data based on the change of pressure applied to the pipette tip when the pipette tip is at different depths in the liquid of the liquid container measured by the pressure sensor;

[0010] Inputting the first detection data, the second detection data and the third detection data into a multimodal fusion model, and obtaining the multimodal fusion model to output the liquid type and liquid level position of the liquid in the liquid container; the multimodal fusion model is trained based on sample fusion features and their corresponding liquid type labels and liquid level position labels;

[0011] The pipette aspiration operation is performed based on the liquid type and the liquid surface position, so that the pipette tip follows the liquid surface to complete the pipetting.

[0012] According to the pipette liquid level following method provided by an embodiment of the present invention, the multimodal fusion model includes a first feature extraction layer, a second feature extraction layer, a feature fusion layer and a fuzzy prediction layer; the first feature extraction layer includes a plurality of convolution layers with different convolution kernel sizes, a pooling layer and a fully connected layer; the second feature extraction layer includes a plurality of hidden units;

[0013] The step of inputting the first detection data, the second detection data, and the third detection data into a multimodal fusion model to obtain a liquid type and a liquid level position of the liquid in the liquid container output by the multimodal fusion model includes:

[0014] Inputting the first detection data, the second detection data, and the third detection data into the multimodal fusion model, performing feature extraction on the first detection data based on the first feature extraction layer, and obtaining optical features output by the first feature extraction layer;

[0015] Based on the second feature extraction layer, feature extraction is performed on the second detection data and the third detection data respectively, so as to obtain capacitance features and pressure features output by the second feature extraction layer;

[0016] Based on the feature fusion layer, the optical feature, the capacitance feature and the pressure feature are feature spliced ​​to obtain a fusion feature output by the feature fusion layer;

[0017] The fusion feature is predicted based on the fuzzy prediction layer in combination with a fuzzy prediction algorithm to obtain the liquid type and liquid level position output by the fuzzy prediction layer.

[0018] According to the pipette liquid level following method provided by an embodiment of the present invention, the fusion feature is predicted based on the fuzzy prediction layer in combination with the fuzzy prediction algorithm to obtain the liquid type and liquid level position output by the fuzzy prediction layer, including:

[0019] The fuzzy prediction layer performs the following steps:

[0020] The fused feature is converted into a corresponding fused feature vector F; wherein the fused feature vector F is expressed as F=(f1, f2, f3), f1 represents an optical feature, f2 represents a capacitance feature, and f3 represents a pressure feature;

[0021] The fused feature vector F is mapped to a high-dimensional feature space through a high-dimensional mapping function to obtain the mapping of the fused feature vector F in the high-dimensional feature space Among them, the mapping of the fusion feature vector F in the high-dimensional feature space It is expressed as: in, represents a high-dimensional mapping function, a 1i ,a 2i ,b i ,c i Represents the parameters learned through training data;

[0022] Perform hierarchical clustering on the feature space after high-dimensional mapping to obtain K cluster centers C k ,k=1,2,...,K,C k =(C1,C2,...,C K ), calculate the mapping of each fused feature vector in high-dimensional space The distance d to each cluster center k (F), distance d k (F) is expressed as:

[0023] According to the mapping of each fused feature vector in high-dimensional space The distance d to each cluster center k (F), determine the membership function μ k (F), membership function μ k (F) is expressed as:

[0024]

[0025] Among them, the membership function μ k (F) indicates the degree to which the fused feature vector belongs to each cluster;

[0026] The manifold learning algorithm is used to analyze the sample fusion features in the training data and their corresponding liquid types and liquid surface positions, and fuzzy rules are generated according to the local neighborhood relationship on the manifold; the weight of the fuzzy rule ω k The weight ω is determined based on the sample density of the local area on the manifold. k It is expressed as:

[0027] n krepresents the number of samples in the local area corresponding to the kth fuzzy subset on the manifold, and N represents the total number of samples;

[0028] Based on the membership function μ k (F) and weight ω k A prediction is performed to obtain the liquid type and liquid level position output by the fuzzy prediction layer.

[0029] According to the pipette liquid level following method provided by an embodiment of the present invention, the K cluster centers C k , the steps to determine are as follows:

[0030] Each fused feature vector after high-dimensional mapping is taken as a separate cluster, where the number of initial clusters is equal to the number of samples N;

[0031] Cluster merging criterion: Calculate Q for each pair of clusters a and clustering Q b The distance d(Q a ,Q b ), distance d(Q a ,Q b )The specific calculation formula is as follows:

[0032]

[0033] Among them, F i and F j Represents the two fused feature vectors after high-dimensional mapping, d ij Represents two fused feature vectors F i and F j The distance between them, K(·) represents the Gaussian kernel function, K(F i ,F j ) represents the fusion feature vector F i and the fusion feature vector F j The kernel function value between represents the high-dimensional mapping function, σ represents the kernel bandwidth parameter, |Q a |and|Q b | respectively represent cluster Q a and the number of samples in cluster Q;

[0034] In each iteration, the two clusters with the closest distance are found according to the cluster merging criterion, and the two clusters are merged into a new cluster Q new , the center of the new cluster C new The calculation formula is as follows:

[0035] Among them, C a Represents cluster Q a The center of C b Represents cluster Cb center.

[0036] Repeat the above merging process to continuously reduce the number of clusters until the predetermined number of clusters K is reached. The center C of the K clusters finally obtained is k =(C1,C2,...,C K ).

[0037] According to the pipette liquid level following method provided by an embodiment of the present invention, the total loss function of the modal fusion model includes a feature reconstruction loss function, a fuzzy prediction error loss function, a fuzzy rule consistency loss function and a multimodal feature interaction loss function; the feature reconstruction loss function includes an optical feature reconstruction loss function, a capacitance feature reconstruction loss function and a pressure feature reconstruction loss function; the fuzzy prediction error loss function includes a liquid type prediction loss function and a liquid level position prediction loss function.

[0038] According to the pipette liquid level following method provided by an embodiment of the present invention, the optical feature reconstruction loss function L recon-opt It is expressed as:

[0039]

[0040] Where W(·) represents the Wasserstein distance between two distributions, n represents the total number of training samples, and O i and represent the original optical features and reconstructed optical features of the i-th sample respectively;

[0041] The capacitance feature reconstruction loss function L recon-cap It is expressed as:

[0042]

[0043] Among them, C i and Represent the original capacitance feature and reconstructed capacitance feature of the i-th sample respectively;

[0044] The pressure feature reconstruction loss function L recon-pres It is expressed as:

[0045]

[0046] Among them, P i and Represent the original pressure feature and reconstructed pressure feature of the i-th sample respectively;

[0047] The liquid type prediction loss function L pres-type It is expressed as:

[0048]

[0049] Where J represents the number of categories or attributes related to liquid types in the fuzzy rules, K represents the number of fuzzy subsets, represents the membership of the i-th sample to the j-th category attribute and the k-th fuzzy subset, represents the predicted value of the liquid type of the i-th sample under the j-th category attribute and the k-th fuzzy subset relationship, It represents the predicted value of the liquid type of the i-th sample under the j-th category attribute and the l-th fuzzy subset relationship;

[0050] The liquid level position prediction loss function L pres-pos It is expressed as:

[0051]

[0052] in, represents the membership of the i-th sample to the k-th fuzzy subset, and represents the values ​​of the predicted liquid level position and the true liquid level position in the j-th dimension respectively;

[0053] The fuzzy rule consistency loss function L rule-cons It is expressed as:

[0054]

[0055] Among them, ω k represents the weight of the kth fuzzy rule, and represents the activation strength of the i-th sample and the j-th sample for the k-th fuzzy rule respectively;

[0056] The multimodal feature interaction loss function L interact It is expressed as:

[0057]

[0058] Where MI(·) represents the calculation of the mutual information between two random variables based on kernel density estimation; F i Represents the fused feature vector of the i-th sample;

[0059] The total loss function represents L all for:

[0060] L all =L recon-opt +L recon-cap +L recon-pres +L pres-type +L pres-pos +L rule-cons +L interact .

[0061] According to the liquid level following method for a pipette provided by an embodiment of the present invention, the liquid aspiration operation of the pipette is performed based on the liquid type and the liquid level position, comprising:

[0062] Determining the suction force and suction speed of the pipette based on the type of the liquid;

[0063] adjusting the pipette to a target position based on the liquid level position;

[0064] The liquid suction operation is performed on the liquid in the liquid container at the target position with the suction force magnitude and the suction speed.

[0065] In a second aspect, the present invention further provides a pipette liquid level following system, which is applied to a pipette, wherein the pipette tip includes an optical sensor, a capacitive sensor and a pressure sensor, and the device includes:

[0066] A control module, for responding to a liquid level detection instruction of the pipette, and controlling the pipette to move along a vertical axis toward the liquid container;

[0067] An optical data acquisition module, used to control the optical sensor to emit light to the surface of the liquid in the liquid container, and generate first detection data based on the received light signal; the first detection data represents the change of the optical property of the liquid surface;

[0068] A capacitance data acquisition module, used for measuring the capacitance change caused by the difference between the dielectric properties of the liquid in the liquid container and the surrounding air based on the capacitance sensor, to generate second detection data;

[0069] A pressure data acquisition module, used for measuring the change of the pressure exerted on the pipette tip at different depths in the liquid of the liquid container based on the pressure sensor, and generating third detection data;

[0070] A prediction module, used for inputting the first detection data, the second detection data and the third detection data into a multimodal fusion model, so as to obtain the liquid type and liquid level position of the liquid in the liquid container output by the multimodal fusion model; the multimodal fusion model is trained based on sample fusion features and their corresponding liquid type labels and liquid level position labels;

[0071] The liquid level following module is used to perform the pipetting operation of the pipette based on the liquid type and the liquid level position, so that the pipette tip of the pipette follows the liquid level to complete the pipetting.

[0072] In a third aspect, the present invention further provides an electronic device, comprising: a memory for storing a computer software program; a processor for reading and executing the computer software program, thereby implementing any of the above-mentioned pipette liquid surface following methods.

[0073] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium, wherein the storage medium stores a computer software program, and when the computer software program is executed by a processor, the pipette liquid surface following method as described above is implemented.

[0074] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described pipette liquid surface following methods.

[0075] The pipette liquid level following method provided in the embodiment of the present invention overcomes the problem that a single sensor is greatly affected by the specific properties of the liquid through multiple sensors, and is combined with first detection data of changes in optical properties of the surface of the liquid in the liquid container provided by the optical sensor, second detection data of capacitance changes caused by the difference in dielectric properties of the liquid in the liquid container and the surrounding air, and third detection data of changes in pressure on the pipette tip at different depths in the liquid in the liquid container, combined with a multimodal fusion model, to more accurately determine the liquid level position and reduce the pipetting volume error caused by differences in liquid properties. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 is a flow chart of a pipette liquid level following method provided by an embodiment of the present invention;

[0077] Figure 2 is a structural diagram of a liquid level following system for a pipette provided in an embodiment of the present invention;

[0078] Figure 3 An embodiment diagram of an electronic device provided by an embodiment of the present invention;

[0079] Figure 4 An embodiment diagram of a computer-readable storage medium provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0080] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0081] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0082] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.

[0083] Optional, see Figure 1 , Figure 1 : is a flow chart of the pipette liquid level following method provided by the present invention. The execution subject of the pipette liquid level following method in the embodiment of the present invention is a pipette. The pipette tip includes an optical sensor, a capacitive sensor and a pressure sensor. The optical sensor is used to detect the change of light reflection or refraction characteristics of the interface between the liquid and the air. The capacitive sensor is used to detect the change of capacitance caused by the change of the dielectric constant of the liquid. The pressure sensor is used to detect the pressure change caused by the static pressure of the liquid at the pipette tip. Therefore, the pipette liquid level following method in the embodiment of the present invention includes:

[0084] Step 10, responding to the liquid level detection instruction of the pipette, controlling the pipette to move along the vertical axis toward the liquid container.

[0085] When it is necessary to control the pipette to follow the liquid level of the liquid in the liquid container, it is necessary to send a liquid level detection instruction to the pipette. At this time, the pipette can be in a stopped state or in a running state. If the pipette is in a stopped state, after receiving the liquid level detection instruction of the pipette, the pipette is turned on, and the pipette is controlled to move along the vertical axis (Z direction) toward the liquid container in response to the liquid level detection instruction of the pipette. If the pipette is in a running state, the pipette is directly responded to the liquid level detection instruction of the pipette, and the pipette is controlled to move along the vertical axis (Z direction) toward the liquid container.

[0086] Step 20: Control the optical sensor to emit light to the surface of the liquid in the liquid container, and generate first detection data based on the received light signal.

[0087] Furthermore, the optical sensor emits light to the surface of the liquid in the liquid container, and after the light contacts the surface of the liquid, it is reflected or refracted toward the optical sensor. Therefore, the optical sensor receives the light reflected or refracted by the surface of the liquid, and generates the first detection data according to the change of the light. Therefore, the first detection data in the embodiment of the present invention reflects the change of the optical properties near the surface of the liquid.

[0088] Step 30, generating second detection data based on the capacitance sensor measuring the capacitance change caused by the difference in dielectric properties of the liquid in the liquid container and the surrounding air.

[0089] Furthermore, the capacitive sensor forms a capacitive coupling with the liquid in the liquid container, and measures the capacitance change caused by the difference in dielectric properties between the liquid in the liquid container and the surrounding air through the capacitive coupling, thereby generating second detection data.

[0090] Step 40: Generate third detection data by measuring the change in pressure exerted on the tip of the pipette at different depths in the liquid of the liquid container based on the pressure sensor.

[0091] Furthermore, the pressure sensor measures the change in pressure to which the pipette tip is subjected when the pipette tip is at different depths in the liquid of the liquid container, and generates third detection data.

[0092] Step 50, input the first detection data, the second detection data and the third detection data into the multimodal fusion model, and obtain the multimodal fusion model to output the liquid type and liquid level position of the liquid in the liquid container.

[0093] Optionally, a pre-trained multimodal fusion model is embedded in the pipette of an embodiment of the present invention, or a pre-trained multimodal fusion model in an external system is called through a communication protocol. The multimodal fusion model is trained based on sample fusion features and their corresponding liquid type labels and liquid level position labels.

[0094] Therefore, the first detection data, the second detection data and the third detection data are input into the multimodal fusion model, and the multimodal fusion model outputs the liquid type and the liquid level position of the liquid in the liquid container.

[0095] Step 60, performing a pipette aspiration operation based on the liquid type and the liquid level position, so that the pipette tip follows the liquid level to complete the pipetting.

[0096] Further, the pipette aspiration operation is performed according to the liquid type and the liquid level position, so that the pipette tip follows the liquid level to complete the pipetting.

[0097] The embodiment of the present invention overcomes the problem that a single sensor is greatly affected by the specific properties of the liquid through the use of multiple sensors, and combines the first detection data of the change in optical properties of the surface of the liquid in the liquid container provided by the optical sensor, the second detection data of the capacitance change caused by the difference in dielectric properties of the liquid in the liquid container and the surrounding air, and the third detection data of the change in pressure on the pipette tip at different depths in the liquid in the liquid container, with a multimodal fusion model, to more accurately determine the liquid level position and reduce the pipetting volume error caused by differences in liquid properties.

[0098] In one embodiment, the multimodal fusion model includes a first feature extraction layer, a second feature extraction layer, a feature fusion layer and a fuzzy prediction layer; the first feature extraction layer includes a plurality of convolution layers with different convolution kernel sizes, a pooling layer and a fully connected layer; the second feature extraction layer includes a plurality of hidden units, wherein the plurality of convolution layers with different convolution kernel sizes include a convolution layer with a convolution kernel of 3*3, a convolution layer with a convolution kernel of 5*5 and a convolution layer with a convolution kernel of 7*7. Therefore, step 50 inputs the first detection data, the second detection data and the third detection data into the multimodal fusion model, and obtains the liquid type and liquid level position of the liquid in the liquid container output by the multimodal fusion model, including:

[0099] Step 501, inputting the first detection data, the second detection data and the third detection data into the multimodal fusion model, performing feature extraction on the first detection data based on the first feature extraction layer, and obtaining the optical features output by the first feature extraction layer;

[0100] Step 502, performing feature extraction on the second detection data and the third detection data based on the second feature extraction layer, respectively, to obtain capacitance features and pressure features output by the second feature extraction layer;

[0101] Step 503, based on the feature fusion layer, the optical features, the capacitance features and the pressure features are feature spliced ​​to obtain the fusion features output by the feature fusion layer.

[0102] Optionally, the first detection data, the second detection data and the third detection data are input into the multimodal fusion model, and the convolution layer with a convolution kernel of 3*3, the convolution layer with a convolution kernel of 5*5 and the convolution layer with a convolution kernel of 7*7 in the first feature extraction layer respectively extract features of the first detection data, and further the features output by multiple convolution layers with different convolution kernel sizes are fused, and after the pooling layer and the fully connected layer, the optical features output by the first feature extraction layer are obtained.

[0103] Further, the plurality of hidden units in the second feature extraction layer perform multiple feature processing on the second detection data and the third detection data respectively to obtain the capacitance feature and the pressure feature output by the second feature extraction layer. The feature fusion layer performs feature splicing on the optical feature, the capacitance feature and the pressure feature to obtain the fusion feature output by the feature fusion layer, such as splicing the optical feature, the capacitance feature and the pressure feature in sequence.

[0104] Step 504 , predicting the fusion features based on the fuzzy prediction layer in combination with the fuzzy prediction algorithm, and obtaining the liquid type and liquid level position output by the fuzzy prediction layer.

[0105] Furthermore, the fuzzy prediction layer combines the fuzzy prediction algorithm to predict the fusion features and obtain the liquid type and liquid level position output by the fuzzy prediction layer. The specific analysis is as follows:

[0106] The fused features are converted into corresponding fused feature vectors F. The fused feature vector F can be expressed as F = (f1, f2, f3), where f1 represents optical features, f2 represents capacitance features, and f3 represents pressure features.

[0107] Furthermore, through the high-dimensional mapping function Map the fused feature vector F to a high-dimensional feature space to obtain the mapping of the fused feature vector F in the high-dimensional feature space Among them, the mapping of the fusion feature vector F in the high-dimensional feature space It can be expressed as:

[0108]

[0109] Among them, a 1i ,a 2i ,b i ,c i Represents the parameters learned through training data.

[0110] Furthermore, the feature space after high-dimensional mapping is hierarchically clustered to obtain K cluster centers C k ,k=1,2,...,K, for the mapping of each fused feature vector in high-dimensional space Calculate the distance d to each cluster center k (F), can be at a distance d k (F) is expressed as:

[0111] Among them, C k represents the kth cluster center.

[0112] Therefore, the membership function μ k (F) can be expressed as:

[0113] Among them, the membership function μ k (F) represents the degree to which the fused feature vector belongs to each cluster (fuzzy subset).

[0114] In one embodiment, K cluster centers C k , the steps to determine are as follows:

[0115] Each fused feature vector after high-dimensional mapping is taken as a separate cluster. Therefore, the number of initial clusters is equal to the number of samples N, and the center of each cluster is the corresponding fused feature vector itself.

[0116] Furthermore, the cluster merging criterion is: calculate each pair of clusters (such as cluster Q a and clustering Q b ) between a ,Q b ), distance d(Q a ,Q b )The specific calculation formula is as follows:

[0117]

[0118] Among them, F i and F j Represents the two fused feature vectors after high-dimensional mapping, d ij Represents two fused feature vectors F i and F j The distance between them, K(·) represents the Gaussian kernel function, K(F i ,F j ) represents the fusion feature vector F i and the fusion feature vector F j The kernel function value between represents the high-dimensional mapping function, σ represents the kernel bandwidth parameter, |Q a |and|Q b | respectively represent cluster Q a and clustering Q b The number of samples in .

[0119] Furthermore, in each iteration, the two closest clusters are found according to the cluster merging criterion, and the two clusters are merged into a new cluster Q new , the center of the new cluster C new The calculation formula is as follows:

[0120] Among them, C a Represents cluster Q a The center of C b Represents cluster C b center.

[0121] Repeat the above merging process to continuously reduce the number of clusters until the predetermined number of clusters K is reached. The center C of the K clusters finally obtained is k =(C1,C2,...,C K ).

[0122] Furthermore, the sample fusion features in the training data and their corresponding liquid types and liquid surface positions are analyzed using manifold learning algorithms (such as local linear embedding, LLE). In the low-dimensional manifold space, the intrinsic geometric structure of the data is found. Fuzzy rules are generated based on the local neighborhood relationship on the manifold. For example, if in the manifold space, most of the samples near a set of fusion features correspond to liquid type T and liquid surface position in the interval [p1, p2], then a fuzzy rule is generated as "If the fusion feature is in the fuzzy subset corresponding to the local manifold area, then the liquid type is T and the liquid surface position is [p1, p2]".

[0123] The weight of the fuzzy rule ω k Determined based on the sample density of the local area on the manifold. Let n k is the number of samples in the local area corresponding to the kth fuzzy subset on the manifold. The total number of samples is N, then the weight ω k It can be expressed as

[0124] For each fuzzy rule k, calculate its activation strength a k :

[0125]

[0126] Where L represents the number of different levels of clustering or feature space partitioning used to generate fuzzy rules, μ kl (F) represents the membership of the fused feature vector to the kth fuzzy subset under the lth partition.

[0127] In one embodiment, there are T types of liquid types, and the liquid level position is a continuous value. For the fuzzy prediction of the liquid type, a T*k matrix M is constructed. L , where the element M Lij =a k *I ij , I ij represents the indicator function. If the liquid type corresponding to the kth fuzzy rule is the i-th type, then I ij =1, otherwise I ij =0.

[0128] For the fuzzy prediction of the liquid level position, construct a vector v p , whose element v pk =a k *p k, where p k Represents the liquid level position value corresponding to the kth fuzzy rule.

[0129] Furthermore, the information entropy H of each category of liquid type is calculated i , and the information entropy H i The smallest category is used as the predicted liquid type, where the information entropy H i The calculation formula is as follows:

[0130]

[0131] The liquid level position is determined by the weighted median method, specifically: p The element v in pk Sort the values ​​from small to large and record the corresponding weights ω k . Calculate the weighted cumulative weight and find the position where the cumulative weight exceeds 0.5. The corresponding value is the predicted liquid level position.

[0132] The embodiment of the present invention generates fusion features based on first detection data of changes in optical properties of the surface of liquid in the liquid container provided by the optical sensor, second detection data of capacitance changes caused by the difference in dielectric properties of the liquid in the liquid container and the surrounding air, and third detection data of changes in pressure on the tip of the pipette at different depths in the liquid in the liquid container. Combined with the multimodal fusion model, the liquid surface position can be determined more accurately, reducing the pipetting volume error caused by differences in liquid properties.

[0133] Optionally, the total loss function of the modal fusion model in an embodiment of the present invention includes a feature reconstruction loss function, a fuzzy prediction error loss function, a fuzzy rule consistency loss function and a multimodal feature interaction loss function; the feature reconstruction loss function includes an optical feature reconstruction loss function, a capacitance feature reconstruction loss function and a pressure feature reconstruction loss function; the fuzzy prediction error loss function includes a liquid type prediction loss function and a liquid level position prediction loss function.

[0134] Among them, the optical feature reconstruction loss function L recon-opt It is expressed as:

[0135]

[0136] Where W(·) represents the Wasserstein distance between two distributions, n represents the total number of training samples, and O i and represent the original optical features and reconstructed optical features of the i-th sample respectively;

[0137] Capacitance feature reconstruction loss function L recon-cap It is expressed as:

[0138]

[0139] Among them, C i and Represent the original capacitance feature and reconstructed capacitance feature of the i-th sample respectively;

[0140] Pressure feature reconstruction loss function L recon-pres It is expressed as:

[0141]

[0142] Among them, P i and Represent the original pressure feature and reconstructed pressure feature of the i-th sample respectively;

[0143] Liquid type prediction loss function L pres-type It is expressed as:

[0144]

[0145] Where J represents the number of categories or attributes related to liquid types in the fuzzy rules, K represents the number of fuzzy subsets, represents the membership of the i-th sample to the j-th category attribute and the k-th fuzzy subset, represents the predicted value of the liquid type of the i-th sample under the j-th category attribute and the k-th fuzzy subset relationship, It represents the predicted value of the liquid type of the i-th sample under the j-th category attribute and the l-th fuzzy subset relationship;

[0146] Liquid level prediction loss function L pres-pos It is expressed as:

[0147]

[0148] in, represents the membership of the i-th sample to the k-th fuzzy subset, and represents the values ​​of the predicted liquid level position and the true liquid level position in the j-th dimension respectively;

[0149] Fuzzy rule consistency loss function L rule-cons It is expressed as:

[0150]

[0151] Among them, ω k represents the weight of the kth fuzzy rule, and represents the activation strength of the i-th sample and the j-th sample for the k-th fuzzy rule respectively;

[0152] Multimodal feature interaction loss function L interact It is expressed as:

[0153]

[0154] Where MI(·) represents the calculation of the mutual information between two random variables based on kernel density estimation; F i Represents the fused feature vector of the i-th sample;

[0155] The total loss function represents L all for:

[0156] L all =L recon-opt +L recon-cap +L recon-pres +L pres-type +L pres-pos +L rule-cons +L interact .

[0157] Further, step 60 performs a pipetting operation of the pipette based on the liquid type and the liquid level position, including:

[0158] Step 601, determining the suction force and suction speed of the pipette based on the liquid type;

[0159] Step 602, adjusting the pipette to a target position based on the liquid level position;

[0160] Step 603, performing a suction operation on the liquid at the target position with a suction force and a suction speed.

[0161] Specifically, the suction force and suction speed of the pipette are determined according to the type of liquid, wherein factors of different liquid densities and liquid viscosities may be considered.

[0162] For different liquid density factors: If the liquid type is a liquid with a higher density (such as mercury, etc.), determine the appropriate suction force that the pipette needs to apply based on its density value. Liquids with high density require greater suction to overcome gravity and surface tension to ensure that the tip can smoothly absorb the liquid. For example, for high-density liquids such as mercury, the suction force is much greater than when absorbing low-density liquids such as water. For liquids with low density and high volatility (such as certain organic solvents), the suction force and the speed of liquid absorption should be properly controlled. Excessive suction may cause the liquid to shake violently in the tip, generate bubbles, or even evaporate too quickly, affecting the accuracy of pipetting. At the same time, the speed of liquid absorption should be moderate to prevent liquid splashing.

[0163] Regarding the viscosity of the liquid: When the liquid type is a liquid with high viscosity (such as glycerol), the aspiration speed of the pipette needs to be adjusted. Viscous liquids flow slowly, and aspiration speed that is too fast may cause the tip to not be completely filled with liquid, resulting in residual air. Through experiments or by referring to liquid viscosity data, the aspiration speed can be reduced so that the viscous liquid can be smoothly aspirated into the tip. For liquids with low viscosity (such as ethanol solutions), the aspiration speed can be relatively fast.

[0164] Furthermore, the pipette is adjusted to the target position according to the liquid level position. The specific analysis is as follows:

[0165] Initial approach to the liquid surface: Use the liquid surface position information to control the pipette to move slowly toward the liquid container along the vertical axis. When approaching the liquid surface (e.g., a safe approach distance set from the liquid surface, such as a few millimeters), reduce the movement speed to avoid the tip from inserting too deep into the liquid or hitting the bottom of the container due to inertia or excessive movement. At the same time, continue to monitor the changes in the liquid surface position, because during the approach process, the liquid may produce slight fluctuations due to the approach of the pipette, especially for liquids that are easy to shake (such as low-viscosity liquids with low surface tension). If obvious fluctuations in the liquid surface are detected, temporarily stop the movement of the pipette and wait for the liquid surface to stabilize.

[0166] Follow the liquid surface during aspiration: Perform aspiration operation on the liquid at the target position with suction force and suction speed. During the aspiration process, if the liquid surface changes position due to evaporation, liquid flow caused by pipette aspiration, etc. (known by real-time monitoring of the liquid surface position), adjust the vertical axis position of the pipette. For example, if the liquid level drops, move the pipette downward accordingly, maintain the relative position of the pipette tip and the liquid surface, and ensure that the pipette tip is always at the appropriate aspiration depth. If the liquid level rises temporarily due to reasons such as liquid shaking, suspend the downward movement of the pipette to prevent the pipette tip from being inserted too deep, and continue normal aspiration after the liquid level drops.

[0167] Aspiration completion stage: When the set aspiration volume (determined according to the pipetting operation requirements) is reached, the aspiration operation is stopped. At this time, the pipette tip should be slowly withdrawn from the liquid according to the characteristics of the liquid type. For viscous liquids, it should be withdrawn more slowly to prevent the liquid from drawing or remaining on the container wall; for non-viscous liquids, the withdrawal speed can be slightly faster, but it is also necessary to prevent the liquid from dripping outside the container.

[0168] Furthermore, the pipette liquid level following system provided by the present invention is described below, and the pipette liquid level following system described below and the pipette liquid level following method described above can be referred to each other.

[0169] Optional, see Figure 2 , Figure 2: is a structural diagram of the liquid level following system of the pipette provided by the present invention. The liquid level following system of the pipette is applied to the pipette. The pipette tip includes an optical sensor, a capacitive sensor and a pressure sensor, including:

[0170] The control module 210 is used to respond to the liquid level detection instruction of the pipette and control the pipette to move along the vertical axis toward the liquid container;

[0171] The optical data acquisition module 220 is used to control the optical sensor to emit light to the surface of the liquid in the liquid container, and generate first detection data based on the received light signal; the first detection data represents the change of the optical characteristics of the liquid surface;

[0172] The capacitance data acquisition module 230 is used to measure the capacitance change caused by the difference in dielectric properties of the liquid in the liquid container and the surrounding air based on the capacitance sensor to generate second detection data;

[0173] The pressure data acquisition module 240 is used to measure the change of the pressure on the pipette tip when it is in the liquid of the liquid container at different depths based on the pressure sensor, and generate third detection data;

[0174] The prediction module 250 is used to input the first detection data, the second detection data and the third detection data into the multimodal fusion model, so as to obtain the liquid type and liquid level position of the liquid in the liquid container output by the multimodal fusion model; the multimodal fusion model is trained based on the sample fusion features and their corresponding liquid type labels and liquid level position labels;

[0175] The liquid level following module 260 is used to perform the pipetting operation of the pipette based on the liquid type and the liquid level position, so that the pipette tip follows the liquid level to complete the pipetting.

[0176] The embodiment of the present invention overcomes the problem that a single sensor is greatly affected by the specific properties of the liquid through the use of multiple sensors, and combines the first detection data of the change in optical properties of the surface of the liquid in the liquid container provided by the optical sensor, the second detection data of the capacitance change caused by the difference in dielectric properties of the liquid in the liquid container and the surrounding air, and the third detection data of the change in pressure on the pipette tip at different depths in the liquid in the liquid container, with a multimodal fusion model, to more accurately determine the liquid level position and reduce the pipetting volume error caused by differences in liquid properties.

[0177] See also Figure 3 , Figure 3 FIG. 1 is an embodiment diagram of an electronic device provided by an embodiment of the present invention. Figure 3As shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the following steps are implemented:

[0178] In response to the liquid level detection instruction of the pipette, the pipette is controlled to move the liquid container along the vertical axis;

[0179] Controlling the optical sensor to emit light to the surface of the liquid in the liquid container, and generating first detection data based on the received light signal; the first detection data represents the change of the optical property of the liquid surface;

[0180] The capacitance sensor measures the capacitance change caused by the difference between the dielectric properties of the liquid in the liquid container and the surrounding air, thereby generating second detection data;

[0181] The pressure sensor measures the change in pressure exerted on the tip of the pipette when the tip is in the liquid of the liquid container at different depths, thereby generating third detection data;

[0182] The first detection data, the second detection data and the third detection data are input into a multimodal fusion model, and the multimodal fusion model outputs the liquid type and the liquid level position of the liquid in the liquid container; the multimodal fusion model is trained based on the sample fusion features and their corresponding liquid type labels and liquid level position labels;

[0183] The pipette aspiration operation is performed based on the liquid type and the liquid level position so that the pipette tip follows the liquid level to complete the pipetting.

[0184] See also Figure 4 , Figure 4 Detailed description of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400, on which a computer program 311 is stored. When the computer program 311 is executed by a processor, the following steps are implemented:

[0185] In response to the liquid level detection instruction of the pipette, the pipette is controlled to move the liquid container along the vertical axis;

[0186] Controlling the optical sensor to emit light to the surface of the liquid in the liquid container, and generating first detection data based on the received light signal; the first detection data represents the change of the optical property of the liquid surface;

[0187] The capacitance sensor measures the capacitance change caused by the difference between the dielectric properties of the liquid in the liquid container and the surrounding air, thereby generating second detection data;

[0188] The pressure sensor measures the change in pressure exerted on the tip of the pipette when the tip is in the liquid of the liquid container at different depths, thereby generating third detection data;

[0189] The first detection data, the second detection data and the third detection data are input into a multimodal fusion model, and the multimodal fusion model outputs the liquid type and the liquid level position of the liquid in the liquid container; the multimodal fusion model is trained based on the sample fusion features and their corresponding liquid type labels and liquid level position labels;

[0190] The pipette aspiration operation is performed based on the liquid type and the liquid level position so that the pipette tip follows the liquid level to complete the pipetting.

[0191] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the pipette liquid level following method provided by the above methods, the method comprising:

[0192] In response to the liquid level detection instruction of the pipette, the pipette is controlled to move the liquid container along the vertical axis;

[0193] Controlling the optical sensor to emit light to the surface of the liquid in the liquid container, and generating first detection data based on the received light signal; the first detection data represents the change of the optical property of the liquid surface;

[0194] The capacitance sensor measures the capacitance change caused by the difference between the dielectric properties of the liquid in the liquid container and the surrounding air, thereby generating second detection data;

[0195] The pressure sensor measures the change in pressure exerted on the tip of the pipette when the tip is in the liquid of the liquid container at different depths, thereby generating third detection data;

[0196] The first detection data, the second detection data and the third detection data are input into a multimodal fusion model, and the multimodal fusion model outputs the liquid type and the liquid level position of the liquid in the liquid container; the multimodal fusion model is trained based on the sample fusion features and their corresponding liquid type labels and liquid level position labels;

[0197] The pipette aspiration operation is performed based on the liquid type and the liquid level position so that the pipette tip follows the liquid level to complete the pipetting.

[0198] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0199] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0200] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A pipette liquid level following method, characterized in that: Applied to a pipette, the pipette tip of which includes an optical sensor, a capacitive sensor and a pressure sensor, the method comprising: In response to a liquid level detection instruction of the pipette, controlling the pipette to move along the vertical axis toward the liquid container; Controlling the optical sensor to emit light to the surface of the liquid in the liquid container, and generating first detection data based on the received light signal; the first detection data represents the change of the optical property of the liquid surface; Generate second detection data based on the capacitance change caused by the capacitance sensor measuring the difference between the dielectric properties of the liquid in the liquid container and the surrounding air; Generate third detection data based on the change of pressure applied to the pipette tip when the pipette tip is at different depths in the liquid of the liquid container measured by the pressure sensor; Inputting the first detection data, the second detection data and the third detection data into a multimodal fusion model, and obtaining the multimodal fusion model to output the liquid type and liquid level position of the liquid in the liquid container; the multimodal fusion model is trained based on sample fusion features and their corresponding liquid type labels and liquid level position labels; The pipette aspiration operation is performed based on the liquid type and the liquid surface position, so that the pipette tip follows the liquid surface to complete the pipetting.

2. The pipette liquid level following method according to claim 1, characterized in that: The multimodal fusion model includes a first feature extraction layer, a second feature extraction layer, a feature fusion layer and a fuzzy prediction layer; the first feature extraction layer includes a plurality of convolution layers with different convolution kernel sizes, a pooling layer and a fully connected layer; the second feature extraction layer includes a plurality of hidden units; The step of inputting the first detection data, the second detection data, and the third detection data into a multimodal fusion model to obtain a liquid type and a liquid level position of the liquid in the liquid container output by the multimodal fusion model includes: Inputting the first detection data, the second detection data, and the third detection data into the multimodal fusion model, performing feature extraction on the first detection data based on the first feature extraction layer, and obtaining optical features output by the first feature extraction layer; Based on the second feature extraction layer, feature extraction is performed on the second detection data and the third detection data respectively, so as to obtain capacitance features and pressure features output by the second feature extraction layer; Based on the feature fusion layer, the optical feature, the capacitance feature and the pressure feature are feature spliced ​​to obtain a fusion feature output by the feature fusion layer; The fusion feature is predicted based on the fuzzy prediction layer in combination with a fuzzy prediction algorithm to obtain the liquid type and liquid level position output by the fuzzy prediction layer.

3. The liquid level following method of a pipette according to claim 2, characterized in that: The step of predicting the fusion feature based on the fuzzy prediction layer in combination with a fuzzy prediction algorithm to obtain the liquid type and liquid level position output by the fuzzy prediction layer includes: The fuzzy prediction layer performs the following steps: The fused feature is converted into a corresponding fused feature vector F; wherein the fused feature vector F is expressed as F=(f1, f2, f3), f1 represents an optical feature, f2 represents a capacitance feature, and f3 represents a pressure feature; The fused feature vector F is mapped to a high-dimensional feature space through a high-dimensional mapping function to obtain the mapping of the fused feature vector F in the high-dimensional feature space Among them, the mapping of the fusion feature vector F in the high-dimensional feature space It is expressed as: in, represents a high-dimensional mapping function, a 1i ,a 2i ,b i ,c i Represents the parameters learned through training data; Perform hierarchical clustering on the feature space after high-dimensional mapping to obtain K cluster centers C k ,k=1,2,...,K,C k =(C1,C2,...,C K ), calculate the mapping of each fused feature vector in high-dimensional space The distance d to each cluster center k (F), distance d k (F) is expressed as: According to the mapping of each fused feature vector in high-dimensional space The distance d to each cluster center k (F), determine the membership function μ k (F), membership function μ k (F) is expressed as: Among them, the membership function μ k (F) indicates the degree to which the fused feature vector belongs to each cluster; The manifold learning algorithm is used to analyze the sample fusion features in the training data and their corresponding liquid types and liquid surface positions, and fuzzy rules are generated according to the local neighborhood relationship on the manifold; the weight of the fuzzy rule ω k The weight ω is determined based on the sample density of the local area on the manifold. k It is expressed as: n k represents the number of samples in the local area corresponding to the kth fuzzy subset on the manifold, and N represents the total number of samples; Based on the membership function μ k (F) and weight ω k A prediction is performed to obtain the liquid type and liquid level position output by the fuzzy prediction layer.

4. The liquid level following method of a pipette according to claim 3, characterized in that: The K cluster centers C k , the steps to determine are as follows: Each fused feature vector after high-dimensional mapping is taken as a separate cluster, where the number of initial clusters is equal to the number of samples N; Cluster merging criterion: Calculate Q for each pair of clusters a and clustering Q b The distance d(Q a ,Q b ), distance d(Q a ,Q b )The specific calculation formula is as follows: Among them, F i and F j Represents the two fused feature vectors after high-dimensional mapping, d ij Represents two fused feature vectors F i and F j The distance between them, K(·) represents the Gaussian kernel function, K(F i ,F j ) represents the fusion feature vector F i and the fusion feature vector F j The kernel function value between represents the high-dimensional mapping function, σ represents the kernel bandwidth parameter, |Q a |and|Q b | respectively represent cluster Q a and clustering Q b The number of samples in ; In each iteration, the two closest clusters are found according to the cluster merging criterion and the two clusters are merged into a new cluster Q new , the center of the new cluster C new The calculation formula is as follows: Among them, C a Represents cluster Q a The center of C b Represents cluster C b the center of; Repeat the above merging process to continuously reduce the number of clusters until the predetermined number of clusters K is reached. The center C of the K clusters finally obtained is k =(C1,C2,...,C K ).

5. The liquid level following method of a pipette according to claim 1, characterized in that: The total loss function of the modal fusion model includes a feature reconstruction loss function, a fuzzy prediction error loss function, a fuzzy rule consistency loss function and a multimodal feature interaction loss function; the feature reconstruction loss function includes an optical feature reconstruction loss function, a capacitance feature reconstruction loss function and a pressure feature reconstruction loss function; the fuzzy prediction error loss function includes a liquid type prediction loss function and a liquid level position prediction loss function.

6. The liquid level following method of a pipette according to claim 1, characterized in that: The optical feature reconstruction loss function L recon-opt It is expressed as: Where W(·) represents the Wasserstein distance between two distributions, n represents the total number of training samples, and O i and represent the original optical features and reconstructed optical features of the i-th sample respectively; The capacitance feature reconstruction loss function L recon-cap It is expressed as: Among them, C i and Represent the original capacitance feature and reconstructed capacitance feature of the i-th sample respectively; The pressure feature reconstruction loss function L recon-pres It is expressed as: Among them, P i and Represent the original pressure feature and reconstructed pressure feature of the i-th sample respectively; The liquid type prediction loss function L pres-type It is expressed as: Where J represents the number of categories or attributes related to liquid types in the fuzzy rules, K represents the number of fuzzy subsets, represents the membership of the i-th sample to the j-th category attribute and the k-th fuzzy subset, represents the predicted value of the liquid type of the i-th sample under the j-th category attribute and the k-th fuzzy subset relationship, It represents the predicted value of the liquid type of the i-th sample under the j-th category attribute and the l-th fuzzy subset relationship; The liquid level position prediction loss function L pres-pos It is expressed as: in, represents the membership of the i-th sample to the k-th fuzzy subset, and represents the values ​​of the predicted liquid level position and the true liquid level position in the j-th dimension respectively; The fuzzy rule consistency loss function L rule-cons It is expressed as: Among them, ω k represents the weight of the kth fuzzy rule, and represents the activation strength of the i-th sample and the j-th sample for the k-th fuzzy rule respectively; The multimodal feature interaction loss function L interact It is expressed as: Where MI(·) represents the calculation of the mutual information between two random variables based on kernel density estimation; F i Represents the fused feature vector of the i-th sample; The total loss function represents L all for: L all =L recon-opt +L recon-cap +L recon-pres +L pres-type +L pres-pos +L rule-cons +L interact .

7. The liquid level following method of a pipette according to any one of claims 1 to 6, characterized in that: The step of performing the liquid aspiration operation of the pipette based on the liquid type and the liquid level position comprises: Determining the suction force and suction speed of the pipette based on the type of the liquid; adjusting the pipette to a target position based on the liquid level position; The liquid suction operation is performed on the liquid in the liquid container at the target position with the suction force magnitude and the suction speed.

8. A pipette liquid level following system, characterized in that: Applied to a pipette, the pipette tip of which includes an optical sensor, a capacitive sensor and a pressure sensor, and the device includes: A control module, for responding to a liquid level detection instruction of the pipette, and controlling the pipette to move along a vertical axis toward the liquid container; An optical data acquisition module, used to control the optical sensor to emit light to the surface of the liquid in the liquid container, and generate first detection data based on the received light signal; the first detection data represents the change of the optical property of the liquid surface; A capacitance data acquisition module, used for measuring the capacitance change caused by the difference between the dielectric properties of the liquid in the liquid container and the surrounding air based on the capacitance sensor, to generate second detection data; A pressure data acquisition module, used for measuring the change of the pressure exerted on the pipette tip at different depths in the liquid of the liquid container based on the pressure sensor, and generating third detection data; A prediction module, used for inputting the first detection data, the second detection data and the third detection data into a multimodal fusion model, so as to obtain the liquid type and liquid level position of the liquid in the liquid container output by the multimodal fusion model; the multimodal fusion model is trained based on sample fusion features and their corresponding liquid type labels and liquid level position labels; The liquid level following module is used to perform the pipetting operation of the pipette based on the liquid type and the liquid level position, so that the pipette tip of the pipette follows the liquid level to complete the pipetting.

9. An electronic device, comprising: Memory for storing computer software programs; A processor, used to read and execute the computer software program, characterized in that when the processor executes the computer software program, it implements the pipette liquid surface following method as described in any one of claims 1 to 8.

10. A non-transitory computer-readable storage medium, wherein a computer software program is stored in the non-transitory computer-readable storage medium, and when the computer software program is executed by a processor, the pipette liquid level following method as claimed in any one of claims 1 to 8 is implemented.

Images