CN116127385A

CN116127385A - Banana powder and preparation method thereof

Info

Publication number: CN116127385A
Application number: CN202310202637.8A
Authority: CN
Inventors: 窦敏; 黄全
Original assignee: Gaozhou Shishengyuan Biotechnology Development Co ltd
Current assignee: Gaozhou Shishengyuan Biotechnology Development Co ltd
Priority date: 2023-03-06
Filing date: 2023-03-06
Publication date: 2023-05-16

Abstract

A banana powder and its preparation method are disclosed. Firstly, arranging stirring speed values at a plurality of preset time points, then, obtaining stirring speed time sequence feature vectors through a one-dimensional convolutional neural network model, then, respectively, obtaining a plurality of mixed state feature graphs through a first convolutional neural network model by using mixed state detection images at a plurality of preset time points, then, carrying out feature graph expansion on the plurality of mixed state feature graphs, obtaining mixed state time sequence semantic understanding feature vectors through a context encoder, then, calculating the response estimation of the mixed state time sequence semantic understanding feature vectors relative to the stirring speed time sequence feature vectors to obtain a classification feature matrix, and finally, passing the classification feature matrix through a classifier to obtain a classification result for representing that the stirring speed value at the current time point is required to be increased or reduced. Thus, the preparation quality and efficiency of the banana powder can be improved.

Description

Banana powder and preparation method thereof

Technical Field

The application relates to the field of intelligent control, and more particularly, to banana powder and a preparation method thereof.

Background

At present, banana products in China mainly comprise banana jam, banana wine, banana ice cream, banana can, banana yogurt, banana powder and the like, wherein the banana powder is very convenient to eat, and can be directly added into sterilized water to be brewed into beverages or quantitatively added into other foods as a health-care food raw material. The product is produced in China relatively little, and mainly because of high content of banana pectin, high viscosity and difficult dehydration, the currently adopted production methods are spray drying and freeze drying. Spray drying is a process of peeling, color protecting, pulping, homogenizing and spray drying ripe bananas, and is easy to adhere to walls and high in energy consumption in the drying process. The freeze drying is to peel, protect color, slice, stir and crush the ripe bananas, freeze-dry and screen the ripe bananas to obtain the banana powder, and the method can prepare the banana powder with better quality.

At present, in the process of preparing banana powder by actually adopting a freeze drying mode, the added banana powder and other raw materials are required to be stirred and mixed to prepare flour, but when stirring control is carried out in the existing scheme, the stirring speed is controlled within a certain range, the suitability between the stirring speed and the stirring state of the mixed material is not considered, so that the mixed material is difficult to form flour, and the preparation quality and efficiency of the banana powder are reduced.

Thus, an intelligent banana powder preparation scheme is desired.

Disclosure of Invention

The present application has been made in order to solve the above technical problems. The embodiment of the application provides banana powder and a preparation method thereof. Firstly, arranging stirring speed values at a plurality of preset time points, then, obtaining stirring speed time sequence feature vectors through a one-dimensional convolutional neural network model, then, respectively, obtaining a plurality of mixed state feature graphs through a first convolutional neural network model by using mixed state detection images at a plurality of preset time points, then, carrying out feature graph expansion on the plurality of mixed state feature graphs, obtaining mixed state time sequence semantic understanding feature vectors through a context encoder, then, calculating the response estimation of the mixed state time sequence semantic understanding feature vectors relative to the stirring speed time sequence feature vectors to obtain a classification feature matrix, and finally, passing the classification feature matrix through a classifier to obtain a classification result for representing that the stirring speed value at the current time point is required to be increased or reduced. Thus, the preparation quality and efficiency of the banana powder can be improved.

According to one aspect of the present application, there is provided a banana powder and a method for preparing the same, comprising:

acquiring mixing state detection images of a mixture of banana powder, wet starch and light cream at a plurality of preset time points in a preset time period;

arranging the stirring speed values at a plurality of preset time points into stirring speed input vectors according to a time dimension, and then obtaining stirring speed time sequence feature vectors through a one-dimensional convolutional neural network model;

respectively passing the mixed state detection images at a plurality of preset time points through a first convolution neural network model comprising a depth feature fusion module to obtain a plurality of mixed state feature images;

performing feature map expansion on the plurality of mixed state feature maps to obtain a plurality of mixed state expansion feature vectors;

passing the plurality of mixed state expanded feature vectors through a context encoder based on a converter to obtain mixed state sequential semantic understanding feature vectors;

calculating the response estimation of the mixed state time sequence semantic understanding feature vector relative to the stirring speed time sequence feature vector to obtain a classification feature matrix; and

And passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating that the stirring speed value at the current time point should be increased or decreased.

In the above banana powder and the preparation method thereof, the step of arranging the stirring speed values at the plurality of predetermined time points into the stirring speed input vector according to the time dimension and then obtaining the stirring speed time sequence feature vector through a one-dimensional convolutional neural network model comprises the following steps:

each layer of the one-dimensional convolutional neural network model is used for respectively carrying out input data in forward transfer of the layer:

performing convolution processing on the input data to obtain a convolution feature vector;

pooling the convolution feature vectors to obtain pooled feature vectors; and

non-linear activation is carried out on the pooled feature vectors to obtain activated feature vectors;

the output of the last layer of the one-dimensional convolutional neural network model is the stirring speed time sequence feature vector, and the input of the first layer of the one-dimensional convolutional neural network model is the stirring speed input vector.

In the above banana powder and the preparation method thereof, the step of passing the mixed state detection images at the predetermined time points through the first convolutional neural network model including the depth feature fusion module to obtain a plurality of mixed state feature diagrams includes:

Respectively inputting the mixed state detection images of the plurality of preset time points into the first convolutional neural network model to extract a plurality of shallow feature maps from the shallow layer of the first convolutional neural network model and a plurality of deep feature maps from the deep layer of the first convolutional neural network model;

and cascading the shallow feature maps and the deep feature maps by using the depth feature fusion module to obtain the mixed state feature maps.

In the banana powder and the preparation method thereof, the step of passing the plurality of mixed state unfolding feature vectors through a context encoder based on a converter to obtain mixed state time sequence semantic understanding feature vectors comprises the following steps:

performing global-based context semantic coding on the plurality of mixed state expansion feature vectors by using the context encoder based on the converter to obtain a plurality of context mixed state feature vectors; and

and cascading the plurality of context mixed state feature vectors to obtain the mixed state sequential semantic understanding feature vector.

In the banana powder and the preparation method thereof, the performing global context semantic coding on the plurality of mixed state expansion feature vectors by using the context encoder based on the converter to obtain a plurality of context mixed state feature vectors comprises the following steps:

One-dimensional arrangement is carried out on the plurality of mixed state expansion feature vectors so as to obtain global mixed state feature vectors;

calculating the product between the global mixed state feature vector and the transpose vector of each mixed state expansion feature vector in the mixed state expansion feature vectors to obtain a plurality of self-attention association matrixes;

respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices;

obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and

and weighting each mixed state unfolding feature vector in the mixed state unfolding feature vectors by taking each probability value in the probability values as a weight so as to obtain the context mixed state feature vectors.

In the banana powder and the preparation method thereof, calculating the response estimation of the mixed state time sequence semantic understanding feature vector relative to the stirring speed time sequence feature vector to obtain a classification feature matrix comprises the following steps:

Based on the stirring speed time sequence feature vector and the mixed state time sequence semantic understanding feature vector, carrying out feature distribution optimization on the mixed state time sequence semantic understanding feature vector to obtain an optimized mixed state time sequence semantic understanding feature vector;

constructing a Gaussian density map of the optimized mixed state time sequence semantic understanding feature vector and the stirring speed time sequence feature vector to obtain a first Gaussian density map and a second Gaussian density map;

calculating a responsiveness estimate of the first gaussian density map relative to the second gaussian density map to obtain a responsive gaussian density map; and

and carrying out Gaussian discretization on the Gaussian distribution of each position in the response Gaussian density map to obtain the classification characteristic matrix.

In the above banana powder and the preparation method thereof, performing feature distribution optimization on the mixed state sequential semantic understanding feature vector based on the stirring speed sequential feature vector and the mixed state sequential semantic understanding feature vector to obtain an optimized mixed state sequential semantic understanding feature vector, including:

calculating incoherent sparse response type fusion of the stirring speed time sequence feature vector and the mixed state time sequence semantic understanding feature vector according to the following formula to obtain the optimized mixed state time sequence semantic understanding feature vector;

Wherein, the formula is:

wherein V is ₁ 、V ₂ And V ₂ ' respectively representing the stirring speed time sequence feature vector, the mixing state time sequence semantic understanding feature vector and the optimized mixing state time sequence semantic understanding feature vector, |·|| ₁ And|| | ₂ Respectively representing a first norm and a second norm of the vector, L is the length of the vector,

and +.A vector product and a vector point product are shown, respectively, and all vectors are in the form of row vectors.

In the banana powder and the preparation method thereof, the classification feature matrix is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the stirring speed value at the current time point should be increased or decreased, and the method comprises the following steps:

expanding the classification feature matrix into classification feature vectors according to row vectors or column vectors;

performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and

and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.

According to another aspect of the present application, there is provided a banana powder produced by the method of producing banana powder as described in any one of the preceding.

Compared with the prior art, the banana powder and the preparation method thereof have the advantages that firstly, stirring speed values at a plurality of preset time points are arranged and then are subjected to one-dimensional convolutional neural network model to obtain stirring speed time sequence feature vectors, then, mixed state detection images at the preset time points are respectively subjected to first convolutional neural network model to obtain a plurality of mixed state feature images, then, feature image expansion is carried out on the mixed state feature images and the mixed state feature images are subjected to a context encoder to obtain mixed state time sequence semantic understanding feature vectors, then, the response estimation of the mixed state time sequence semantic understanding feature vectors relative to the stirring speed time sequence feature vectors is calculated to obtain a classification feature matrix, and finally, the classification feature matrix is subjected to a classifier to obtain a classification result for indicating that the stirring speed value at the current time point is required to be increased or reduced. Thus, the preparation quality and efficiency of the banana powder can be improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. The following drawings are not intended to be drawn to scale, with emphasis instead being placed upon illustrating the principles of the present application.

Fig. 1 is a schematic view of a banana powder and its preparation method according to an embodiment of the present application.

Fig. 2 is a flow chart of banana powder and its preparation method according to an embodiment of the present application.

Fig. 3 is a schematic diagram of the structure of banana powder and its preparation method according to an embodiment of the present application.

Fig. 4 is a flowchart of substep S130 in banana powder and its preparation method according to an embodiment of the present application.

Fig. 5 is a flowchart of substep S150 in banana powder and its preparation method according to an embodiment of the present application.

Fig. 6 is a flowchart of sub-step S151 in the banana powder and its preparation method according to an embodiment of the present application.

Fig. 7 is a flowchart of substep S160 in banana powder and its preparation method according to an embodiment of the present application.

Fig. 8 is a flowchart of substep S170 in banana powder and its preparation method according to an embodiment of the present application.

Fig. 9 is a block diagram of banana powder and its preparation system according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some, but not all embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present application without making any inventive effort, are also within the scope of the present application.

As used in this application and in the claims, the terms "a," "an," "the," and/or "the" are not specific to the singular, but may include the plural, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.

Although the present application makes various references to certain modules in a system according to embodiments of the present application, any number of different modules may be used and run on a user terminal and/or server. The modules are merely illustrative, and different aspects of the systems and methods may use different modules.

Flowcharts are used in this application to describe the operations performed by systems according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in order precisely. Rather, the various steps may be processed in reverse order or simultaneously, as desired. Also, other operations may be added to or removed from these processes.

Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some of the embodiments of the present application and not all of the embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.

As described above, in the process of preparing banana powder by actually adopting the freeze-drying method, stirring and mixing are required for the added banana powder and other raw materials to prepare flour, but in the conventional scheme, when stirring control is performed, only the stirring speed is controlled within a certain range, and the suitability between the stirring speed and the stirring state of the mixed material is not considered, so that the mixed material is difficult to form flour, and the preparation quality and efficiency of the banana powder are reduced. Thus, an intelligent banana powder preparation scheme is desired.

Specifically, in the technical scheme of the application, a preparation method of banana powder is provided, which comprises the following steps: 1. preparing bananas: selecting yellow ripe bananas, and cutting the bananas into slices by hands; 2. stirring: placing banana powder into an electric stirrer, adding a little wet starch and light cream, and stirring to flour shape; 3. pressing: pouring the prepared banana powder into a pressing machine, and pressing the pressing head to reduce the shells so that the banana powder is fine; 4. and (3) drying: drying the pressed banana powder in a room temperature or low temperature dryer; 5. screening, namely pouring the dried banana powder into a screen to screen the banana powder out of the powder; 6. and (3) bottling: and bottling or canning the sieved banana powder, and sealing.

Accordingly, in consideration of the fact that the adaptive control of the stirring speed during the actual preparation of the banana powder should be adapted to the mixing state change condition of the mixed material composed of the banana powder, the wet starch and the light cream, that is, the stirring speed value is adaptively adjusted based on the mixing state change characteristics of the mixed material, the stirring effect is optimized so that the stirred mixed material presents flour. In the process, the difficulty is how to establish the mapping relation between the mixing state change characteristics of the mixed materials and the time sequence change characteristics of the stirring speed values, so as to adaptively adjust the stirring speed values based on the state time sequence change conditions of the mixed materials, improve the stirring effect, and improve the preparation quality and efficiency of the banana powder.

In recent years, deep learning and neural networks have been widely used in the fields of computer vision, natural language processing, text signal processing, and the like. In addition, deep learning and neural networks have also shown levels approaching and even exceeding humans in the fields of image classification, object detection, semantic segmentation, text translation, and the like.

The development of deep learning and neural networks provides new solutions and schemes for mining complex mapping relations between the mixing state change characteristics of the mixed materials and the time sequence change characteristics of the stirring speed values. Those of ordinary skill in the art will appreciate that deep learning based deep neural network models can adjust parameters of the deep neural network model by appropriate training strategies, such as by a gradient descent back-propagation algorithm, to enable modeling of complex nonlinear correlations between things, which is obviously suitable for modeling and establishing complex mappings between mixing state change characteristics of the mixture and time-series change characteristics of the stirring velocity values.

Specifically, in the technical scheme of the application, firstly, stirring speed values at a plurality of preset time points in a preset time period and mixed state detection images of mixed materials consisting of banana powder, wet starch and light cream at the preset time points are obtained. Next, the characteristic information that has a correlation between the stirring speed values at the respective predetermined time points is considered as a result of the change law of the stirring speed values having dynamics in the time dimension. Therefore, in the technical scheme of the application, after the stirring speed values at the plurality of preset time points are arranged into the stirring speed input vector according to the time dimension, feature mining is performed in the one-dimensional convolutional neural network model, so that time sequence dynamic associated feature distribution information of the stirring speed values in the time dimension is extracted, and therefore the stirring speed time sequence feature vector is obtained.

Then, it is considered that since the mixed state detection images of the plurality of predetermined time points are image data, feature mining of the mixed state detection images of the mixed materials of the respective predetermined time points is performed using a convolutional neural network model having excellent expression in terms of implicit feature extraction of the images. In particular, in order to more accurately detect the mixed state of the mixed materials consisting of the banana powder, the wet starch and the light cream when extracting the hidden features of the mixed state detection image at each predetermined time point, the shallow features such as texture, color and the like of the mixed state detection image should be focused, and have important significance for detecting the mixed state change condition of the mixed materials, while the convolutional neural network becomes blurred and even submerged by noise with the deepening of the depth of the convolutional neural network during encoding. Therefore, in the technical scheme of the application, the mixed state detection images at a plurality of preset time points are processed by using the first convolution neural network model comprising the depth feature fusion module to obtain a plurality of mixed state feature images. It should be understood that, compared to a standard convolutional neural network model, the first convolutional neural network model according to the application can retain the shallow layer features and deep layer features of the mixed state detection image at each predetermined time point, so that not only feature information is richer, but also features with different depths can be retained, so as to improve the detection precision of the mixed state change of the mixed material.

Further, considering that the implicit characteristic of the mixed state of the mixed material due to the respective predetermined time points has an association relationship not only at different positions in the mixed state detection image of the respective predetermined points, it also has a dynamically changing characteristic in the time dimension, that is, the mixed state characteristic information of the mixed state of the mixed material in the mixed state detection image of the respective predetermined time points has a dynamically association relationship in the time dimension. Therefore, in the technical solution of the present application, in order to sufficiently extract dynamic change feature information of the mixing state feature of the mixed material at each predetermined time point in the time dimension, the feature map expansion is further performed on the plurality of mixing state feature maps to obtain a plurality of mixing state expansion feature vectors, and the plurality of mixing state expansion feature vectors are encoded in a context encoder based on a converter, so as to extract dynamic association feature distribution information of the mixing state feature of the mixed material at each predetermined time point in the time dimension based on the time sequence global, thereby obtaining a mixing state time sequence semantic understanding feature vector.

And then, calculating the response estimation of the time sequence semantic understanding feature vector of the mixing state relative to the time sequence feature vector of the stirring speed so as to represent the time sequence dynamic change feature of the stirring speed and the time sequence dynamic association feature distribution information of the mixing state of the mixing material.

In particular, in order to improve the accuracy of real-time control of the stirring speed value, data enhancement is required for the time-series dynamic change characteristic of the stirring speed and the time-series dynamic correlation characteristic of the mixing state of the mixed material in a high-dimensional feature space, in consideration of both the fluctuation and uncertainty of the mixing state information of the mixed material and the stirring speed value in the time dimension. It should be understood that, as a learning target of the neural network model, the gaussian density map may represent a joint distribution in the case where a plurality of feature values constitute an overall distribution due to probability density thereof, that is, a feature distribution is taken as a priori distribution, and probability density under the influence of correlation with other a priori distribution positions at each a priori distribution position is obtained as a posterior distribution, thereby describing the feature distribution more accurately in a higher dimension. Therefore, in the technical scheme of the application, the time sequence dynamic change characteristic of the stirring speed and the time sequence dynamic association characteristic of the mixing state of the mixed material can be subjected to data enhancement through the mixing state information of the mixed material and the prior distribution of the stirring speed value, namely Gaussian distribution.

Specifically, firstly, respectively constructing a mixed state time sequence semantic understanding feature vector and a Gaussian density chart of the stirring speed time sequence feature vector to obtain a mixed state Gaussian density chart and a stirring speed Gaussian density chart; then, calculating the response estimation of the mixing state Gaussian density diagram relative to the stirring speed Gaussian density diagram so as to represent the correlation characteristic distribution information between the time sequence dynamic change characteristic of the stirring speed value and the mixing state time sequence dynamic correlation characteristic of the mixed material, thereby obtaining a response Gaussian density diagram; and then, carrying out Gaussian discretization processing on the responsive Gaussian density map so as not to generate information loss when the data features are amplified, thereby obtaining a classification feature matrix.

Then, the classification feature matrix is passed through a classifier to obtain a classification result indicating whether the stirring speed value at the current time point should be increased or decreased. That is, in the technical solution of the present application, the label of the classifier includes that the stirring speed value at the current time point should be increased (first label) and that the stirring speed value at the current time point should be decreased (second label), wherein the classifier determines to which classification label the classification feature matrix belongs through a soft maximum function. It should be understood that, in the technical solution of the present application, the classification label of the classifier is a control policy label of the stirring speed value, so after the classification result is obtained, the stirring speed value at the current time point can be adaptively adjusted based on the classification result, so as to improve the effect of stirring and mixing into flour, and improve the preparation quality and efficiency of the banana powder.

In particular, in the technical solution of the present application, when calculating the responsiveness estimation of the mixed state time-series semantic understanding feature vector with respect to the stirring speed time-series feature vector based on a gaussian density diagram, the calculation of the position-by-position responsiveness estimation of the feature value granularity of the mixed state time-series semantic understanding feature vector with respect to the stirring speed time-series feature vector is focused due to the characteristics of the gaussian density diagram. Therefore, if the responsiveness of the mixed state timing semantic understanding feature vector to the vector magnitude of the stirring speed timing feature vector can be further determined, the accuracy of the classification result of the classification feature matrix can be improved.

Thus, the applicant of the present application recorded the stirring speed timing feature vector as, for example, V ₁ As a source vector, the mixture isState timing semantic understanding feature vectors, e.g. denoted V ₂ As a response vector, calculating its incoherent sparse response fusion to optimize the mixed state timing semantic understanding feature vector, e.g., the optimized mixed state timing semantic understanding feature vector is denoted as V ₂ ' expressed as:

wherein I II ₁ And|| | ₂ Representing the first and second norms of the vector, L being the length of the vector,

Here, the incoherent sparse response fusion obtains incoherent sparse fusion representation among vectors through fuzzy bit distribution responsiveness of vector differences represented by a norm and true differential embedding responsiveness based on modulo constraint of differential vectors in the case of authenticity distribution (group-truth distribution) with initial response vectors as feature inter-domain responsiveness fusion, so as to extract response relation of probability distribution descriptiveness after feature vector fusion, thereby improving the mixed state timing semantic understanding feature vector V as incoherent sparse response fusion optimization ₂ ' fusion express effect on source vector and response vector with response relationship. In this way, the mixed state timing semantic understanding feature vector V after optimization is calculated based on the Gaussian density map ₂ ' time sequence characteristic vector V relative to the stirring speed ₁ The accuracy of the classification result of the obtained classification feature matrix can be improved. Therefore, the stirring speed value can be adaptively adjusted in real time and accurately based on the state change condition of the mixed materials, so that the stirring effect is improved, and the preparation quality and efficiency of the banana powder are improved.

Fig. 1 is an application scene diagram of banana powder and its preparation method according to an embodiment of the present application. As shown in fig. 1, in this application scenario, first, stirring speed values (e.g., D1 illustrated in fig. 1) at a plurality of predetermined time points within a predetermined period of time and mixing state detection images (e.g., D2 illustrated in fig. 1) of a mixture of banana powder, wet starch and whipped cream at the plurality of predetermined time points are acquired, and then, the stirring speed values at the plurality of predetermined time points and the mixing state detection images of the mixture of banana powder, wet starch and whipped cream at the plurality of predetermined time points within the predetermined period of time are input to a server (e.g., S illustrated in fig. 1) in which banana powder and a preparation algorithm thereof are deployed, wherein the server is capable of processing the stirring speed values at the plurality of predetermined time points and the mixing state detection images of the mixture of banana powder, wet starch and whipped cream at the plurality of predetermined time points using the banana powder and a preparation algorithm thereof to obtain a stirring speed value for indicating that the stirring speed value at the current time point should be increased or should be decreased.

Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.

Fig. 2 is a flow chart of banana powder and its preparation method according to an embodiment of the present application. As shown in fig. 2, the banana powder and the preparation method thereof according to the embodiment of the present application include the steps of: s110, acquiring stirring speed values at a plurality of preset time points in a preset time period and mixed state detection images of mixed materials consisting of banana powder, wet starch and light cream at the preset time points; s120, arranging the stirring speed values of the plurality of preset time points into a stirring speed input vector according to a time dimension, and then obtaining a stirring speed time sequence feature vector through a one-dimensional convolutional neural network model; s130, respectively passing the mixed state detection images at a plurality of preset time points through a first convolution neural network model comprising a depth feature fusion module to obtain a plurality of mixed state feature images; s140, performing feature map expansion on the plurality of mixed state feature maps to obtain a plurality of mixed state expansion feature vectors; s150, the plurality of mixed state unfolding feature vectors pass through a context encoder based on a converter to obtain mixed state time sequence semantic understanding feature vectors; s160, calculating the response estimation of the mixed state time sequence semantic understanding feature vector relative to the stirring speed time sequence feature vector to obtain a classification feature matrix; and S170, passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating that the stirring speed value at the current time point is increased or decreased.

Fig. 3 is a schematic diagram of the structure of banana powder and its preparation method according to an embodiment of the present application. As shown in fig. 3, in the network architecture, first, stirring speed values at a plurality of predetermined time points within a predetermined period of time and mixed state detection images of a mixed material consisting of banana powder, wet starch and whipped cream at the plurality of predetermined time points are acquired; then, arranging the stirring speed values of the plurality of preset time points into stirring speed input vectors according to a time dimension, and obtaining stirring speed time sequence feature vectors through a one-dimensional convolutional neural network model; then, the mixed state detection images at a plurality of preset time points are respectively passed through a first convolution neural network model comprising a depth feature fusion module to obtain a plurality of mixed state feature images; then, performing feature map expansion on the plurality of mixed state feature maps to obtain a plurality of mixed state expansion feature vectors; then, the plurality of mixed state unfolding feature vectors pass through a context encoder based on a converter to obtain mixed state time sequence semantic understanding feature vectors; then, calculating the response estimation of the mixed state time sequence semantic understanding feature vector relative to the stirring speed time sequence feature vector to obtain a classification feature matrix; finally, the classification feature matrix is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating that the stirring speed value at the current time point should be increased or decreased.

More specifically, in step S110, stirring speed values at a plurality of predetermined time points within a predetermined period of time and mixed state detection images of a mixed material composed of banana powder, wet starch and whipped cream at the plurality of predetermined time points are acquired. In the actual preparation process of banana powder, the self-adaptive control of the stirring speed during stirring is adapted to the mixing state change condition of the mixed material consisting of banana powder, wet starch and light cream, that is, the stirring speed value is self-adaptively adjusted based on the mixing state change characteristic of the mixed material, so that the stirring effect is optimized, and the stirred mixed material presents flour. In the method, the mixing speed value is adaptively adjusted based on the state time sequence change condition of the mixed material by establishing the mapping relation between the mixing state change characteristic of the mixed material and the time sequence change characteristic of the mixing speed value, so that the mixing effect is improved, and the preparation quality and efficiency of the banana powder are improved.

More specifically, in step S120, the stirring speed values at the plurality of predetermined time points are arranged into a stirring speed input vector according to a time dimension, and then the stirring speed input vector is passed through a one-dimensional convolutional neural network model to obtain a stirring speed time sequence feature vector. Since the stirring speed values have a dynamic change rule in the time dimension, that is, characteristic information having a correlation between the stirring speed values at the respective predetermined time points. Therefore, in the technical scheme of the application, after the stirring speed values at the plurality of preset time points are arranged into the stirring speed input vector according to the time dimension, feature mining is performed in the one-dimensional convolutional neural network model, so that time sequence dynamic associated feature distribution information of the stirring speed values in the time dimension is extracted, and therefore the stirring speed time sequence feature vector is obtained.

It should be understood that a one-dimensional convolutional neural network model refers to a convolutional neural network having one-dimensional convolutional kernels. The convolutional neural network (Convolutional Neural Network, CNN) is an artificial neural network and has wide application in the fields of image recognition and the like. The convolutional neural network may include an input layer, a hidden layer, and an output layer, where the hidden layer may include a convolutional layer, a pooling layer, an activation layer, a full connection layer, etc., where the previous layer performs a corresponding operation according to input data, outputs an operation result to the next layer, and obtains a final result after the input initial data is subjected to a multi-layer operation. The convolutional neural network model has excellent performance in the aspect of image local feature extraction by taking a convolutional kernel as a feature filtering factor, and has stronger feature extraction generalization capability and fitting capability compared with the traditional image feature extraction algorithm based on statistics or feature engineering.

Accordingly, in a specific example, the method for obtaining the time sequence feature vector of the stirring speed by using the one-dimensional convolutional neural network model after arranging the stirring speed values of the plurality of preset time points into the stirring speed input vector according to the time dimension includes: each layer of the one-dimensional convolutional neural network model is used for respectively carrying out input data in forward transfer of the layer: performing convolution processing on the input data to obtain a convolution feature vector; pooling the convolution feature vectors to obtain pooled feature vectors; and performing nonlinear activation on the pooled feature vectors to obtain activated feature vectors; the output of the last layer of the one-dimensional convolutional neural network model is the stirring speed time sequence feature vector, and the input of the first layer of the one-dimensional convolutional neural network model is the stirring speed input vector.

More specifically, in step S130, the mixed state detection images at the plurality of predetermined time points are respectively passed through a first convolutional neural network model including a depth feature fusion module to obtain a plurality of mixed state feature maps. Since the mixed state detection images at the plurality of predetermined time points are image data, feature mining of the mixed state detection images of the mixed materials at the respective predetermined time points is performed using a convolutional neural network model having excellent performance in implicit feature extraction of the images. In particular, in order to more accurately detect the mixed state of the mixed materials consisting of the banana powder, the wet starch and the light cream when extracting the hidden features of the mixed state detection image at each predetermined time point, the shallow features such as texture, color and the like of the mixed state detection image should be focused, and have important significance for detecting the mixed state change condition of the mixed materials, while the convolutional neural network becomes blurred and even submerged by noise with the deepening of the depth of the convolutional neural network during encoding. Therefore, in the technical scheme of the application, the mixed state detection images at a plurality of preset time points are processed by using the first convolution neural network model comprising the depth feature fusion module to obtain a plurality of mixed state feature images. It should be understood that, compared to a standard convolutional neural network model, the first convolutional neural network model according to the application can retain the shallow layer features and deep layer features of the mixed state detection image at each predetermined time point, so that not only feature information is richer, but also features with different depths can be retained, so as to improve the detection precision of the mixed state change of the mixed material.

Accordingly, in a specific example, as shown in fig. 4, the step of passing the mixed state detection images at the plurality of predetermined time points through the first convolutional neural network model including the depth feature fusion module to obtain a plurality of mixed state feature diagrams includes: s131, respectively inputting the mixed state detection images of the predetermined time points into the first convolutional neural network model to extract a plurality of shallow feature maps from the shallow layer of the first convolutional neural network model and a plurality of deep feature maps from the deep layer of the first convolutional neural network model; and S132, cascading the shallow feature maps and the deep feature maps by using the depth feature fusion module to obtain the mixed state feature maps.

More specifically, in step S140, feature map expansion is performed on the plurality of hybrid-state feature maps to obtain a plurality of hybrid-state expansion feature vectors. Considering that the implicit characteristic of the mixed state of the mixed material at each predetermined point in time has an association relationship not only at different positions in the mixed state detection image at each predetermined point in time but also has a dynamic change characteristic, that is, the mixed state characteristic information of the mixed material at each predetermined point in time has a dynamic association relationship. Therefore, in the technical solution of the present application, in order to sufficiently extract the dynamic change feature information of the mixing state feature of the mixed material at each predetermined time point in the time dimension, the feature map expansion is further performed on the plurality of mixing state feature maps so as to obtain a plurality of mixing state expansion feature vectors.

More specifically, in step S150, the plurality of mixed state expansion feature vectors are passed through a context encoder based on a converter to obtain a mixed state timing semantic understanding feature vector. And encoding the plurality of mixed state unfolding feature vectors in a context encoder based on a converter to extract mixed state feature of the mixed materials at each preset time point, wherein the mixed state feature is based on dynamic associated feature distribution information of time sequence global in a time dimension, so as to obtain a mixed state time sequence semantic understanding feature vector.

Accordingly, in one specific example, as shown in fig. 5, passing the plurality of mixed state expansion feature vectors through a context encoder based on a converter to obtain a mixed state timing semantic understanding feature vector includes: s151, performing global-based context semantic coding on the plurality of mixed state expansion feature vectors by using the context encoder based on the converter to obtain a plurality of context mixed state feature vectors; and S152, cascading the plurality of context mixed state feature vectors to obtain the mixed state timing semantic understanding feature vector.

Accordingly, in one specific example, as shown in fig. 6, performing global-based context semantic encoding on the plurality of hybrid-state expansion feature vectors using the converter-based context encoder to obtain a plurality of context hybrid-state feature vectors, including: s1511, performing one-dimensional arrangement on the plurality of mixed state expansion feature vectors to obtain a global mixed state feature vector; s1512, calculating the product between the global mixed state eigenvector and the transpose vector of each mixed state expansion eigenvector in the mixed state expansion eigenvectors to obtain a plurality of self-attention association matrices; s1513, respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; s1514, obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and S1515, weighting each of the plurality of mixed state expansion feature vectors by using each of the plurality of probability values as a weight to obtain the plurality of context mixed state feature vectors.

More specifically, in step S160, a responsiveness estimate of the mixed state timing semantic understanding feature vector relative to the stirring speed timing feature vector is calculated to obtain a classification feature matrix. And the correlation characteristic distribution information of the time sequence dynamic change characteristic of the stirring speed and the time sequence dynamic correlation characteristic of the mixing state of the mixed materials is expressed.

Accordingly, in one specific example, as shown in fig. 7, calculating a responsiveness estimate of the mixed state timing semantic understanding feature vector relative to the stirring speed timing feature vector to obtain a classification feature matrix includes: s161, carrying out feature distribution optimization on the mixed state time sequence semantic understanding feature vector based on the stirring speed time sequence feature vector and the mixed state time sequence semantic understanding feature vector so as to obtain an optimized mixed state time sequence semantic understanding feature vector; s162, constructing a Gaussian density map of the optimized mixed state time sequence semantic understanding feature vector and the stirring speed time sequence feature vector to obtain a first Gaussian density map and a second Gaussian density map; s163, calculating the response estimation of the first Gaussian density map relative to the second Gaussian density map to obtain a response Gaussian density map; and S164, performing Gaussian discretization on the Gaussian distribution of each position in the response Gaussian density map to obtain the classification characteristic matrix.

Accordingly, in one specific example, based on the stirring speed timing feature vector and the mixed state timing semantic understanding feature vector, performing feature distribution optimization on the mixed state timing semantic understanding feature vector to obtain an optimized mixed state timing semantic understanding feature vector, including: calculating incoherent sparse response type fusion of the stirring speed time sequence feature vector and the mixed state time sequence semantic understanding feature vector according to the following formula to obtain the optimized mixed state time sequence semantic understanding feature vector; wherein, the formula is:

Here, the incoherent sparse response fusion obtains incoherent sparse fusion representation among vectors through fuzzy bit distribution responsiveness of vector differences represented by a norm and true differential embedding responsiveness based on modulo constraint of differential vectors in the case of authenticity distribution (group-truth distribution) with initial response vectors as feature inter-domain responsiveness fusion, so as to extract response relation of probability distribution descriptiveness after feature vector fusion, thereby improving the mixed state timing semantic understanding feature vector V as incoherent sparse response fusion optimization ₂ ' fusion express effect on source vector and response vector with response relationship. Thus, the optimization is calculated based on the Gaussian density mapThe mixed state timing semantic understanding feature vector V ₂ ' time sequence characteristic vector V relative to the stirring speed ₁ The accuracy of the classification result of the obtained classification feature matrix can be improved. Therefore, the stirring speed value can be adaptively adjusted in real time and accurately based on the state change condition of the mixed materials, so that the stirring effect is improved, and the preparation quality and efficiency of the banana powder are improved.

More specifically, in step S170, the classification feature matrix is passed through a classifier to obtain a classification result indicating whether the stirring speed value at the current time point should be increased or decreased. That is, in the technical solution of the present application, the label of the classifier includes that the stirring speed value at the current time point should be increased (first label) and that the stirring speed value at the current time point should be decreased (second label), wherein the classifier determines to which classification label the classification feature matrix belongs through a soft maximum function. It should be understood that, in the technical solution of the present application, the classification label of the classifier is a control policy label of the stirring speed value, so after the classification result is obtained, the stirring speed value at the current time point can be adaptively adjusted based on the classification result, so as to improve the effect of stirring and mixing into flour, and improve the preparation quality and efficiency of the banana powder.

Accordingly, in one specific example, as shown in fig. 8, the classification feature matrix is passed through a classifier to obtain a classification result, where the classification result is used to indicate that the stirring speed value at the current time point should be increased or decreased, and includes: s171, the classification feature matrix is unfolded into classification feature vectors according to row vectors or column vectors; s172, performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and S173, inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.

Accordingly, in one specific example, the banana powder described herein is made by a method of making banana powder as described in any one of the foregoing.

In summary, according to the banana powder and the preparation method thereof according to the embodiments of the present application, firstly, stirring speed values at a plurality of predetermined time points are arranged and then a one-dimensional convolutional neural network model is used to obtain stirring speed time sequence feature vectors, then, mixed state detection images at a plurality of predetermined time points are respectively passed through a first convolutional neural network model to obtain a plurality of mixed state feature diagrams, then, feature diagram expansion is performed on the plurality of mixed state feature diagrams and a context encoder is used to obtain mixed state time sequence semantic understanding feature vectors, then, response estimation of the mixed state time sequence semantic understanding feature vectors relative to the stirring speed time sequence feature vectors is calculated to obtain a classification feature matrix, and finally, the classification feature matrix is passed through a classifier to obtain a classification result for indicating that the stirring speed value at the current time point should be increased or decreased. Thus, the preparation quality and efficiency of the banana powder can be improved.

Fig. 9 is a block diagram of banana powder and its preparation system 100 according to an embodiment of the present application. As shown in fig. 9, the banana powder and its preparing system 100 according to the embodiment of the present application includes: a data acquisition module 110, configured to acquire stirring speed values at a plurality of predetermined time points within a predetermined time period and a mixed state detection image of a mixed material consisting of banana powder, wet starch and whipped cream at the plurality of predetermined time points; the one-dimensional convolutional encoding module 120 is configured to arrange the stirring speed values at the plurality of predetermined time points into a stirring speed input vector according to a time dimension, and then obtain a stirring speed time sequence feature vector through a one-dimensional convolutional neural network model; the depth feature fusion module 130 is configured to pass the mixed state detection images at the plurality of predetermined time points through a first convolutional neural network model including the depth feature fusion module to obtain a plurality of mixed state feature graphs; the feature map expansion module 140 is configured to perform feature map expansion on the plurality of hybrid state feature maps to obtain a plurality of hybrid state expansion feature vectors; a context encoding module 150 for passing the plurality of mixed state expanded feature vectors through a converter-based context encoder to obtain a mixed state temporal semantic understanding feature vector; a responsiveness estimation module 160, configured to calculate a responsiveness estimate of the mixed state timing semantic understanding feature vector relative to the stirring speed timing feature vector to obtain a classification feature matrix; and a classification module 170, configured to pass the classification feature matrix through a classifier to obtain a classification result, where the classification result is used to indicate that the stirring speed value at the current time point should be increased or decreased.

In one example, in the banana powder and its preparation system 100, the one-dimensional convolutional encoding module 120 is configured to: each layer of the one-dimensional convolutional neural network model is used for respectively carrying out input data in forward transfer of the layer: performing convolution processing on the input data to obtain a convolution feature vector; pooling the convolution feature vectors to obtain pooled feature vectors; and performing nonlinear activation on the pooled feature vectors to obtain activated feature vectors; the output of the last layer of the one-dimensional convolutional neural network model is the stirring speed time sequence feature vector, and the input of the first layer of the one-dimensional convolutional neural network model is the stirring speed input vector.

In one example, in the banana powder and its preparation system 100, the depth feature fusion module 130 is configured to: respectively inputting the mixed state detection images of the plurality of preset time points into the first convolutional neural network model to extract a plurality of shallow feature maps from the shallow layer of the first convolutional neural network model and a plurality of deep feature maps from the deep layer of the first convolutional neural network model; and cascading the shallow feature maps and the deep feature maps by using the depth feature fusion module to obtain the mixed state feature maps.

In one example, in the banana powder and its preparation system 100 described above, the context encoding module 150 is configured to: performing global-based context semantic coding on the plurality of mixed state expansion feature vectors by using the context encoder based on the converter to obtain a plurality of context mixed state feature vectors; and cascading the plurality of context mixed state feature vectors to obtain the mixed state timing semantic understanding feature vector.

In one example, in the banana powder and its preparation system 100, the global-based context semantic encoding of the plurality of mixed state expansion feature vectors using the converter-based context encoder to obtain a plurality of context mixed state feature vectors includes: one-dimensional arrangement is carried out on the plurality of mixed state expansion feature vectors so as to obtain global mixed state feature vectors; calculating the product between the global mixed state feature vector and the transpose vector of each mixed state expansion feature vector in the mixed state expansion feature vectors to obtain a plurality of self-attention association matrixes; respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and weighting each of the plurality of mixed state expansion feature vectors with each of the plurality of probability values as a weight to obtain the plurality of context mixed state feature vectors.

In one example, in the banana powder and its preparation system 100 described above, the responsiveness estimation module 160 is configured to: based on the stirring speed time sequence feature vector and the mixed state time sequence semantic understanding feature vector, carrying out feature distribution optimization on the mixed state time sequence semantic understanding feature vector to obtain an optimized mixed state time sequence semantic understanding feature vector; constructing a Gaussian density map of the optimized mixed state time sequence semantic understanding feature vector and the stirring speed time sequence feature vector to obtain a first Gaussian density map and a second Gaussian density map; calculating a responsiveness estimate of the first gaussian density map relative to the second gaussian density map to obtain a responsive gaussian density map; and performing Gaussian discretization on the Gaussian distribution of each position in the responsive Gaussian density map to obtain the classification feature matrix.

In one example, in the banana powder and its preparation system 100, based on the stirring speed sequential feature vector and the mixed state sequential semantic understanding feature vector, performing feature distribution optimization on the mixed state sequential semantic understanding feature vector to obtain an optimized mixed state sequential semantic understanding feature vector includes: calculating incoherent sparse response type fusion of the stirring speed time sequence feature vector and the mixed state time sequence semantic understanding feature vector according to the following formula to obtain the optimized mixed state time sequence semantic understanding feature vector; wherein, the formula is:

In one example, in the banana powder and its preparation system 100 described above, the classification module 170 is configured to: expanding the classification feature matrix into classification feature vectors according to row vectors or column vectors; performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.

Here, it will be understood by those skilled in the art that the specific functions and operations of the above-described banana powder and the respective units and modules in the preparation system 100 thereof have been described in detail in the above description of the banana powder and the preparation method thereof with reference to fig. 1 to 8, and thus, repeated descriptions thereof will be omitted.

As described above, the banana powder and its preparing system 100 according to the embodiment of the present application may be implemented in various wireless terminals, such as a server having banana powder and its preparing algorithm, etc. In one example, banana powder and its preparation system 100 according to embodiments of the present application may be integrated into a wireless terminal as one software module and/or hardware module. For example, the banana powder and its preparation system 100 may be a software module in the operating system of the wireless terminal or may be an application developed for the wireless terminal; of course, the banana powder and its preparation system 100 may also be one of the many hardware modules of the wireless terminal.

Alternatively, in another example, the banana powder and its preparation system 100 and the wireless terminal may also be separate devices, and the banana powder and its preparation system 100 may be connected to the wireless terminal through a wired and/or wireless network and transmit interactive information in a agreed data format.

According to another aspect of the present application, there is also provided a non-volatile computer-readable storage medium having stored thereon computer-readable instructions which, when executed by a computer, can perform a method as described above.

Program portions of the technology may be considered to be "products" or "articles of manufacture" in the form of executable code and/or associated data, embodied or carried out by a computer readable medium. A tangible, persistent storage medium may include any memory or storage used by a computer, processor, or similar device or related module. Such as various semiconductor memories, tape drives, disk drives, or the like, capable of providing storage functionality for software.

All or a portion of the software may sometimes communicate over a network, such as the internet or other communication network. Such communication may load software from one computer device or processor to another. For example: a hardware platform loaded from a server or host computer of the video object detection device to a computer environment, or other computer environment implementing the system, or similar functioning system related to providing information needed for object detection. Thus, another medium capable of carrying software elements may also be used as a physical connection between local devices, such as optical, electrical, electromagnetic, etc., propagating through cable, optical cable, air, etc. Physical media used for carrier waves, such as electrical, wireless, or optical, may also be considered to be software-bearing media. Unless limited to a tangible "storage" medium, other terms used herein to refer to a computer or machine "readable medium" mean any medium that participates in the execution of any instructions by a processor.

This application uses specific words to describe embodiments of the application. Reference to "a first/second embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic is associated with at least one embodiment of the present application. Thus, it should be emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various positions in this specification are not necessarily referring to the same embodiment. Furthermore, certain features, structures, or characteristics of one or more embodiments of the present application may be combined as suitable.

Furthermore, those skilled in the art will appreciate that the various aspects of the invention are illustrated and described in the context of a number of patentable categories or circumstances, including any novel and useful procedures, machines, products, or materials, or any novel and useful modifications thereof. Accordingly, aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, micro-code, etc.) or by a combination of hardware and software. The above hardware or software may be referred to as a "data block," module, "" engine, "" unit, "" component, "or" system. Furthermore, aspects of the present application may take the form of a computer product, comprising computer-readable program code, embodied in one or more computer-readable media.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

The foregoing is illustrative of the present invention and is not to be construed as limiting thereof. Although a few exemplary embodiments of this invention have been described, those skilled in the art will readily appreciate that many modifications are possible in the exemplary embodiments without materially departing from the novel teachings and advantages of this invention. Accordingly, all such modifications are intended to be included within the scope of this invention as defined in the following claims. It is to be understood that the foregoing is illustrative of the present invention and is not to be construed as limited to the specific embodiments disclosed, and that modifications to the disclosed embodiments, as well as other embodiments, are intended to be included within the scope of the appended claims. The invention is defined by the claims and their equivalents.

Claims

1. A banana powder and a preparation method thereof, which is characterized by comprising the following steps:

2. The banana powder and its preparation process as claimed in claim 1, wherein the mixing speed values at the predetermined time points are arranged in time dimension as mixing speed input vector and then passed through one-dimensional convolutional neural network model to obtain mixing speed time sequence feature vector, comprising:

pooling the convolution feature vectors to obtain pooled feature vectors; and

3. The banana powder and its preparing process according to claim 2, wherein the step of passing the mixed state detection images at the predetermined time points through the first convolutional neural network model including the depth feature fusion module to obtain the mixed state feature maps includes:

4. A banana powder and its preparation method according to claim 3, characterised in that passing the plurality of mixed state expansion feature vectors through a transducer-based context encoder to obtain a mixed state timing semantic understanding feature vector comprises:

5. Banana powder and its preparation method according to claim 4, characterized in that the global-based context semantic coding of the multiple mixed state expansion feature vectors using the converter-based context encoder to obtain multiple context mixed state feature vectors comprises:

6. Banana powder and its preparation method according to claim 5, characterized in that calculating the response estimate of the mixed state timing semantic understanding feature vector relative to the stirring speed timing feature vector to get a classification feature matrix comprises:

7. The banana powder and its preparation method according to claim 6, wherein the feature distribution optimization of the mixed state timing semantic understanding feature vector based on the stirring speed timing feature vector and the mixed state timing semantic understanding feature vector to obtain an optimized mixed state timing semantic understanding feature vector comprises:

Wherein, the formula is:

8. Banana powder and its preparation according to claim 7, characterised in that the classification feature matrix is passed through a classifier to obtain a classification result, which is used to indicate whether the stirring speed value at the current time point should be increased or decreased, comprising:

9. Banana powder, characterized in that it is produced by the process of preparation of banana powder according to any one of claims 1-8.