Fresh-keeping parameter dynamic optimization method and system based on waxberry state matching analysis

Abstract

The invention relates to the technical field of fruit fresh-keeping, and particularly discloses a fresh-keeping parameter dynamic optimization method and system based on waxberry state matching analysis, and the method comprises the steps: firstly constructing a waxberry state prototype image-recommended fresh-keeping parameter prior database based on experimental data; and then performing image matching analysis on each red bayberry state prototype image in the prior database and the red bayberry real-time state image at the current time point by adopting a machine vision technology based on deep learning so as to query adaptive fresh-keeping parameters from the prior database, thereby realizing intelligent control on real-time environment parameters in the database. The accuracy and efficiency of waxberry preservation can be effectively improved, meanwhile, economic losses caused by improper preservation are reduced, and a more scientific and accurate preservation solution is provided for waxberry production and sales.

Description

Technical Field

[0001] The present application relates to the technical field of fruit preservation, and more specifically, to a method and system for dynamically optimizing preservation parameters based on bayberry state matching analysis. Background Art

[0002] Bayberry is a seasonal fruit with high nutritional value and unique taste, and enjoys a high reputation in the market. However, due to its thin peel and high water content, it is very susceptible to the influence of the external environment and deteriorates, making it difficult to keep fresh.

[0003] During the preservation process of bayberry, the state changes of bayberry (such as color change, degree of decay, surface damage, etc.) are important bases for judging its freshness and formulating preservation strategies. However, the traditional preservation method mainly sets the cold storage preservation parameters through manual observation and experience judgment in the hope of maintaining the freshness of bayberry. However, this method is not only easily affected by the subjective experience judgment of the operator, but also difficult to adapt to the dynamic changes of the state of bayberry and the diversity of preservation needs, thus affecting the preservation effect and economic benefits.

[0004] Therefore, a dynamic optimization scheme for preservation parameters based on bayberry state matching analysis is expected. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a method and system for dynamic optimization of preservation parameters based on bayberry state matching analysis, which first constructs a priori database of bayberry state prototype images-recommended preservation parameters based on experimental data, and then uses machine vision technology based on deep learning to perform image matching analysis on each bayberry state prototype image in the priori database and the bayberry real-time state image at the current time point to query the adapted preservation parameters from the priori database. In this way, intelligent control of real-time environmental parameters in the warehouse is achieved. In this way, the accuracy and efficiency of bayberry preservation can be effectively improved, while reducing the economic losses caused by improper preservation, providing a more scientific and accurate preservation solution for the production and sales of bayberry. The warehouse refers to the inside of the AI ​​preservation warehouse.

[0006] Accordingly, according to one aspect of the present application, a method for dynamic optimization of preservation parameters based on bayberry state matching analysis is provided, which comprises: Construct a priori database of prototype images of bayberry states and recommended preservation parameters; Obtain the real-time status image of the bayberry in the warehouse collected by the camera; Extracting the bayberry state features from the bayberry real-time state image in the warehouse and the bayberry state prototype image-each bayberry state prototype image in the prior database of recommended preservation parameters to obtain a set of the bayberry real-time state semantic coding feature map and the bayberry state prototype semantic coding feature map; Performing spatial enhancement processing on the real-time state semantic coding feature map of the bayberry in the warehouse to obtain the real-time state semantic enhancement coding feature map of the bayberry in the warehouse; Based on the semantic similarity between the real-time state semantic enhancement coding feature graph of bayberry in the warehouse and each bayberry state prototype semantic coding feature graph in the set of bayberry state prototype semantic coding feature graphs, the bayberry preservation control strategy is determined.

[0007] According to another aspect of the present application, a dynamic optimization system for preservation parameters based on bayberry state matching analysis is provided, which includes: A priori database construction module is used to construct a priori database of prototype images of bayberry states and recommended preservation parameters; The real-time status monitoring module of bayberry is used to obtain the real-time status image of bayberry in the warehouse collected by the camera; A bayberry state feature extraction module is used to extract bayberry state features from the bayberry real-time state image in the warehouse and the bayberry state prototype image-each bayberry state prototype image in the prior database of recommended preservation parameters to obtain a set of the bayberry real-time state semantic coding feature map and the bayberry state prototype semantic coding feature map; A feature enhancement module is used to perform spatial enhancement processing on the real-time state semantic coding feature map of the bayberry in the warehouse to obtain a real-time state semantic enhancement coding feature map of the bayberry in the warehouse; The comparison and matching module is used to determine the bayberry preservation control strategy based on the semantic similarity between the bayberry real-time state semantic enhancement coding feature map in the library and the bayberry state prototype semantic coding feature map set.

[0008] Compared with the prior art, the method and system for dynamic optimization of preservation parameters based on bayberry state matching analysis provided in the present application first construct a priori database of bayberry state prototype images-recommended preservation parameters based on experimental data, and then use machine vision technology based on deep learning to perform image matching analysis on each bayberry state prototype image in the priori database and the bayberry real-time state image at the current time point to query the adapted preservation parameters from the priori database. In this way, intelligent control of real-time environmental parameters in the warehouse is achieved. In this way, the accuracy and efficiency of bayberry preservation can be effectively improved, while reducing the economic losses caused by improper preservation, providing a more scientific and accurate preservation solution for the production and sale of bayberry. The warehouse refers to the interior of the AI ​​preservation warehouse. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 The present invention is a flowchart of a method for dynamically optimizing preservation parameters based on bayberry state matching analysis according to an embodiment of the present application.

[0011] Figure 2 This is a data flow diagram of a method for dynamically optimizing preservation parameters based on bayberry state matching analysis according to an embodiment of the present application.

[0012] Figure 3 This is a flowchart of step S140 in the method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to an embodiment of the present application.

[0013] Figure 4 This is a flowchart of step S150 in the method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to an embodiment of the present application.

[0014] Figure 5 It is a block diagram of a system for dynamic optimization of preservation parameters based on bayberry state matching analysis according to an embodiment of the present application.

[0015] Figure 6 This is a schematic diagram of the results of measuring the ethanol and acetaldehyde contents of bayberry samples according to an embodiment of the present application.

[0016] Figure 7 This is a schematic diagram of a photo comparison of the preservation effects of bayberry samples according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0018] Figure 1 The present invention is a flowchart of a method for dynamically optimizing preservation parameters based on bayberry state matching analysis according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the dynamic optimization method of preservation parameters based on bayberry state matching analysis includes the following steps: S110, constructing a priori database of bayberry state prototype images-recommended preservation parameters; S120, acquiring a real-time state image of the bayberry in the warehouse collected by a camera; S130, performing bayberry state feature extraction on the real-time state image of the bayberry in the warehouse and each bayberry state prototype image in the priori database of the bayberry state prototype images-recommended preservation parameters to obtain a set of semantic coding feature maps of the real-time state of the bayberry in the warehouse and bayberry state prototype semantic coding feature maps; S140, performing spatial enhancement processing on the semantic coding feature map of the real-time state of the bayberry in the warehouse to obtain a semantic enhanced coding feature map of the real-time state of the bayberry in the warehouse; S150, determining the bayberry preservation control strategy based on the semantic similarity between the real-time state semantic enhanced coding feature map of the bayberry in the warehouse and each bayberry state prototype semantic coding feature map in the set of the bayberry state prototype semantic coding feature map.

[0019] The step S110 is to construct a priori database of the prototype image of the state of bayberry and the recommended preservation parameters. It should be understood that the priori database is constructed based on a large amount of experimental data, which contains bayberry images of different maturity and different states and their corresponding recommended preservation parameters, covering the state changes of bayberry under different growth cycles and different environmental conditions, and can provide a comprehensive data reference for the preservation control of bayberry. Specifically, the construction process of the priori database of the prototype image of the state of bayberry and the recommended preservation parameters may include the following steps: First, it is necessary to collect bayberry samples at different maturity, different states and different environmental conditions. These samples should be able to represent the various states that may occur in the storage process of bayberry. Then, for each collected bayberry sample, high-quality images are taken under different environmental conditions (such as different temperatures, humidity, etc.) to form a prototype image of the state of bayberry. These images will serve as the basis for subsequent analysis. Then, a series of experimental tests are carried out on each sample to determine its optimal preservation parameters under specific environmental conditions. This may involve measuring the changes in indicators such as the physical properties (such as hardness, color), chemical composition (such as sugar content, acidity), and microbial activity of bayberry. It is worth mentioning that during the specific implementation process, it is also necessary to verify the effectiveness of the database through practical applications, and continuously adjust and optimize the database content based on feedback to ensure that it can accurately reflect the actual preservation needs of bayberry.

[0020] The step S120 is to obtain the real-time state image of the bayberry in the warehouse collected by the camera. It should be understood that the bayberry will be affected by various environmental factors such as temperature and humidity during storage, causing its biological state to change accordingly, and the demand for preservation parameters will also change accordingly. Therefore, in the technical solution of the present application, the real-time state image of the bayberry in the warehouse is collected by the camera in real time, so that the current state of the bayberry can be understood in time, so as to make corresponding preservation adjustments and ensure the preservation effect of the bayberry.

[0021] Specifically, the process of collecting the real-time state image of the bayberry in the warehouse can include the following steps in the specific implementation: first, select a suitable position inside the storage warehouse to install the camera to ensure that the camera can cover all the areas where the bayberry is stored. Then, adjust the angle and height of the camera to ensure that the captured image is clear and complete, and can reflect the overall state of the bayberry. At the same time, set the resolution and frame rate of the camera to ensure that the image quality and transmission speed meet the requirements of real-time monitoring. It is worth mentioning that it is necessary to ensure that the lighting conditions in the storage warehouse are suitable to avoid excessive or weak light affecting the image quality. In this way, the camera regularly or continuously captures the image of the bayberry in the warehouse, and the frequency can be adjusted according to actual needs. For example, it can be set to capture an image once a minute, or trigger the capture when a change in environmental conditions is detected. In terms of data transmission, the camera transmits the captured image to the central processing system via a wired or wireless network, and the central processing system stores the received image for subsequent processing and analysis. The stored image data should have a timestamp to facilitate tracking the change process of the bayberry state.

[0022] In step S130, the real-time state image of the bayberry in the warehouse and each prototype image of the bayberry state in the prior database of the prototype image of the bayberry state-recommended fresh-keeping parameters are respectively subjected to bayberry state feature extraction to obtain a set of semantic coding feature maps of the real-time state of the bayberry in the warehouse and semantic coding feature maps of the prototype of the bayberry state. It should be understood that in order to realize the comparative analysis between the real-time state image of the bayberry in the warehouse and each prototype image of the bayberry state in the prior database, so as to query the most suitable fresh-keeping parameters from the prior database, it is necessary to further perform image feature extraction on the real-time state image of the bayberry in the warehouse and each prototype image of the bayberry state to mine key features such as maturity, color, and texture of the bayberry.

[0023] In a specific example of the present application, the real-time state image of the bayberry in the warehouse and the prototype image of the bayberry state - each prototype image of the bayberry state in the prior database of the recommended preservation parameters are respectively input into the bayberry state feature extractor based on the deep separable convolutional neural network model to obtain the real-time state semantic coding feature map of the bayberry in the warehouse and the set of the prototype semantic coding feature map of the bayberry state. That is, the deep separable convolutional neural network model is used to achieve efficient extraction of bayberry state features. In particular, the deep separable convolutional neural network can reduce the amount of calculation and model parameters while maintaining a high feature extraction capability by decomposing the standard convolution into deep convolution and point-by-point convolution. This method can extract key information from the image layer by layer and improve the recognition accuracy of the bayberry state features. The final generated set of the real-time state semantic coding feature map of the bayberry in the warehouse and the prototype semantic coding feature map of the bayberry state contains high-level semantic information such as the color, shape, and texture of the bayberry.

[0024] The step S140 performs spatial enhancement processing on the semantic coding feature map of the real-time state of the bayberry in the warehouse to obtain the semantic enhanced coding feature map of the real-time state of the bayberry in the warehouse. It should be understood that, considering that in an actual storage environment, the real-time state image of the bayberry in the warehouse may produce noise or blur due to the influence of ambient light, shadow or other factors, thereby affecting the accuracy of subsequent feature comparison analysis. Therefore, in order to improve the accuracy of feature expression, the present application further performs feature enhancement processing on the semantic coding feature map of the real-time state of the bayberry in the warehouse, and based on the contextual semantic association information of the semantic coding feature map of the real-time state of the bayberry in the warehouse, performs spatial attention enhancement on it to highlight the key features of the expression of the bayberry, while suppressing noise and irrelevant information, thereby obtaining the semantic enhanced coding feature map of the real-time state of the bayberry in the warehouse.

[0025] Figure 3 FIG. 1 is a flow chart of step S140 in the method for dynamically optimizing preservation parameters based on bayberry state matching analysis according to an embodiment of the present application. Figure 3 As shown, the step S140 includes: S141, calculating the pixel-level semantic association of the semantic coding feature map of the real-time state of the bayberry in the warehouse to obtain the pixel-granularity semantic association topological matrix of the real-time state of the bayberry in the warehouse; S142, based on the pixel spatial position information of the semantic coding feature map of the real-time state of the bayberry in the warehouse, performing spatial attention attenuation modulation on the pixel-granularity semantic association topological matrix of the real-time state of the bayberry in the warehouse to obtain the bayberry state semantic association space soft constraint topological feature matrix; S143, based on the bayberry state semantic association space soft constraint topological feature matrix, performing pixel-granularity spatial enhancement processing on the semantic coding feature map of the real-time state of the bayberry in the warehouse to obtain the semantic enhancement coding feature map of the real-time state of the bayberry in the warehouse.

[0026] Specifically, the step S141 includes: performing feature dispersion on the semantic coding feature map of the real-time state of bayberry in the warehouse along the channel dimension to obtain a set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse; calculating the semantic association score between any two pixel granularity feature vectors of the real-time state of bayberry in the warehouse in the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse to obtain the pixel granularity semantic association topological matrix of the real-time state of bayberry in the warehouse. In a specific example of the present application, the semantic association score between any two pixel granularity feature vectors of the real-time state of bayberry in the warehouse in the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse is calculated, including: performing nonlinear mapping on each pixel granularity feature vector of the real-time state of bayberry in the warehouse in the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse to obtain a set of mapped pixel granularity feature vectors of the real-time state of bayberry in the warehouse; calculating the Poincare distance between any two mapped pixel granularity feature vectors of the real-time state of bayberry in the warehouse in the set of mapped pixel granularity feature vectors as the semantic association score to obtain the semantic association topological matrix of the pixel granularity of the real-time state of bayberry in the warehouse.

[0027] That is, first, the image of the real-time state of bayberry in the warehouse is feature scattered along the channel dimension to split it into a feature vector set of pixel granularity. Then, by calculating the semantic association score between any two pixel granularity feature vectors of the real-time state of bayberry in the warehouse in the set, a semantic association topological matrix of the real-time state of bayberry in the warehouse with pixel granularity is constructed, thereby revealing the intrinsic semantic relevance of the image. In particular, in this process, considering that the hyperbolic space can more effectively capture the nonlinear relationship between data compared with the Euclidean space, the present application first maps the pixel granularity feature vectors of the real-time state of bayberry in each warehouse from the Euclidean space to the Poincare space, so that it adapts to the geometric characteristics of the hyperbolic space while maintaining the original semantic information, and calculates the Poincare distance between any two mapped pixel granularity feature vectors of the real-time state of bayberry in the warehouse as the semantic association score between the two, so as to utilize the geometric characteristics of the Poincare space to enhance the nonlinear association between features, and improve the model's ability to distinguish features of different categories, thereby more accurately simulating the intrinsic semantic association structure of the image.

[0028] Specifically, the step S142 includes: performing spatial attention attenuation modulation on the real-time state pixel granularity semantic association topology matrix of the bayberry in the warehouse based on the spatial distance between any two real-time state pixel granularity feature vectors of the bayberry in the warehouse in the set of real-time state pixel granularity feature vectors of the bayberry in the warehouse to obtain the bayberry state semantic association space soft constraint topology matrix; performing void convolution coding on the bayberry state semantic association space soft constraint topology matrix to obtain the bayberry state semantic association space soft constraint topology feature matrix. More specifically, the spatial attention attenuation modulation is performed on the pixel granularity semantic association topology matrix of the real-time state of bayberry in the warehouse, including: taking the natural exponential function value of the Euclidean distance between any two pixel granularity feature vectors of the real-time state of bayberry in the warehouse in the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse as the spatial distance factor to obtain the spatial distance topology matrix of the pixel granularity of the real-time state of bayberry in the warehouse; multiplying the spatial distance topology matrix of the pixel granularity of the real-time state of bayberry in the warehouse by a preset attenuation modulation constant to obtain the spatial attention attenuation modulation matrix, and calculating the position point-by-position division of the pixel granularity semantic association topology matrix of the real-time state of bayberry in the warehouse and the spatial attention attenuation modulation matrix to obtain the bayberry state semantic association space soft constraint topology matrix.

[0029] It should be understood that, considering that in practical applications, adjacent image regions often have stronger correlations than distant regions, this application further performs spatial attention attenuation modulation on the above-mentioned semantic association topological matrix of the real-time state of bayberry in the library based on the spatial distance between any two pixel granularity feature vectors of the real-time state of bayberry in the library, so as to simulate the association attenuation characteristics between local regions of the image, so as to pay more attention to the feature association of adjacent regions and weaken the feature association of distant regions, thereby improving the distinguishability of features. Furthermore, by using the hole convolution technology, the semantic association topological matrix adjusted by spatial soft constraints is subjected to hole convolution encoding to capture the semantic association topological pattern between the pixel granularity features of the real-time state of bayberry in each library, so as to achieve a deeper understanding of the contextual semantic association structure of the bayberry state, and generate the topological feature matrix of the semantic association space soft constraint of the bayberry state.

[0030] Specifically, the step S143 includes: inputting the set of pixel granularity feature vectors of the real-time status of bayberry in the warehouse and the soft-constraint topological feature matrix of the semantic association space of the bayberry status into the graph convolutional coding module to obtain a set of shallow pixel granularity feature vectors of the bayberry status enhanced by context semantic association; and performing feature reshaping on the set of shallow pixel granularity feature vectors of the bayberry status enhanced by context semantic association to obtain the semantically enhanced coding feature map of the real-time status of bayberry in the warehouse.

[0031] That is, the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse and the soft-constrained topological feature matrix of the semantic association space of the bayberry state are processed through graph convolution coding technology, and the graph convolution characteristics are utilized to update the feature representation of each pixel granularity of the image based on the semantic association information between pixels. By learning the intrinsic semantic association structure of the image, the global semantic context association feature expression of the image pixel granularity feature is enhanced, and by reshaping the feature shape of each pixel granularity feature after the context semantic association is strengthened, its original feature structure is restored to generate a semantically enhanced coding feature map of the real-time state of bayberry in the warehouse, thereby achieving refined enhancement of the local detail features of the image.

[0032] In summary, the step S140 includes: processing the real-time state semantic coding feature map of the bayberry in the warehouse with the following feature enhancement formula to obtain the real-time state semantic enhancement coding feature map of the bayberry in the warehouse, wherein the feature enhancement formula is:

[0033]

[0034]

[0035]

[0036]

[0037]

[0038]

[0039] in, is a semantic coding feature graph of the real-time state of bayberry in the warehouse, , and respectively represent the height, width and number of channels of the semantic coding feature graph of the real-time state of bayberry in the warehouse, represents the feature shape reshaping, The set of pixel granularity feature vectors representing the real-time status of bayberry in the warehouse, and The pixel granularity feature vectors of the real-time state of bayberry in the warehouse are respectively represented by and The pixel granularity feature vector of the real-time status of bayberry in the warehouse, The value of is the number of feature vectors in the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse, and , The value of is the length of the pixel granularity feature vector of the real-time state of bayberry in the warehouse, is the weight transformation matrix, is the weight transformation vector, and Respectively represent the pixel granularity feature vectors of the real-time state of bayberry in the library after mapping. and The pixel granularity feature vector of the real-time state of bayberry in the library after mapping, represents the inverse hyperbolic cosine function in hyperbolic space, represents the norm of the eigenvector, Indicates the The pixel granularity feature vector of the real-time state of bayberry in the library after mapping and the The Poincare distance between the pixel granularity feature vectors of the real-time state of bayberry in the library after mapping. The pixel-granular semantic association topology matrix represents the real-time status of bayberry in the warehouse. The semantic relevance score of the location, is the preset attenuation modulation constant, Indicates the The pixel granularity feature vector of the real-time state of bayberry in the warehouse and the The Euclidean distance between the pixel granularity feature vectors of the real-time status of bayberry in the warehouse. is the soft constraint topological matrix of the Yangmei state semantic association space, is the soft constraint topological matrix of the Yangmei state semantic association space The eigenvalues ​​of the positions, represents the dilated convolutional coding, is the soft constraint topological feature matrix of the Yangmei state semantic association space, represents graph convolution processing, The set of shallow pixel-level feature vectors representing the contextual semantic associations that enhance the state of the bayberry. The semantically enhanced coding feature graph represents the real-time status of bayberries in the warehouse.

[0040] The step S150 is to determine the waxberry preservation control strategy based on the semantic similarity between the waxberry real-time state semantic enhancement coding feature graph in the warehouse and each waxberry state prototype semantic coding feature graph in the set of the waxberry state prototype semantic coding feature graphs. Figure 4 FIG. 1 is a flow chart of step S150 in the method for dynamically optimizing preservation parameters based on bayberry state matching analysis according to an embodiment of the present application. Figure 4As shown, the step S150 includes: S151, respectively inputting each of the bayberry state prototype semantic coding feature maps in the set of the bayberry state real-time state semantic enhancement coding feature map and the bayberry state prototype semantic coding feature map into the bayberry state semantic comparison network based on the hash function to obtain a set of bayberry state semantic comparison coefficients; S152, extracting the recommended preservation parameters corresponding to the minimum value of the bayberry state semantic comparison coefficient in the set of bayberry state semantic comparison coefficients; S153, determining the bayberry preservation control strategy based on the comparison between the recommended preservation parameters and the real-time environmental parameters in the warehouse.

[0041] Specifically, in step S151, each of the prototype semantic coding feature graphs of the state of the bayberry in the library is respectively input into the bayberry state semantic comparison network based on the hash function to obtain a set of bayberry state semantic comparison coefficients. It should be understood that, considering that in a large-scale database, directly calculating the similarity between high-dimensional image features is not only very time-consuming, but also has high computational complexity. In this regard, the present application introduces a fast similarity retrieval method based on hash technology, by inputting the real-time state semantic enhancement coding feature graph of the bayberry in the library and each of the prototype semantic coding feature graphs of the state of the bayberry in the library into the hash function respectively, so as to map each high-dimensional image feature to the hash space, and generate a low-dimensional binary hash code corresponding to each image feature, so as to quickly calculate the similarity between the image features based on the Hamming distance between each binary hash code, and generate the bayberry state semantic comparison coefficient between the real-time state image of the bayberry in the library and each prototype image of the bayberry state, so as to achieve fast and accurate similarity retrieval.

[0042] Preferably, calculating the bayberry state semantic contrast coefficient between the bayberry real-time state semantic enhancement coding feature map and the bayberry state prototype semantic coding feature map comprises: respectively expanding the bayberry real-time state semantic enhancement coding feature map and the bayberry state prototype semantic coding feature map into a bayberry real-time state semantic enhancement coding feature vector and a bayberry state prototype semantic coding feature vector, and further cascading them into a bayberry real-time state-bayberry state prototype cascade feature vector; calculating the product of a first bayberry real-time state-bayberry state prototype cascade feature value of the bayberry real-time state-bayberry state prototype cascade feature vector and the bayberry state semantic contrast coefficient to obtain a first bayberry real-time state-bayberry state prototype cascade feature vector. The product of the probability of the cascade feature of the real-time state of the bayberry in the warehouse-the prototype cascade feature of the bayberry state; the product of the real-time state of the bayberry in the second warehouse-the prototype cascade feature value of the bayberry state in the warehouse and the difference between one and the bayberry state semantic contrast coefficient is calculated to obtain the inverse probability product of the real-time state of the bayberry in the second warehouse-the prototype cascade feature of the bayberry state; the quotient of the real-time state of the bayberry in the first warehouse-the prototype cascade feature value of the bayberry state is divided by the difference between one and the semantic contrast coefficient of the bayberry state to obtain the inverse probability quotient of the real-time state of the bayberry in the first warehouse-the prototype cascade feature of the bayberry state; the product of the real-time state of the bayberry in the second warehouse-the prototype cascade feature value of the bayberry state and the semantic contrast coefficient of the bayberry state is calculated. to obtain the probability quotient of the cascade feature of the real-time state of the bayberry in the second warehouse - the prototype cascade feature of the bayberry state; after adding the product of the probability of the cascade feature of the real-time state of the bayberry in the first warehouse - the prototype cascade feature of the bayberry state in the second warehouse and the inverse probability product of the real-time state of the bayberry in the first warehouse - the prototype cascade feature of the bayberry state, subtract the difference between the inverse probability quotient of the real-time state of the bayberry in the first warehouse - the prototype cascade feature of the bayberry state in the second warehouse and the probability quotient of the real-time state of the bayberry in the second warehouse - the prototype cascade feature of the bayberry state, to obtain the corrected characteristic value of the real-time state of the bayberry in the warehouse - the prototype cascade feature of the bayberry state; for each first real-time state of the bayberry in the warehouse - the prototype cascade feature value of the real-time state of the bayberry in the second warehouse and the real-time state of the bayberry in the second warehouse state-bayberry state prototype cascade characteristic value, the bayberry real-time state in the warehouse-bayberry state prototype cascade correction feature matrix composed of the bayberry real-time state in the warehouse-bayberry state prototype cascade correction characteristic value is matrix-multiplied with the bayberry real-time state in the warehouse-bayberry state prototype cascade characteristic vector to obtain the corrected bayberry real-time state in the warehouse-bayberry state prototype cascade feature vector; the corrected bayberry real-time state in the warehouse-bayberry state prototype cascade feature vector is split into a corrected bayberry real-time state semantic enhancement coding feature vector and a corrected bayberry state prototype semantic coding feature vector, and restored to a corrected bayberry real-time state semantic enhancement coding feature map and a corrected bayberry state prototype semantic coding feature map;The corrected real-time state semantic enhancement coding feature graph of bayberry in the warehouse and the corrected prototype semantic coding feature graph of bayberry state are input into the bayberry state semantic comparison network based on the hash function to obtain an updated bayberry state semantic comparison coefficient. ;

[0043] The correction process of the real-time state of bayberry in the warehouse and the bayberry state prototype cascade feature vector can be expressed by the following correction formula:

[0044]

[0045]

[0046] in, represents the real-time state of bayberry in the warehouse - the prototype cascade feature vector of bayberry state, is the semantic contrast coefficient of the bayberry state, and The first and second cascade feature vectors of the real-time state of bayberry in the warehouse and the prototype state of bayberry are represented respectively. and eigenvalues, Indicates the real-time status of bayberry in each warehouse - the cascade correction characteristic value of the bayberry status prototype, The real-time state of bayberry in the warehouse and the cascade correction feature matrix of the prototype of the bayberry state composed of the real-time state of bayberry in each warehouse and the prototype of the bayberry state cascade correction feature value are represented. represents the matrix multiplication operation, Represents the corrected real-time state of bayberry in the warehouse - bayberry state prototype cascade feature vector.

[0047] That is, when the real-time semantic enhancement coding feature map of the bayberry state in the warehouse and the bayberry state prototype semantic coding feature map respectively represent the image semantic features of the real-time state image of the bayberry in the warehouse and the bayberry state prototype image, when the real-time semantic enhancement coding feature map of the bayberry state in the warehouse is subjected to feature enhancement based on spatial soft constraints and mapped to the common image semantic similarity space, the real-time semantic enhancement coding feature map of the bayberry state in the warehouse and the bayberry state prototype semantic coding feature map will also cause soft constraint enhancement diversity due to the difference in image semantic space distribution, affect the iterative consistency of the common space mapping, and reduce the calculation accuracy of the bayberry state semantic contrast coefficient.

[0048] Therefore, the present application targets the regional-boundary integral relationship of the joint high-dimensional feature manifold of the real-time semantically enhanced coding feature map of the bayberry state in the warehouse and the prototype semantically encoded feature map of the bayberry state, and defines the common image semantic similarity space mapping probability distribution of the feature set of the real-time semantically enhanced coding feature map of the bayberry state in the warehouse and the prototype semantically encoded feature map of the bayberry state as a manifold constraint boundary to approximate the high-dimensional manifold representation of the feature set of the real-time semantically enhanced coding feature map of the bayberry state in the warehouse and the prototype semantically encoded feature map of the bayberry state, which is composed of eigenvalue pairs, so as to avoid the ambiguity in the mapping of the diversified feature representations of the real-time semantically enhanced coding feature map of the bayberry state in the warehouse and the prototype semantically encoded feature map of the bayberry state to the local manifold representation in the high-dimensional feature convergence space, improve the execution iteration consistency of each local feature distribution in the mapping task, so as to improve the feature mapping effect and improve the calculation accuracy of the bayberry state semantic contrast coefficient.

[0049] Specifically, the step S152 extracts the recommended preservation parameters corresponding to the minimum value of the bayberry state semantic contrast coefficient in the set of bayberry state semantic contrast coefficients. It should be understood that the bayberry state semantic contrast coefficient reflects the semantic similarity between the real-time state image of the bayberry in the warehouse and each bayberry state prototype image in the priori database, wherein the minimum bayberry state semantic contrast coefficient corresponds to the most matching bayberry state prototype image. Therefore, by extracting the recommended preservation parameters corresponding to the bayberry state prototype image that best matches the real-time state of the bayberry in the warehouse as the bayberry preservation solution, the best preservation effect of the bayberry during the storage process can be ensured.

[0050] Specifically, the step S153 determines the bayberry preservation control strategy based on the comparison between the recommended preservation parameters and the real-time environmental parameters in the warehouse. That is, according to the comparison between the recommended preservation parameters and the real-time environmental parameters in the warehouse, the temperature, humidity and other condition parameters in the storage environment can be automatically adjusted to make the storage environment more suitable for the preservation of bayberry. For example, if the recommended preservation parameters indicate that bayberry needs to be stored at a lower temperature, and the current ambient temperature in the warehouse is high, the temperature in the warehouse can be lowered to meet the best preservation requirements of bayberry. On the contrary, if the recommended preservation parameters suggest a higher humidity, and the current ambient humidity is low, the system will increase the humidity accordingly. In this way, the preservation strategy of bayberry can be dynamically adjusted according to the matching degree between the real-time environmental parameters and the recommended preservation parameters, thereby realizing intelligent preservation management, extending its shelf life and maintaining the best quality.

[0051] In summary, the dynamic optimization method of preservation parameters based on bayberry state matching analysis according to the embodiment of the present application is explained, which first constructs a priori database of bayberry state prototype images-recommended preservation parameters based on experimental data, and then uses machine vision technology based on deep learning to perform image processing on each bayberry state prototype image in the priori database and the bayberry real-time state image at the current time point to mine the bayberry state semantic features in the image, and then, based on the semantic similarity between the bayberry real-time state image and each bayberry state prototype image, the most suitable preservation parameters are queried from the priori database, thereby realizing intelligent control of real-time environmental parameters in the warehouse. In this way, the accuracy and efficiency of bayberry preservation can be effectively improved, while reducing the economic losses caused by improper preservation, providing a more scientific and accurate preservation solution for the production and sales of bayberry.

[0052] Figure 5 FIG. 1 is a block diagram of a dynamic optimization system for preservation parameters based on bayberry state matching analysis according to an embodiment of the present application. Figure 5 As shown, according to the embodiment of the present application, the dynamic optimization system 100 of preservation parameters based on the matching analysis of the state of bayberry includes: a priori database construction module 110, which is used to construct a priori database of bayberry state prototype images-recommended preservation parameters; a bayberry real-time state monitoring module 120, which is used to obtain the real-time state image of the bayberry in the warehouse collected by the camera; a bayberry state feature extraction module 130, which is used to extract bayberry state features from the real-time state image of the bayberry in the warehouse and each bayberry state prototype image in the priori database of the bayberry state prototype images-recommended preservation parameters to obtain a set of semantic coding feature maps of the real-time state of the bayberry in the warehouse and the prototype semantic coding feature maps of the bayberry state; a feature enhancement module 140, which is used to perform spatial enhancement processing on the semantic coding feature map of the real-time state of the bayberry in the warehouse to obtain a semantic enhanced coding feature map of the real-time state of the bayberry in the warehouse; a comparison and matching module 150, which is used to determine the bayberry preservation control strategy based on the semantic similarity between the real-time state semantic enhanced coding feature map of the bayberry in the warehouse and each bayberry state prototype semantic coding feature map in the set of the bayberry state prototype semantic coding feature map.

[0053] Here, those skilled in the art can understand that the specific operations of each module in the above-mentioned dynamic optimization system for preservation parameters based on bayberry state matching analysis have been described in the above reference. Figures 1 to 4 The description of the dynamic optimization method of preservation parameters based on bayberry state matching analysis has been introduced in detail, and therefore, its repeated description will be omitted.

[0054] Example 1 This application adopts the above-mentioned dynamic optimization method of preservation parameters based on bayberry state matching analysis to construct a bayberry preservation and storage AI library to manage the preservation of bayberries in the library. At the same time, a preservation control method based on fixed preservation parameters is used to construct a bayberry preservation and storage ordinary library for comparative experiments to measure and record the ethanol and acetaldehyde content of bayberries.

[0055] Among them, the determination method of ethanol and acetaldehyde content is as follows: Weigh 2 g of waxberry pulp powder fully ground with liquid nitrogen into a 10 mL centrifuge tube, add 2 mL of saturated NaCl solution, and vortex to mix. Pipette 3 mL of homogenate into a headspace extraction bottle, add 10 μL of 2% sec-butanol, and mix. After heating in a 60°C water bath, the detection operation is the same as the standard sample. Perform 3 technical replicates for each sample.

[0056] like Figure 6 As shown, the above-mentioned determination method was used to measure the bayberry after it was stored for 20 days and 38 days, respectively. The experimental results showed that with the extension of storage time, the ethanol content increased significantly, and the ethanol and acetaldehyde contents in the AI ​​library and the ordinary library were significantly different. It can be seen that the bayberry intelligent preservation control method described in the present application can effectively slow down the metabolic activity of bayberry during storage, thereby extending its shelf life.

[0057] like Figure 7 As shown, after 31 days of storage, the bayberry fruits in the AI ​​warehouse still have good flavor and are still commercially available, while the bayberry fruits in the ordinary warehouse are seriously moldy and have lost their commercial value.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for dynamic optimization of preservation parameters based on bayberry state matching analysis, characterized in that: include: Construct a priori database of prototype images of bayberry states and recommended preservation parameters; Obtain the real-time status image of the bayberry in the warehouse collected by the camera; Extracting the bayberry state features from the bayberry real-time state image in the warehouse and the bayberry state prototype image-each bayberry state prototype image in the prior database of recommended preservation parameters to obtain a set of the bayberry real-time state semantic coding feature map and the bayberry state prototype semantic coding feature map; Performing spatial enhancement processing on the real-time state semantic coding feature map of the bayberry in the warehouse to obtain the real-time state semantic enhancement coding feature map of the bayberry in the warehouse; Determine a bayberry preservation control strategy based on the semantic similarity between the bayberry real-time state semantic enhancement coding feature graph in the warehouse and each bayberry state prototype semantic coding feature graph in the set of bayberry state prototype semantic coding feature graphs; Among them, the spatial enhancement processing is performed on the semantic coding feature map of the real-time state of bayberry in the warehouse, including: calculating the pixel-level semantic association of the semantic coding feature map of the real-time state of bayberry in the warehouse to obtain the pixel-granularity semantic association topological matrix of the real-time state of bayberry in the warehouse; based on the pixel spatial position information of the semantic coding feature map of the real-time state of bayberry in the warehouse, the spatial attention attenuation modulation is performed on the pixel-granularity semantic association topological matrix of the real-time state of bayberry in the warehouse to obtain the bayberry state semantic association space soft constraint topological feature matrix; based on the bayberry state semantic association space soft constraint topological feature matrix, the pixel-granularity spatial enhancement processing is performed on the semantic coding feature map of the real-time state of bayberry in the warehouse to obtain the semantic enhanced coding feature map of the real-time state of bayberry in the warehouse.

2. The method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to claim 1, characterized in that: The bayberry real-time state image in the warehouse and the bayberry state prototype image-each bayberry state prototype image in the prior database of recommended preservation parameters are respectively subjected to bayberry state feature extraction to obtain a set of the bayberry real-time state semantic coding feature map in the warehouse and the bayberry state prototype semantic coding feature map, including: The real-time state image of the bayberry in the warehouse and the prototype image of the bayberry state - each prototype image of the bayberry state in the prior database of recommended preservation parameters are respectively input into the bayberry state feature extractor based on the deep separable convolutional neural network model to obtain a set of the semantic coding feature map of the real-time state of the bayberry in the warehouse and the semantic coding feature map of the prototype state of the bayberry.

3. The method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to claim 2, characterized in that: Calculating the pixel-level semantic association of the real-time state semantic coding feature map of the bayberry in the warehouse to obtain the pixel-level semantic association topology matrix of the real-time state of the bayberry in the warehouse, including: The feature dispersion of the semantic coding feature map of the real-time state of the bayberry in the warehouse is performed along the channel dimension to obtain a set of pixel granularity feature vectors of the real-time state of the bayberry in the warehouse; The semantic association score between any two pixel granularity feature vectors of the real-time state of bayberry in the warehouse in the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse is calculated to obtain the pixel granularity semantic association topological matrix of the real-time state of bayberry in the warehouse.

4. The method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to claim 3, characterized in that: Calculating the semantic association score between any two of the real-time state pixel granularity feature vectors of the bayberry in the warehouse in the set of the real-time state pixel granularity feature vectors of the bayberry in the warehouse to obtain the real-time state pixel granularity semantic association topology matrix of the bayberry in the warehouse, including: Performing nonlinear mapping on each of the set of the real-time state pixel granularity feature vectors of the bayberry in the warehouse to obtain a set of mapped real-time state pixel granularity feature vectors of the bayberry in the warehouse; The Poincare distance between any two pixel granularity feature vectors of the real-time state of bayberry in the mapped warehouse in the set of pixel granularity feature vectors of the real-time state of bayberry in the mapped warehouse is calculated as the semantic association score to obtain the pixel granularity semantic association topological matrix of the real-time state of bayberry in the warehouse.

5. The method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to claim 4, characterized in that: Based on the pixel spatial position information of the real-time state semantic coding feature map of the bayberry in the warehouse, the real-time state pixel granularity semantic association topological matrix of the bayberry in the warehouse is subjected to spatial attention attenuation modulation to obtain the bayberry state semantic association space soft constraint topological feature matrix, including: Based on the spatial distance between any two pixel granularity feature vectors of the real-time state of bayberry in the warehouse in the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse, the pixel granularity semantic association topological matrix of the real-time state of bayberry in the warehouse is subjected to spatial attention attenuation modulation to obtain the soft constraint topological matrix of the semantic association space of the bayberry state; The soft constraint topological matrix of the bayberry state semantic association space is subjected to dilated convolution coding to obtain the soft constraint topological feature matrix of the bayberry state semantic association space.

6. The method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to claim 5, characterized in that: Based on the spatial distance between any two pixel granularity feature vectors of the real-time state of bayberry in the warehouse in the set of pixel granularity feature vectors of the real-time state of bayberry in the warehouse, the spatial attention attenuation modulation is performed on the semantic association topological matrix of the real-time state of bayberry in the warehouse to obtain the soft constraint topological matrix of the semantic association space of the bayberry state, including: The natural exponential function value of the Euclidean distance between any two of the real-time state pixel granularity feature vectors of the bayberry in the warehouse in the set of the real-time state pixel granularity feature vectors of the bayberry in the warehouse is used as a spatial distance factor to obtain a spatial distance topological matrix of the real-time state pixel granularity of the bayberry in the warehouse; The preset attenuation modulation constant is multiplied by the spatial distance topological matrix of the real-time pixel granularity state of the bayberry in the warehouse to obtain the spatial attention attenuation modulation matrix, and the pixel granularity semantic association topological matrix of the real-time pixel granularity state of the bayberry in the warehouse and the spatial attention attenuation modulation matrix are calculated and divided by the position points to obtain the bayberry state semantic association space soft constraint topological matrix.

7. The method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to claim 6, characterized in that: Based on the soft constraint topological feature matrix of the bayberry state semantic association space, pixel granularity spatial enhancement processing is performed on the bayberry real-time state semantic coding feature map in the warehouse to obtain the bayberry real-time state semantic enhancement coding feature map in the warehouse, including: Input the set of pixel granularity feature vectors of the real-time state of bayberry in the library and the bayberry state semantic association space soft constraint topological feature matrix into the graph convolutional coding module to obtain a set of shallow pixel granularity feature vectors of the bayberry state enhanced by context semantic association; The set of shallow pixel granularity feature vectors of the state of the waxberry enhanced by the contextual semantic association is reshaped to obtain a semantically enhanced coding feature map of the real-time state of the waxberry in the library.

8. The method for dynamic optimization of preservation parameters based on bayberry state matching analysis according to claim 7, characterized in that: Based on the semantic similarity between the real-time state semantic enhancement coding feature graph of bayberry in the warehouse and each bayberry state prototype semantic coding feature graph in the set of bayberry state prototype semantic coding feature graphs, a bayberry preservation control strategy is determined, including: Inputting each of the bayberry state prototype semantic coding feature graphs in the set of the bayberry real-time state semantic enhancement coding feature graph in the library and the bayberry state prototype semantic coding feature graph into the bayberry state semantic comparison network based on the hash function to obtain a set of bayberry state semantic comparison coefficients; Extracting the recommended preservation parameter corresponding to the minimum value of the bayberry state semantic contrast coefficient from the set of bayberry state semantic contrast coefficients; Based on the comparison between the recommended preservation parameters and the real-time environmental parameters in the warehouse, the waxberry preservation control strategy is determined.

9. A dynamic optimization system for preservation parameters based on bayberry state matching analysis, characterized in that: include: A priori database construction module is used to construct a priori database of prototype images of bayberry states and recommended preservation parameters; The real-time status monitoring module of bayberry is used to obtain the real-time status image of bayberry in the warehouse collected by the camera; A bayberry state feature extraction module is used to extract bayberry state features from the bayberry real-time state image in the warehouse and the bayberry state prototype image-each bayberry state prototype image in the prior database of recommended preservation parameters to obtain a set of the bayberry real-time state semantic coding feature map and the bayberry state prototype semantic coding feature map; A feature enhancement module is used to perform spatial enhancement processing on the real-time state semantic coding feature map of the bayberry in the warehouse to obtain a real-time state semantic enhancement coding feature map of the bayberry in the warehouse; The comparison and matching module is used to determine the bayberry preservation control strategy based on the semantic similarity between the bayberry real-time state semantic enhancement coding feature map in the library and the bayberry state prototype semantic coding feature map set.

10. A server comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images