CN-121746824-B - Tea fermentation degree identification method, system, equipment and storage medium

CN121746824BCN 121746824 BCN121746824 BCN 121746824BCN-121746824-B

Abstract

The application discloses a tea fermentation degree identification method, a system, equipment and a storage medium, wherein a tea fermentation image to be identified is input into a tea fermentation identification network, feature extraction is carried out on the tea fermentation image to be identified through a first processing layer to obtain a first semantic feature image, and multi-scale processing and edge enhancement processing are carried out on the tea fermentation image to be identified through an edge analysis and enhancement module layer to obtain a detail feature image; and carrying out classification recognition on the second semantic feature map through a classification decision layer to obtain a fermentation degree recognition result of the tea fermentation image to be recognized, and realizing cooperative perception of local microscopic change and overall uniformity in the fermentation process by strengthening recognition of image edges and texture detail features and establishing a cross-regional global context dependency relationship so as to further improve the accuracy of tea fermentation degree recognition.

Inventors

TANG ZHE
QIN JINGRU
QI FANG
Zou Dengpeng

Assignees

中南大学
长沙湘丰智能装备股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260228

Claims (8)

1. The tea fermentation degree identification method is characterized by being applied to electronic equipment, wherein the electronic equipment is integrated with a tea fermentation identification network, the tea fermentation identification network comprises a first processing layer, an edge analysis and enhancement module layer, a second processing layer and a classification decision layer, and the method comprises the following steps: Inputting a tea fermentation image to be identified into the tea fermentation identification network, extracting features of the tea fermentation image to be identified through the first processing layer to obtain a first semantic feature image, and performing multi-scale processing and edge enhancement processing on the tea fermentation image to be identified through the edge analysis and enhancement module layer to obtain a detail feature image; according to the first semantic feature map and the detail feature map, performing global dependency modeling processing through the second processing layer to obtain a second semantic feature map; classifying and identifying the second semantic feature images through the classifying and decision layer to obtain fermentation degree identification results of the tea fermentation images to be identified; The edge analysis and enhancement module layer comprises an initial convolution unit, a multi-stage circulation enhancement unit and a fusion output unit, and the edge analysis and enhancement module layer is used for carrying out multi-scale processing and edge enhancement processing on the tea fermentation image to be identified to obtain a detail characteristic diagram, wherein the detail characteristic diagram comprises the following steps: performing convolution downsampling and channel expansion processing on the tea fermentation image to be identified through the initial convolution unit to obtain a first feature map; Performing multistage circulation processing on the first feature map through the multistage circulation enhancement unit to obtain a multi-scale fusion feature; Carrying out channel fusion and dimension reduction processing on the multi-scale fusion characteristics through the fusion output unit to obtain the detail characteristic diagram; The multi-stage circulation processing is performed on the first feature map through the multi-stage circulation enhancing unit to obtain multi-scale fusion features, including: Performing iterative operation on the image features of the first feature map through the multi-stage loop enhancement unit until an iteration stop condition is met to obtain a plurality of enhancement features, and performing channel splicing on the enhancement features and the image features of the first feature map to obtain the multi-scale fusion feature, wherein any one iterative operation comprises: carrying out average pooling operation on the image features of the first feature map to calculate the difference between the image features and the local average value of the image features so as to obtain edge response features; performing convolution and activation processing on the edge response characteristics to obtain self-adaptive edge enhancement weights; Weighting the edge response characteristic and the self-adaptive edge enhancement weight, and adding back the image characteristic in a residual mode to obtain the enhancement characteristic; the second processing layer comprises a feature fusion unit, a downsampling unit and a global context modeling unit.
2. The method for identifying the fermentation degree of tea leaves according to claim 1, wherein the first processing layer comprises a first processing unit and a second processing unit, the feature extraction is performed on the tea fermentation image to be identified through the first processing layer, and a first semantic feature map is obtained, and the method comprises the following steps: performing downsampling treatment on the tea fermentation image to be identified through the first processing unit to obtain a low-level feature map of the tea fermentation image to be identified; Dividing the low-level feature map into a first sub-feature map and a second sub-feature map according to the dimension of the low-level feature map; Carrying out space feature extraction on the first sub-feature map through a first branch in the second processing unit to obtain a first sub-semantic feature map; Performing identity mapping on the second sub-feature map through a second branch in the second processing unit to obtain a second sub-semantic feature map; And splicing the first sub-semantic feature map and the second sub-semantic feature map to obtain the first semantic feature map.
3. The method for identifying the fermentation degree of tea leaves according to claim 1, wherein the step of performing global dependency modeling processing through the second processing layer according to the first semantic feature map and the detail feature map to obtain a second semantic feature map comprises: The feature fusion unit is used for splicing and fusing the first semantic feature image and the detail feature image in a channel dimension to obtain a fusion feature image; performing space downsampling and channel expansion processing on the fusion feature map through the downsampling unit to obtain a second feature map; And carrying out global context modeling on the second feature map through the global context modeling unit to obtain the second semantic feature map.
4. A method for identifying a degree of fermentation of tea leaves according to claim 3, wherein said performing global context modeling on said second feature map by said global context modeling unit to obtain said second semantic feature map comprises: performing depth separable convolution and weighted fusion on the second feature map to obtain a first intermediate feature map; Global information enhancement is carried out on the first intermediate feature map, and a second intermediate feature map is obtained; Fusing the first intermediate feature map and the second intermediate feature map to obtain a third intermediate feature map; based on a preset activation function, performing convolution processing and residual fusion on the third intermediate feature map for multiple times to obtain a modeling feature map; Carrying out feature refining on the modeling feature map to obtain a refined feature map; and obtaining the second semantic feature map according to the refined feature map.
5. The method for identifying the fermentation degree of tea leaves according to claim 1, wherein the step of performing classification and identification on the second semantic feature map by the classification decision layer to obtain a fermentation degree identification result of the tea fermentation image to be identified comprises the steps of: Performing global average pooling operation on the second semantic feature map to map the two-dimensional feature of each channel in the second semantic feature map into a scalar value so as to obtain a one-dimensional feature vector; mapping the one-dimensional feature vector into classification score vectors with the quantity corresponding to the preset fermentation degree class; Carrying out normalized exponential function processing on the classification score vector to obtain the prediction probability of each preset fermentation degree category; and determining the fermentation degree category corresponding to the maximum value of the prediction probability in the prediction probabilities of the preset fermentation degree categories as a fermentation degree recognition result of the tea fermentation image to be recognized.
6. The tea fermentation degree identification system is characterized by being applied to electronic equipment, wherein the electronic equipment is integrated with a tea fermentation identification network, the tea fermentation identification network comprises a first processing layer, an edge analysis and enhancement module layer, a second processing layer and a classification decision layer, and the system comprises: The first module is used for inputting the tea fermentation image to be identified into the tea fermentation identification network, extracting the characteristics of the tea fermentation image to be identified through the first processing layer to obtain a first semantic characteristic image, and carrying out multi-scale processing and edge enhancement processing on the tea fermentation image to be identified through the edge analysis and enhancement module layer to obtain a detail characteristic image; the second module is used for carrying out global dependency modeling processing through the second processing layer according to the first semantic feature map and the detail feature map to obtain a second semantic feature map; the third module is used for carrying out classification and identification on the second semantic feature map through the classification decision layer to obtain a fermentation degree identification result of the tea fermentation image to be identified; The edge analysis and enhancement module layer comprises an initial convolution unit, a multi-stage circulation enhancement unit and a fusion output unit, and the edge analysis and enhancement module layer is used for carrying out multi-scale processing and edge enhancement processing on the tea fermentation image to be identified to obtain a detail characteristic diagram, wherein the detail characteristic diagram comprises the following steps: performing convolution downsampling and channel expansion processing on the tea fermentation image to be identified through the initial convolution unit to obtain a first feature map; Performing multistage circulation processing on the first feature map through the multistage circulation enhancement unit to obtain a multi-scale fusion feature; Carrying out channel fusion and dimension reduction processing on the multi-scale fusion characteristics through the fusion output unit to obtain the detail characteristic diagram; The multi-stage circulation processing is performed on the first feature map through the multi-stage circulation enhancing unit to obtain multi-scale fusion features, including: Performing iterative operation on the image features of the first feature map through the multi-stage loop enhancement unit until an iteration stop condition is met to obtain a plurality of enhancement features, and performing channel splicing on the enhancement features and the image features of the first feature map to obtain the multi-scale fusion feature, wherein any one iterative operation comprises: carrying out average pooling operation on the image features of the first feature map to calculate the difference between the image features and the local average value of the image features so as to obtain edge response features; performing convolution and activation processing on the edge response characteristics to obtain self-adaptive edge enhancement weights; Weighting the edge response characteristic and the self-adaptive edge enhancement weight, and adding back the image characteristic in a residual mode to obtain the enhancement characteristic; the second processing layer comprises a feature fusion unit, a downsampling unit and a global context modeling unit.
7. A tea fermentation degree identification device comprising at least one control processor and a memory for communication connection with the at least one control processor, the memory storing instructions executable by the at least one control processor to enable the at least one control processor to perform a tea fermentation degree identification method as claimed in any one of claims 1to 5.
8. A computer-readable storage medium storing computer-executable instructions for causing a computer to perform a tea fermentation degree identification method according to any one of claims 1 to 5.

Description

Tea fermentation degree identification method, system, equipment and storage medium Technical Field The application relates to the technical field of tea fermentation identification, in particular to a tea fermentation degree identification method, a system, equipment and a storage medium. Background The tea fermentation is a core process for determining the quality and flavor of tea, the unique characteristics, the aroma and the taste of various fermented tea such as black tea, oolong tea and the like are formed in a fermentation stage, and polyphenols contained in the tea can be dynamically converted into key components such as theaflavin, thearubigin and the like during the fermentation stage, so that the final value of the product is directly determined. Therefore, the tea fermentation degree is accurately and real-time identified, the intelligent and standardized tea processing is realized, the product quality consistency is ensured, the key technical support for meeting the requirement of large-scale continuous production is met, and the intelligent agricultural transformation method has important significance for promoting the traditional tea processing industry to realize intelligent agricultural transformation. However, the current tea fermentation degree identification method is dependent on manual feature design in vision, has limited feature expression capability, is difficult to capture complex nonlinear color and texture evolution in the fermentation process, has weak generalization capability, is insensitive to capturing of fine continuous fermentation feature changes, is difficult to consider local detail and global context, and further causes insufficient accuracy of tea fermentation degree identification. Disclosure of Invention The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims. The main purpose of the disclosed embodiments is to provide a tea fermentation degree identification method, a system, a device and a storage medium, which can realize cooperative sensing of local microscopic change and overall uniformity in a fermentation process by strengthening and identifying image edges and texture detail characteristics and establishing a cross-regional global context dependency relationship, so as to improve the tea fermentation degree identification accuracy. The first aspect of the embodiment of the application provides a tea fermentation degree identification method, which is applied to electronic equipment, wherein the electronic equipment is integrated with a tea fermentation identification network, the tea fermentation identification network comprises a first processing layer, an edge analysis and enhancement module layer, a second processing layer and a classification decision layer, and the method comprises the following steps: Inputting a tea fermentation image to be identified into the tea fermentation identification network, extracting features of the tea fermentation image to be identified through the first processing layer to obtain a first semantic feature image, and performing multi-scale processing and edge enhancement processing on the tea fermentation image to be identified through the edge analysis and enhancement module layer to obtain a detail feature image; according to the first semantic feature map and the detail feature map, performing global dependency modeling processing through the second processing layer to obtain a second semantic feature map; and carrying out classification and identification on the second semantic feature images through the classification decision layer to obtain fermentation degree identification results of the tea fermentation images to be identified. In some embodiments of the present application, the first processing layer includes a first processing unit and a second processing unit, and the feature extraction is performed on the tea fermentation image to be identified by the first processing layer to obtain a first semantic feature map, including: performing downsampling treatment on the tea fermentation image to be identified through the first processing unit to obtain a low-level feature map of the tea fermentation image to be identified; Dividing the low-level feature map into a first sub-feature map and a second sub-feature map according to the dimension of the low-level feature map; Carrying out space feature extraction on the first sub-feature map through a first branch in the second processing unit to obtain a first sub-semantic feature map; Performing identity mapping on the second sub-feature map through a second branch in the second processing unit to obtain a second sub-semantic feature map; And splicing the first sub-semantic feature map and the second sub-semantic feature map to obtain the first semantic feature map. In some embodiments of the present application, the edge analysis and enhancement module layer includes an initial convolu