Tea cake fermentation change degree identification method, quality identification method and storage medium

Abstract

The application discloses a method for training a neural network to identify the fermentation change degree of tea cakes, a fermentation change degree identification method, a quality identification method and a storage medium. The training method comprises the following steps: A. performing the following sample acquisition step multiple times to obtain multiple sets of learning samples, wherein the sample acquisition step comprises: A1. obtaining the image of tea cakes before storage, the image of tea cakes after storage fermentation, and the fermentation change degree of tea cakes during storage; A2. taking the image of tea cakes before storage and the image of tea cakes after storage fermentation as input signals, and taking the fermentation change degree of tea cakes during storage as an output signal to form a set of learning samples; B. using the multiple sets of learning samples to train an artificial neural network to identify the fermentation change degree of tea cakes until the artificial neural network has the identification ability to identify the fermentation change degree of tea cakes, so that intelligent identification is realized, the identification accuracy is high, manual identification is not needed, and the labor cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tea cake fermentation degree recognition, and in particular to a tea cake fermentation change degree recognition method, a quality recognition method and a storage medium. BACKGROUND

[0002] Pu'er tea is a special tea that can be post-fermented under the joint action of moisture and microorganisms, using Yunnan's special large-leaf sun-dried green tea as raw material. During storage, the taste and aroma of tea leaves will change with the increase of storage time, and the taste will become better and better. The storage process of Pu'er tea generally includes aging and tea-awakening stages. Pu'er tea is placed in an aging storage warehouse for aging, and after reaching the aging degree, it is transferred to a tea-awakening storage warehouse for tea-awakening. After tea-awakening, it can be sold. When transferring, it is usually necessary to determine whether the post-fermentation degree of Pu'er tea reaches the aging standard, and only when it reaches the standard can it be transferred to the tea-awakening storage warehouse for tea-awakening. When it is ready for transfer, it also needs to determine the post-fermentation degree of Pu'er tea so as to classify and price it. The current determination method is mainly manual determination, but the number of experienced personnel is small and their training period is long, and the labor cost is high.

[0003] At present, there is a scheme for intelligent recognition of fermentation degree in the semi-fermentation process of tea leaves using machine learning. Specifically, a plurality of sample images of different fermentation degrees are collected, a single sample image is used as an input signal, and the corresponding fermentation degree is used as an output signal to form a set of learning samples. A plurality of sets of learning samples are used to train a neural network, so that the neural network can recognize the fermentation degree of tea leaves according to a single tea leaf image. However, this method is only suitable for fermentation degree recognition in the semi-fermentation process where the color of tea leaves changes obviously. In the post-fermentation process, the fermentation speed of tea leaves is very slow, and the color change of tea leaves is not obvious. If the fermentation degree of tea leaves is simply identified by a single tea leaf image, it is easy to produce misjudgment. Moreover, the color of tea leaves when Pu'er tea reaches the fermentation degree of aging standard is related to the color of tea leaves before storage. The color of tea leaves before storage is different, and the color of tea leaves when it reaches the fermentation degree of aging standard will also be slightly different. Therefore, it is difficult to accurately determine whether Pu'er tea reaches the aging standard by simply relying on a single tea leaf image. SUMMARY

[0004] The technical problem to be solved by the present application is how to intelligently recognize the fermentation change degree of tea cakes during the aging process.

[0005] The method for training an artificial neural network to recognize the fermentation change degree of tea cakes according to the present application comprises the following steps:

[0006] A. The sample acquisition step is executed multiple times to obtain multiple sets of learning samples, and the sample acquisition step comprises:

[0007] A1. obtaining the image of the tea cake before storage, the image of the tea cake after storage and fermentation, and the degree of fermentation change of the tea cake during storage;

[0008] A2. taking the image of the tea cake before storage and the image of the tea cake after storage and fermentation as input signals, and taking the degree of fermentation change of the tea cake during storage as an output signal, to form a set of learning samples for the artificial neural network to identify the degree of fermentation change of the tea cake;

[0009] B. training the artificial neural network to identify the degree of fermentation change of the tea cake using the above-mentioned multiple sets of learning samples, until the artificial neural network has the ability to identify the degree of fermentation change of the tea cake according to the image of the tea cake before storage and the image of the tea cake after storage and fermentation.

[0010] Optionally,

[0011] Among the part of the learning samples, the image of the tea cake after storage and fermentation is an image labeled with mold spots and / or an image labeled with the integrity of the tea cake strips;

[0012] In step A2, whether the tea cake is moldy and / or the integrity of the tea cake strips is also taken as an output signal;

[0013] In step B, in addition to the above-mentioned, until the artificial neural network has the ability to identify the degree of fermentation change of the tea cake according to the image of the tea cake before storage and the image of the tea cake after storage and fermentation, the artificial neural network also has the ability to identify whether the tea cake is moldy and / or the integrity of the tea cake strips according to the image of the tea cake before storage and the image of the tea cake after storage and fermentation.

[0014] The first method for identifying the degree of fermentation change of the tea cake, the image of the tea cake before storage and the image of the tea cake after storage and fermentation are obtained and input into the trained artificial neural network, and the degree of fermentation change of the tea cake is identified by the artificial neural network.

[0015] Optionally, the artificial neural network is trained by the method for training the artificial neural network to identify the degree of fermentation change of the tea cake as described above.

[0016] Optionally, in addition to the above-mentioned, the degree of fermentation change of the tea cake is identified by the artificial neural network, and whether the tea cake is moldy and / or the integrity of the tea cake strips is also identified by the artificial neural network.

[0017] The second method for identifying the degree of fermentation change of the tea cake, comprising the following steps:

[0018] X. finding the point of the tea cake color in the image of the tea cake before storage and the point of the tea cake color in the image of the tea cake after storage and fermentation in the pre-stored tea cake color card, and obtaining the trajectory between the two points;

[0019] Y. looking up the fermentation variation degree matching the track from a pre-established database as the fermentation variation degree of the tea cake.

[0020] The tea cake quality identification method comprises the following steps:

[0021] P. identifying the fermentation variation degree of the tea cake during storage according to the color difference of the tea cake in the image before entering the warehouse and the image after the tea cake is fermented during storage;

[0022] Q. judging whether the tea cake quality meets the standard according to the standard meeting condition, wherein the standard meeting condition comprises that the storage duration of the tea cake and the fermentation variation degree during storage meet a preset corresponding relationship.

[0023] Optionally, in step P:

[0024] Specifically, the fermentation variation degree of the tea cake is identified by using the tea cake fermentation variation degree identification method as described above.

[0025] Alternatively

[0026] Specifically, the fermentation variation degree of the tea cake is identified by using the tea cake fermentation variation degree identification method as described above, and whether the tea cake is mildewed and / or the tea cake strip integrity is identified. The standard meeting condition in step Q comprises that the tea cake is not mildewed and / or the tea cake strip integrity is good.

[0027] Optionally, in step P, the fermentation variation degree of the tea cake is identified by using the second tea cake fermentation variation degree identification method as described above.

[0028] A computer readable storage medium having a computer program stored thereon, wherein the computer program is executed to implement the method for identifying the fermentation variation degree of the tea cake by using the trained artificial neural network, or to implement the tea cake fermentation variation degree identification method, or to implement the tea cake quality identification method.

[0029] The artificial neural network trained by using the training method can identify the fermentation variation degree of the tea cake during storage according to the image before entering the warehouse and the image after the tea cake is fermented during storage, realize intelligent identification, and reduce labor cost without manual identification. The training sample used in the training method is the image before entering the warehouse and the image after the tea cake is fermented during storage as the input signal, and the artificial neural network trained can accurately identify the fermentation variation degree of the tea cake during storage. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 is a schematic diagram of the perspective structure of the tea leaf shooting device;

[0031] Figure 2 is a front view of the tea leaf shooting device;

[0032] Figure 3 It is along Figure 2 Sectional view of line AA in the middle;

[0033] Figure 4 It is along Figure 2 A cross-sectional view along the BB line.

[0034] Explanation of the attached labels: 1. Display stand; 2. Camera; 3. Color reference card; 4. Housing; 5. Ring light. Detailed Implementation

[0035] The present invention will be further described in detail below with reference to specific embodiments.

[0036] Example 1

[0037] To enhance the value of Pu-erh tea, it is often processed into tea cakes for storage and fermentation. During this process, over-fermentation or mold growth may occur. Therefore, manual assessment is necessary to determine whether the fermented tea cakes meet quality standards, thus selecting only those that do.

[0038] The following three factors are mainly considered when judging whether a tea cake meets the quality standards: the degree of fermentation, the integrity of the tea leaves, and whether it is moldy. The degree of fermentation, the integrity of the tea leaves, and the presence of mold can be intelligently identified through machine learning. However, artificial neural networks trained using single tea cake images as input signals are prone to misjudgments in identifying the degree of tea fermentation. The degree of fermentation change reflects the current degree of fermentation, so it can be determined whether the target degree of fermentation has been reached. The present invention identifies the degree of fermentation change during storage by comparing the color of the tea cake before and after storage and fermentation, and then determines whether the identified degree of fermentation change matches the target degree of fermentation change. The tea quality identification device of the present invention includes a controller and a computer-readable storage medium. The computer-readable storage medium stores an executable computer program, and the controller executes the computer program to realize the function of the tea quality identification device. This identification device includes, for example,... Figure 1 The tea-shooting device shown includes a platform 1 and a camera 2 for taking pictures of the tea cake placed on the platform 1 to obtain an image of the tea cake (see [link]). Figure 2 or Figure 3 The camera 2 communicates with the controller.

[0039] The recognition of the fermentation change degree of the tea cake relies on the color of the tea cake in the tea cake image, and the color of the tea cake in the tea cake image taken under different environments will have color difference, and the present application uses the color difference between the color of the tea cake in the image before the tea cake is stored and the color of the tea cake in the image after the tea cake is stored to recognize the fermentation change degree of the tea cake during storage, and the color difference of the two images may be superimposed to make the color difference of the tea cake far different from the actual color difference, which greatly affects the recognition accuracy of the fermentation change degree of the tea cake. In order to prevent the difference in the shooting environment from affecting the recognition accuracy of the fermentation change degree of the tea cake, as shown in Figure 1 or Figure 4 The color reference card 3 surrounds the tea leaf placement area. The tea leaf shooting device comprises a shell 4 surrounding the placement table 1, and the shell 4 is open at the front to expose the placement table 1. A ring-shaped light supplement lamp 5 is arranged on the inner top wall of the shell 4, and the ring-shaped light supplement lamp 5 is placed directly above the placement table 1, and the camera 2 is located in the center of the ring-shaped light supplement lamp 5. The shell 4 also serves as a light shield to block the external environmental light from shining on the placement table 1, thereby reducing the influence of the external environmental light on the color of the image obtained by shooting. When the staff is preparing to store the tea cake for fermentation, the tea cake is placed in the tea leaf placement area of the placement table 1, and then the ring-shaped light supplement lamp 5 is turned on for light supplement, and then the camera 2 is started to shoot the tea cake on the placement table 1 together with the color reference card 3, so that the image before the tea cake is stored is obtained, and the image has the color reference card 3. The camera 2 transmits the image before the tea cake is stored to the controller, and the controller calibrates the color of the image before the tea cake is stored according to the difference between the color value of the color reference card 3 in the image before the tea cake is stored and the standard color value of the color reference card 3, and then uploads the calibrated image before the tea cake is stored to the database. After the above tea cake is stored for a period of time for fermentation, the tea merchant takes the tea cake out of the tea warehouse and places it on the placement table 1, then turns on the ring-shaped light supplement lamp 5 for light supplement, and then starts the camera 2 to shoot the tea cake on the placement table 1 together with the color reference card 3, so that the image after the tea cake is stored for fermentation is obtained, and the image has the color reference card 3. The camera 2 transmits the image after the tea cake is stored for fermentation to the controller, and the controller calibrates the color of the image after the tea cake is stored for fermentation according to the difference between the color value of the color reference card 3 in the image after the tea cake is stored for fermentation and the standard color value of the color reference card 3, and then uploads the calibrated image after the tea cake is stored for fermentation to the database and stores it in association with the corresponding image before the tea cake is stored.

[0040] After a large number of images of tea cakes before storage and images of tea cakes after storage and fermentation are obtained, the staff takes the images of tea cakes before storage and the corresponding images of tea cakes after storage and fermentation from the database, and manually judges the fermentation change degree of tea cakes during storage, and judges whether there are mold spots in the images of tea cakes after storage and fermentation and the completeness of tea cake strips. If there are mold spots, manually mark the mold spots and the completeness of tea cake strips on the images of tea cakes after storage and fermentation; if there are no mold spots, manually mark the completeness of tea cake strips on the images of tea cakes after storage and fermentation. After marking, the staff stores the images of tea cakes before storage, the images of tea cakes after storage and fermentation, and the fermentation change degree of tea cakes during storage as a group of tea cake sample data in the database. The staff repeatedly performs the above steps to obtain a plurality of groups of tea cake sample data.

[0041] In order to enable the artificial neural network to have the recognition ability of recognizing the fermentation change degree of tea cakes, whether the tea cakes are moldy, and the completeness of tea cake strips, a plurality of groups of learning samples need to be used to train the artificial neural network. The controller performs the following learning sample obtaining steps to obtain a group of learning samples:

[0042] A1. A group of tea cake sample data is obtained from the database, that is, the images of tea cakes before storage, the images of tea cakes after storage and fermentation, and the fermentation change degree of tea cakes during storage are obtained;

[0043] A2. The images of tea cakes before storage and the images of tea cakes after storage and fermentation are used as input signals of the current group of learning samples. If the images of tea cakes after storage and fermentation are marked with mold spots, the fermentation change degree of tea cakes during storage, moldy, and the completeness of tea cake strips marked in the images are used as output signals of the current group of learning samples. If the images of tea cakes after storage and fermentation are not marked with mold spots, the fermentation change degree of tea cakes during storage, non-moldy, and the completeness of tea cake strips marked in the images are used as output signals of the current group of learning samples.

[0044] After the tea cakes are stored and fermented, the fermentation change degree, the completeness of tea cake strips, and the moldy condition of the tea cakes will have a plurality of conditions. The controller repeatedly performs the above learning sample obtaining steps to obtain learning samples under a plurality of conditions, and then uses these learning samples to train the artificial neural network until the artificial neural network has the recognition ability of recognizing the fermentation change degree of tea cakes, whether the tea cakes are moldy, and the completeness of tea cake strips according to the images of tea cakes before storage and the images of tea cakes after storage and fermentation.

[0045] After the artificial neural network is trained, the controller can use the artificial neural network to recognize the fermentation change degree of tea cakes after storage and fermentation, whether the tea cakes are moldy, and the completeness of tea cake strips. The recognition steps are as follows:

[0046] The image of the tea cake before storage and the image of the tea cake after storage and fermentation are obtained from the database and input into the artificial neural network, and the fermentation change degree of the tea cake, whether the tea cake is mildewed, and the tea cake strip completeness are identified by the artificial neural network.

[0047] The staff pre-sets the corresponding relationship between the storage time of the tea cake and the fermentation change degree during storage according to work experience and stores it in the database. After identifying the fermentation change degree of the tea cake during storage, whether the tea cake is mildewed, and the tea cake strip completeness by the artificial neural network, the controller can determine whether the quality of the tea cake meets the standard according to the above-mentioned information. Specifically, if it is identified that the tea cake is not mildewed and the tea cake strip completeness is good, the controller determines whether the storage time of the tea cake and the fermentation change degree during storage meet the pre-set corresponding relationship in the database. If yes, it is considered that the quality of the tea cake meets the standard, otherwise, it is considered that the quality of the tea cake does not meet the standard.

[0048] Embodiment Two

[0049] This embodiment is generally the same as embodiment one, and only the differences between the two embodiments will be described below, and the same parts will not be described again.

[0050] In this embodiment, the artificial neural network is not used to identify the fermentation change degree of the tea cake during storage, but other methods are used to identify the fermentation change degree, and the specific process is as follows:

[0051] The staff pre-sets the tea cake color card and stores it in the database. After color calibration of the image of the tea cake after storage and fermentation, the controller takes out the corresponding image of the tea cake before storage from the database, finds the color position of the tea cake before storage and the color position of the tea cake after storage and fermentation in the tea cake color card, obtains the trajectory between the two positions, and then stores the image of the tea cake before storage, the image of the tea cake after storage and fermentation, and the corresponding trajectory in association.

[0052] After a large number of images of the tea cake before storage and the tea cake after storage and fermentation are obtained, the staff takes out the image of the tea cake before storage and its corresponding image of the tea cake after storage and fermentation from the database, and manually judges the fermentation change degree of the tea cake during storage, and then stores the fermentation change degree and the corresponding trajectory as a group of tea cake sample data in the database. The staff performs the above-mentioned steps multiple times to obtain many groups of tea cake sample data. The staff also judges whether the tea cake has mildew spots and the tea cake strip completeness according to the image of the tea cake after storage and fermentation. If there are mildew spots, the staff marks the mildew spots and the tea cake strip completeness on the image of the tea cake after storage and fermentation; if there are no mildew spots, the staff marks the tea cake strip completeness on the image of the tea cake after storage and fermentation.

[0053] After obtaining a plurality of sets of tea cake sample data, the following method can be used to identify the fermentation change degree of the tea cake during storage, and the identification steps are as follows:

[0054] X. Find the color point of the tea cake in the image of the tea cake before storage and the color point of the tea cake in the image of the tea cake after fermentation in storage in the pre-stored tea cake color card, and obtain the trajectory between the two points.

[0055] Y. Find the trajectory with the highest similarity from the tea cake sample data in the pre-established database, and use the fermentation change degree in the tea cake sample data as the fermentation change degree of the tea cake to be identified.

[0056] The artificial neural network of the present embodiment only needs to have the recognition ability of identifying whether the tea cake is mildewed and the tea cake strip integrity after the tea cake is fermented in storage, therefore, the controller only needs to execute the following learning sample obtaining steps to obtain a set of learning samples:

[0057] Obtain the image of the tea cake after fermentation in storage from the database;

[0058] Use the image of the tea cake after fermentation in storage as the input signal of the learning sample, if the image of the tea cake after fermentation in storage is marked with mildew points, use the two of the tea cake being mildewed and the tea cake strip integrity marked in the image as the output signal of the learning sample; if the image of the tea cake after fermentation in storage is not marked with mildew points, use the two of the tea cake not being mildewed and the tea cake strip integrity marked in the image as the output signal of the learning sample.

[0059] After the tea cake is fermented in storage, there are various situations of the tea cake strip integrity and mildew, the controller executes the above learning sample obtaining steps multiple times to obtain learning samples under various situations, and then uses these learning samples to train the artificial neural network until the artificial neural network has the recognition ability of identifying whether the tea cake is mildewed and the tea cake strip integrity according to the image of the tea cake after fermentation in storage.

[0060] After the artificial neural network is trained, the controller can use the artificial neural network to identify whether the tea cake is mildewed and the tea cake strip integrity after the tea cake is fermented in storage, and the identification steps are as follows:

[0061] Obtain the image of the tea cake to be identified after fermentation in storage from the database, input it into the above artificial neural network, and identify whether the tea cake is mildewed and the tea cake strip integrity by the artificial neural network.

[0062] At this point, the fermentation change degree of the tea cake during storage, whether the tea cake is mildewed and the tea cake strip integrity have been identified, and the controller can determine whether the tea cake quality meets the standard.

[0063] The above description is only an embodiment of the present application, and does not limit the patent protection scope. Any non-essential changes or substitutions made by those skilled in the art based on the present application still fall within the patent protection scope.

Claims

1. A method for identifying the fermentation variation degree of tea cakes, characterized in that The method comprises the following steps: X. Find the color position of the tea cake in the image before the tea cake is stored and the color position of the tea cake in the image after the tea cake is stored and fermented in the pre-stored tea cake color card, and obtain the track between the two positions; Y. Find the fermentation change degree matched with the track from the pre-established database as the fermentation change degree of the tea cake.

2. A method for identifying the quality of tea cakes, characterized by The method comprises the following steps: P. According to the difference of the tea cake color in the image before the tea cake is stored and the image after the tea cake is stored and fermented, the tea cake fermentation change degree recognition method of claim 1 is used to identify the fermentation change degree of the tea cake during storage; R. Input the image of the tea cake after storage and fermentation into the trained artificial neural network, and identify whether the tea cake is mildewed and the tea cake strip integrity by the artificial neural network; Q. Determine whether the tea cake quality meets the standard according to the standard meeting condition, and the standard meeting condition comprises: (1) The storage time of the tea cake and the fermentation change degree during storage meet the pre-set corresponding relationship; (2) It is identified that the tea cake is not mildewed and the tea cake strip integrity is good.

3. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed to realize the tea cake fermentation change degree recognition method of claim 1, or realize the tea cake quality recognition method of claim 2.

Images