CN116485418B - Tracing method and system for tea refining production - Google Patents
Tracing method and system for tea refining production Download PDFInfo
- Publication number
- CN116485418B CN116485418B CN202310735477.3A CN202310735477A CN116485418B CN 116485418 B CN116485418 B CN 116485418B CN 202310735477 A CN202310735477 A CN 202310735477A CN 116485418 B CN116485418 B CN 116485418B
- Authority
- CN
- China
- Prior art keywords
- detection data
- sse
- dimensional detection
- value
- tea
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 131
- 241001122767 Theaceae Species 0.000 title claims abstract description 129
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 48
- 238000007670 refining Methods 0.000 title claims abstract description 28
- 238000001514 detection method Methods 0.000 claims abstract description 241
- 230000008569 process Effects 0.000 claims abstract description 95
- 238000012216 screening Methods 0.000 claims abstract description 48
- 230000006870 function Effects 0.000 claims description 19
- 238000009825 accumulation Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 18
- 238000003064 k means clustering Methods 0.000 claims description 15
- 230000008859 change Effects 0.000 claims description 13
- 238000010586 diagram Methods 0.000 claims description 12
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 abstract description 18
- 230000000877 morphologic effect Effects 0.000 abstract description 2
- 238000012163 sequencing technique Methods 0.000 abstract 1
- 238000007726 management method Methods 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 238000005259 measurement Methods 0.000 description 6
- 238000011895 specific detection Methods 0.000 description 5
- 238000012847 principal component analysis method Methods 0.000 description 4
- 238000000855 fermentation Methods 0.000 description 3
- 230000004151 fermentation Effects 0.000 description 3
- 238000012512 characterization method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000011946 reduction process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007664 blowing Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000796 flavoring agent Substances 0.000 description 1
- 235000019634 flavors Nutrition 0.000 description 1
- 235000011389 fruit/vegetable juice Nutrition 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/018—Certifying business or products
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
- G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Health & Medical Sciences (AREA)
- Manufacturing & Machinery (AREA)
- Primary Health Care (AREA)
- Human Resources & Organizations (AREA)
- Finance (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Entrepreneurship & Innovation (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of product production traceability management, in particular to a tea refining production traceability method and system. The method comprises the steps of obtaining detection values of each tea under a certain detection index, forming one-dimensional detection data under the detection index, calculating variance contribution rate of the one-dimensional detection data, sequencing the variance contribution rate of the one-dimensional detection data of the same process flow from large to small, accumulating the variance contribution rate in the process flow sequence and the sequence from large to small, determining preliminary screening one-dimensional detection data according to accumulated values, selecting different one-dimensional detection data from the preliminary screening one-dimensional detection data to form different combinations, comparing the morphological differences of SSE line patterns obtained by each combination to determine optimal combinations as multi-dimensional data after optimal dimension reduction, and clustering the optimal combinations to complete tracing classification management of tea products. The invention realizes more optimal selection of multidimensional data after dimension reduction, and can obviously improve the reasonable and accurate traceability classification management of tea products.
Description
Technical Field
The invention relates to the technical field of product production traceability management, in particular to a tea refining production traceability method and system.
Background
The tea tracing refers to the whole process tracing and management of links such as tea production, processing, packaging, circulation and the like through technical means, so that the quality, safety and authenticity of the tea are ensured, and the trust and satisfaction of consumers are improved. The tea tracing system is characterized in that a unique identity code is assigned to each product by means of a one-object one-code technology, and then the object code correlation technology is applied to correlate production process information of tea with the background of the tracing system, so that the whole process and the whole path of tea production are visually presented to consumers, and the consumers can clearly know the information of the products, the production process, the production mode and the like of the tea.
The giving of the tracing codes has great influence on the credibility of commodities facing customers and improving the market prospect of enterprises, and because the tea production amount is large and the package is small, the tea with the tracing data is commonly used for tracing codes according to picking time periods, production batches, detection indexes, transfer batches and the like, namely a plurality of bags of tea produced by the same production batch are commonly used for tracing codes. All record data of the batch tea leaves with the tracing codes are average characterization values, but in the actual production process, even all indexes of the batch tea leaves in the production and processing process are necessarily different, so that the production and processing processes of all tea products in the batch are characterized by the same tracing code, the problem that the tracing classification management of the products is unreasonable and fine exists, and inaccurate tracing data and doubtful authenticity are caused.
Disclosure of Invention
The invention provides a tracing method and a tracing system for refined production of tea, which are used for solving the technical problems of unreasonable and accurate tracing classification management of the existing tea products, and the adopted technical scheme is as follows:
the invention relates to a tracing method for tea refining production, which comprises the following steps:
dividing the same batch of tea leaves into a plurality of parts, obtaining detection values of each part of tea leaves under each detection index in the refining processing process, and forming a sequence of detection values of a certain detection index in each part of tea leaves as one-dimensional detection data of the detection index, so as to obtain a plurality of one-dimensional detection data with the same number as the number of items of the detection index;
calculating variance contribution rate of each one-dimensional detection data, sorting the variance contribution rate of each one-dimensional detection data under the same process flow from large to small, and accumulating the variance contribution rates with the same sorting order under each process flow according to the sequence of the process flow; in the accumulating process, firstly accumulating the variance contribution rate of the front of the ordering order, and then accumulating the variance contribution rate of the rear of the ordering order; stopping accumulation when the accumulated value of the variance contribution rate is not smaller than a preset accumulation threshold value, and taking the one-dimensional detection data corresponding to each accumulated variance contribution rate as preliminary screening one-dimensional detection data;
Selecting one-dimensional detection data from the preliminary screening one-dimensional detection data, and determining various one-dimensional detection data combinations which are set in number and are not identical in one-dimensional detection data in the selection process, wherein each combination is used as a to-be-determined characteristic data set; the value range of the set number is from the preset number to the total number of one-dimensional detection data in the preliminary screening one-dimensional detection data; carrying out k-means clustering for a plurality of times from one cluster number to a preset cluster number on each undetermined characteristic data set, determining the square sum SSE value of the distances between the cluster center point and the data points under each cluster number, and forming an SSE line graph of the current undetermined characteristic data set based on the SSE value obtained by the corresponding cluster number; taking a undetermined characteristic data set corresponding to the SSE line graph with large elbow point protrusion degree, large elbow point early-appearance degree and small smoothness degree as an optimal characteristic data set;
and (3) re-carrying out k-means clustering on the multidimensional detection data in the optimal characteristic data set, and finishing tracing classification of the tea products according to the clustering result.
The beneficial effects of the invention are as follows:
the invention forms a sequence by the detection values of each tea under each detection index to form one-dimensional detection data, namely, the acquisition and construction of multi-dimensional detection data in the tea production process are completed, and the limitation of technological processes is added in the process of accumulating the variance contribution rate of each one-dimensional detection data after sorting from large to small, so that the one-dimensional detection data with larger variances selected after the multi-dimensional data is subjected to dimension reduction can be prevented from being concentrated in one or a plurality of technological processes, and the problem that the classification accuracy is lower due to the fact that the tea classification judgment is not comprehensive enough is avoided; and then, optimally selecting the dimension-reduced preliminary screening one-dimensional detection data among the undetermined characteristic data sets according to the morphological characteristics of the SSE line graph, so that the determination of the multidimensional data which is most reasonable and accurate in classification is realized, and the reasonable and accurate traceability classification management of the tea products is improved to the greatest extent.
Further, the method for acquiring the one-dimensional detection data combination comprises the following steps:
y one-dimensional detection data are selected from the preliminary screening one-dimensional detection data to obtainA plurality of undetermined characteristic data sets, wherein Y is the set number, and the value range of Y is +.>，/>For said preset number,/->And the total number of the one-dimensional detection data in the one-dimensional detection data is screened for preliminarily.
Further, the accumulated value of the variance contribution rate is:
wherein W represents the variance contribution ratio accumulated value of one-dimensional detection data, R represents the number of detection indexes under different process flows, namely the number of corresponding obtained one-dimensional detection data under different process flows, L represents the number of process flows,and the variance contribution rate of the r one-dimensional detection data after the variance contribution rates of all one-dimensional detection data in the z-th process flow are sequenced from large to small is represented.
Further, the method for determining the prominence degree of the elbow point, the early-stage degree of the elbow point and the smoothness degree of the SSE line graph comprises the following steps:
calculating a first ratio between the gradient difference degree of the adjacent line segments on two sides of the elbow point on the SSE line graph and the average value of the gradient difference degree of the adjacent line segments on two sides of each k value on the SSE line graph, wherein the greater the first ratio is, the greater the elbow point protrusion degree is;
Calculating a second ratio between a k value corresponding to the elbow point on the SSE line graph and the preset clustering quantity, wherein the smaller the second ratio is, the greater the early-reaching degree of the elbow point is;
and calculating the mean square error of the slopes of all the line segments on the SSE line graph, wherein the larger the mean square error of the slopes of all the line segments is, the smaller the smoothness of the SSE line graph is.
Further, the method for determining the optimal characteristic data set comprises the following steps:
establishing an optimal objective function with respect to the degree of elbow prominence, the degree of elbow presences, and the degree of smoothness, the optimal objective function being proportional to the degree of elbow prominence and the degree of elbow presences and inversely proportional to the degree of smoothness;
and determining the value of an optimal objective function of each undetermined characteristic data set, and taking the undetermined characteristic data set with the maximum value of the optimal objective function as the optimal characteristic data set.
Further, the optimal objective function is:
wherein ,representing the optimal objective function>Representing the slope of the line segment formed between the kth and the kth+1th SSE value on the resulting SSE line graph for the pending profile data set, +.>Representing the k-1 th and k-th SSE values on the resulting SSE line graph corresponding to the pending feature data setSlope of line segment formed between +. >And->The values of the kth SSE value and the (k+1) SSE value on the SSE line diagram respectively represent the values of the to-be-determined characteristic data set, the to-be-determined characteristic data set corresponds to the abscissa k value on the SSE line diagram, and the specific clustering quantity in the process of clustering the to-be-determined characteristic data set to obtain the SSE value corresponding to the abscissa k value is represented as k>Representation of +.>Is also represented by the maximum value of the undetermined characteristic data set corresponding to the change value of the slope of the line segments on both sides of the elbow point on the obtained SSE line graph, ">Representing the average value obtained by averaging the absolute values of the slope differences between any two adjacent line segments on the SSE line graph corresponding to the feature data set to be determined, < + >>K-values representing the correspondence of the undetermined characteristic data set to the elbow points on the resulting SSE line graph,/-, respectively>Representing the preset number of clusters, +.>Mean square error representing the slope of each line segment on the resulting SSE line graph for the pending feature data set,/>Representation of +.>Is a maximum value of (a).
Further, the method for re-clustering the multi-dimensional detection data in the optimal characteristic data set comprises the following steps:
and taking the k value of the elbow point on the SSE line graph corresponding to the optimal characteristic data set as the optimal clustering quantity, and carrying out k-means clustering on the optimal characteristic data set according to the optimal clustering quantity.
The invention also provides a tea refining production traceability system, which comprises a memory and a processor, wherein the processor is used for executing instructions stored in the memory to realize the tea refining production traceability method introduced above and achieve the same technical effects as the method.
Drawings
FIG. 1 is a flow chart of a tracing method for tea refining production of the invention;
FIG. 2 is a schematic diagram of the invention for dividing the number of tea leaves in the same batch on a conveyor belt;
FIG. 3 is a schematic diagram of the present invention in which one-dimensional detection data under each process flow is sorted from large to small according to variance contribution rate and then the variance contribution rates are accumulated;
fig. 4 is a SSE line graph of the present invention.
Detailed Description
The conception of the invention is as follows:
firstly dividing the same batch of tea leaves into a plurality of parts, then obtaining detection values of each part under each detection index, forming a sequence of the detection values of each part of tea leaves under the same detection index to serve as one-dimensional detection data corresponding to the detection index, and then calculating variance contribution rate of each one-dimensional detection data; when the original multidimensional detection data formed by all one-dimensional detection data is subjected to dimension reduction processing according to the principle component analysis idea, the limitation condition that the order of process flows is added instead of accumulation in a mode of increasing all variance contribution rates is avoided, the primary dimension reduction data obtained when the dimension reduction is performed according to the existing principle component analysis idea are concentrated in a certain process flow or a certain process flows, and then the dimension reduction processing is performed again on the data after the primary dimension reduction.
The following describes a tracing method and system for tea refining production in detail with reference to the accompanying drawings and examples.
Method embodiment:
the embodiment of the tracing method for the refined production of the tea leaves has the overall flow shown in figure 1, and the specific process is as follows:
dividing the same batch of tea leaves into a plurality of parts, acquiring detection values of each part of tea leaves under each detection index in the refining processing process, and taking a sequence formed by the detection values of certain detection indexes in each part of tea leaves as one-dimensional detection data of the detection indexes, thereby obtaining a plurality of one-dimensional detection data with the same number as the number of the detection indexes.
In this embodiment, the most central refining process production data in the process of tracing the tea product is taken as an example for illustration, and it is easy to understand that in other embodiments, data in the non-refining process production process in the tea production process may also be analyzed.
Tea refining processes include, but are not limited to: screening (screening tea into different grades), shearing (when finer strips are needed, cutting tea with a cutter), fermenting (controlling color, flavor and taste of tea by fermentation), extracting stems (separating tea substances separated partially and fully releasing juice), covering fire (drying once again when insufficient), and winnowing (blowing refined tea with wind to separate the broken pieces and fine pieces). The tea obtained after a series of refining processes is the refined tea which can be marketed, and each processing procedure comprises a plurality of tea detection indexes, such as detection indexes of the screening process including tea shape, size, thickness, thinness and thickness, and the detection indexes of the fermentation process including fermentation time, temperature, humidity, PH value, and the like.
It can be seen that each treatment process in the tea refining process comprises a plurality of detection indexes, namely multidimensional detection data, and it is easy to understand that even the same batch of tea leaves is produced in the tea production process, the processing conditions of all parts are slightly different, namely the same detection indexes of all parts of tea leaves are different in specific numerical values in the detection data. That is, even the same batch of tea leaves is actually subjected to processing production, and further classification and different tracing codes are required to be assigned to different types.
In order to improve the fineness of tracing classification management of tea products, further reduce the difference degree of each item of detection data of tea in the same tracing code, improve the rationality and fineness of tracing classification management of tea products, firstly divide the same batch of tea into a plurality of parts, and then acquire specific detection data under all detection indexes in the production process of each part of tea, thereby completing the fine classification of tea according to the analysis of all detection data and giving corresponding tracing codes according to the classification.
Specifically, as shown in fig. 2, in this embodiment, after finishing refining the same batch on the production line, the tea leaves to be packaged are uniformly spread on a conveyor belt, and then for the purpose of refined classification to the greatest extent, the number of parts of the tea leaves in the batch is divided according to the gram weight of each bag when the tea leaves are bagged, so that the gram weight of each part is the same as the gram weight when the tea leaves are bagged subsequently. Of course, it is easy to understand that, in this embodiment, in order to maximally implement the refined classification of tea, the gram weight of each divided tea is set to the gram weight of each bag of tea, and in other embodiments, the gram weight of each divided tea may be set to the sum of the gram weights of multiple bags of tea, and only the division of multiple tea is ensured, so that the refined classification of tea can be implemented as well compared with the existing tracing management method.
After division of the tea leaves is completed, sampling is carried out from each tea leaf, and samples are sent into an automatic detection flow, so that detection data of the tea leaves of the samples are obtained, and the detection data of each sample can be regarded as detection data of the tea leaves in the corresponding tea leaves. The detection data specifically comprises detection data under each detection index in different processing technologies, after all detection data of all tea leaves are obtained, the specific detection data, namely specific detection values, of the same detection index in all tea leaves are constructed into a one-dimensional sequence, one-dimensional detection data corresponding to the detection index can be obtained, all detection data of each tea leaf can be constructed into one-dimensional detection data divided according to the detection index, the number of the one-dimensional detection data is the same as the number of items of the detection index, namely, the number of the detection indexes of the tea leaves in the refining processing process is the same, and finally the number of the one-dimensional detection data can be obtained. In this embodiment, for convenience of analysis, the sorting of the specific detection values in each one-dimensional detection data is arranged according to the order of the number of copies divided by the tea leaves, and of course, in other embodiments, the sorting of the specific detection values may be performed in other manners.
Step two, calculating variance contribution rates of all one-dimensional detection data, sorting the variance contribution rates of all one-dimensional detection data under the same technological process from large to small, and accumulating the variance contribution rates with the same sorting order under all the technological process according to the order of the technological process; in the accumulating process, firstly accumulating the variance contribution rate of the front of the ordering order, and then accumulating the variance contribution rate of the rear of the ordering order; stopping accumulation when the accumulated value of the variance contribution rate is not smaller than a preset accumulation threshold value, and taking the one-dimensional detection data corresponding to each accumulated variance contribution rate as preliminary screening one-dimensional detection data.
The number of the one-dimensional detection data obtained in the first step is the same as the number of the detection indexes of the tea leaves in the refining process, and in the refining process, each processing technology comprises a plurality of detection indexes, and in the refining process, the refining process also comprises a plurality of processing technologies, so that the number of the final detection indexes is large, namely the number of the one-dimensional detection data is large, and each one-dimensional detection data represents one characteristic dimension of the state of the tea leaves, so that the whole tea leaf detection data formed by all the one-dimensional detection data is actually multi-dimensional detection data with high dimension.
In the process of tracing the refined production of the tea leaves, the tea leaves are subjected to clustering classification according to the whole detection data, tracing codes are given according to classification results, the basis for clustering to complete classification is the multi-dimensional detection data formed by all one-dimensional detection data, and when the classification is completed, the conventional multi-dimensional data k-means clustering algorithm firstly needs to perform data dimension reduction, namely, high-dimensional data keep some most important features, and noise and unimportant features are removed.
According to the principle of principal component analysis, when multidimensional data is subjected to dimension reduction, one or more one-dimensional detection data with larger variance in one-dimensional detection data set are generally selected as characteristic data, and a model is constructed to measure the difference or distance between the data so as to finish clustering classification. In a simple sense, the feature data is one or more one-dimensional detection data with larger differences between different tea samples, and the one-dimensional detection data with smaller differences between different tea samples is regarded as useless features and discarded.
The existing principal component analysis method is to select a plurality of dimension data to be reserved as characteristic data according to a variance contribution rate accumulation sum, wherein the variance contribution rate calculation method is to divide the variance of each one-dimensional detection data by the variance sum of all one-dimensional detection data, accumulate the variance contribution rate accumulation sum according to the order of the variance contribution rate from large to small, and generally select one-dimensional detection data accumulated when the variance contribution rate accumulation sum reaches 80% as one-dimensional detection data in a feature data set after dimension reduction.
The main component analysis selects one-dimensional detection data with larger internal variance as characteristic data, but when the tea refining process is carried out, the same batch of tea leaves have deviation on a production line due to a certain process, so that most of one-dimensional detection data detected by the process have larger differences, namely, most of one-dimensional detection data under the process have higher variance contribution rate, and therefore, when the characteristic data is selected according to the existing main component analysis method based on larger variance, the one-dimensional detection data in the feature data group after dimension reduction is concentrated in one or more process flows with high probability. The ratio or content of the detection data in the residual process flow is reduced to a great extent, so that the dimension of the multi-dimensional detection data is reduced directly by the existing principal component analysis method, the judgment of the similarity of the multi-dimensional data is obviously incomplete, namely, for tea tracing, the similarity calculation of the multi-dimensional data is carried out only according to the measurement data with larger internal variance, and the difference among sample data is certainly amplified, so that the selection method of the characteristic data in the dimension reduction process of the multi-dimensional detection data is required to be optimized. Therefore, we adjust the accumulation mode of the contribution rate of the contrast in the existing principal component analysis method:
According to the process flow of tea fine processing, when the former process flow has the difference of the tea produced in the same batch due to the deviation of parameters or processing technology, the follow-up process flow is necessarily influenced, so that the priority of detected data in the former process flow is higher, and the variance contribution rate is accumulated according to the process flow sequence. In this embodiment, the variance contribution rates of the one-dimensional detection data corresponding to the detection indexes under a certain process flow are ranked according to the mode from large to small, so that the ranking of the variance contribution rates of the one-dimensional detection data under all the process flows is completed. As shown in fig. 3, A1, A2 and A3 respectively represent variance contribution rates of the first, second and third one-dimensional detection data after sorting from large to small in the first process flow, B1, B2 and B3 respectively represent variance contribution rates of the first, second and third one-dimensional detection data after sorting from large to small in the second process flow, and C1, C2 and C3 respectively represent variance contribution rates of the first, second and third one-dimensional detection data after sorting from large to small in the third process flow. The variance contribution rate of the one-dimensional detection data of the remaining rows (i.e. under the remaining subsequent process flows) is the same in the ordering condition.
When the specific variance contribution rate accumulation is performed, in this embodiment, the variance contribution rate of the one-dimensional detection data ordered first in each process flow is sequentially accumulated according to the process flow sequence, then the variance contribution rate of the one-dimensional detection data ordered second in each process flow is sequentially accumulated according to the process flow sequence, then the variance contribution rate of the one-dimensional detection data ordered third in each process flow is sequentially accumulated according to the process flow sequence, and so on, each round is that the variance contribution rate of the one-dimensional detection data ordered one level lower in order in each process flow than the one-dimensional detection data ordered in the previous round is sequentially accumulated according to the process flow sequence. Referring to fig. 3 again, the variance contribution rates of the first left column in fig. 3 are sequentially accumulated in the order from top to bottom, and the variance contribution rates of the second left column are sequentially accumulated in the order from top to bottom after the accumulation is completed, so that the variance contribution rates are sequentially accumulated.
According to the above, the accumulated value of the variance contribution rate is calculated as:
wherein W represents the variance contribution ratio accumulated value of one-dimensional detection data, R represents the number of detection indexes under different process flows, namely the number of corresponding obtained one-dimensional detection data under different process flows, L represents the number of process flows, And the variance contribution rate of the r one-dimensional detection data after the variance contribution rates of all one-dimensional detection data in the z-th process flow are sequenced from large to small is represented.
When the sum W of the variance contribution rates of the one-dimensional detection data is larger than a set accumulation threshold value, stopping accumulation and counting all the one-dimensional detection data corresponding to the accumulated variance contribution rates, taking all the one-dimensional detection data obtained by counting as one-dimensional detection data after preliminary screening, forming a preliminary screening characteristic data set by the obtained preliminary screening one-dimensional detection data, and obviously, if the obtained preliminary screening one-dimensional detection data has N pieces, the preliminary screening characteristic data set comprises N pieces of one-dimensional detection data.
Regarding setting the accumulation threshold, the value of the accumulation threshold is preferably 80%, namely 0.8 in this embodiment, and in other embodiments, other values can be also taken according to specific requirements on the subdivision degree of the tea.
The specific accumulation mode of the variance contribution rate of the one-dimensional detection data in the embodiment completes the primary screening of the one-dimensional detection data, avoids the condition that the one-dimensional detection data with larger variances is concentrated in one or a few process flows, enables the similarity between the multi-dimensional detection data of the tea samples to be more comprehensive when the similarity is calculated, and can greatly reduce the calculated amount in the follow-up optimization calculation and reduce the time complexity of iterative calculation on the other hand through the primary screening of the one-dimensional detection data.
Selecting one-dimensional detection data from the preliminary screening one-dimensional detection data, and determining various one-dimensional detection data combinations which are set in number and are not identical in one-dimensional detection data in the selection process, wherein each combination is used as a to-be-determined characteristic data set; the value range of the set number is from the preset number to the total number of one-dimensional detection data in the preliminary screening one-dimensional detection data; carrying out k-means clustering for a plurality of times from one cluster number to a preset cluster number on each undetermined characteristic data set, determining the square sum SSE value of the distances between the cluster center point and the data points under each cluster number, and forming an SSE line graph of the current undetermined characteristic data set based on the SSE value obtained by the corresponding cluster number; and taking the undetermined characteristic data set corresponding to the SSE line graph with the prominent elbow point, early elbow point early appearance degree and small smoothness degree as the optimal characteristic data set.
After the primary screening one-dimensional detection data is obtained in the second step, the embodiment also considers that in order to avoid the unreasonable situation that a large number of classification results appear in the same batch of tea leaves due to the fact that the primary screening one-dimensional detection data contains too many one-dimensional detection data with larger internal variances, namely, in the head part with the largest variance contribution rate in the primary screening one-dimensional detection data, the tea leaves are excessively subdivided due to the fact that the variances of the one-dimensional detection data are too large, so that tracing classification is unreasonable, the embodiment continues to perform secondary screening on the primary screening one-dimensional detection data, and selects part of the one-dimensional detection data from the primary screening one-dimensional detection data as the characteristic data which is finally and practically used for clustering.
Before explaining the principle of further screening the preliminary screening one-dimensional detection data in this embodiment in detail, it is necessary to introduce the content of the prior art means based on which the principle is obtained by improvement.
The k-means clustering algorithm is known to be required to preset the k value of the clustering number, and the k value of the clustering number is determined in a common elbow method mode, namely, a better k value is obtained by iterating the k value and then evaluating the clustering results under different k values. However, in the data dimension reduction process, the distance measurement parameter constructed according to the characteristic data can cause the clustering result distortion when the k-means iterates the k value, the real elbow point is difficult to determine in the elbow graph, and the k value cannot be obtained.
The elbow method comprises the following steps:
the iterative k-values k-means clusters the data sets and calculates the Sum of Squares (SSE) of the distances between the center point of each cluster and the data points. SSE values corresponding to different values of the clustering number k are drawn into a line graph shown in figure 4. The curve shape in the line graph is observed to find the first position showing "elbow", which generally refers to the position where the rapid decrease trend of the SSE value is changed, that is, the position where the change trend of the SSE value is most obvious, or the position where the curve in the line graph starts to be stable, that is, the position corresponding to the clustering number k value of 4 in FIG. 4. After the elbow position is found, the number k of clusters at the elbow point can be used as the optimal number of clusters.
It is readily understood that the quality of the clustering result may affect the highlighting or blurring of the elbow points in the elbow map, i.e. the highlighting or blurring of the elbow points in the elbow map may be used to reflect the quality of the clustering result. The above-mentioned procedure of the existing elbow method is to determine the optimal k value by observing the change condition of the SSE image formed with the change of the k value, specifically by finding the most obvious inflection point in the SSE image formed with the change of the k value in the case of multi-dimensional data determination, that is, in the case that the number of data dimensions and the specific data content in each dimension are determined.
However, in the contrary, considering that the k-means multidimensional clustering is not just data feature degradation and k values, if we change the purpose, we do not need to determine the best k value through elbow points on SSE images for known multidimensional data, but select different numbers of one-dimensional detection data in the preliminary screening one-dimensional detection data, and after the one-dimensional detection data are formed into the pending feature data groups with different dimensions, the comparison evaluation is carried out on each SSE line graph formed by each pending feature data group in the k value iteration process, so that the quality of multidimensional clustering results according to each pending feature data group can be directly reflected, and then the optimal feature data group is determined from the pending feature data groups.
The specific process for completing secondary screening of the primary screening one-dimensional detection data by utilizing the principle is as follows:
firstly we set a preset number Y for selecting one-dimensional detection data from the preliminary screening one-dimensional detection data, and considering the accuracy of tea clustering subdivision to be ensured, the embodiment prefers the preset numberFor 5, it is of course possible in other embodiments to take the preset number by other values, such as by a value of 2, depending on the specific requirements for the accuracy of the tea cluster subdivision. After the preset number value is determined to be 5, 5 one-dimensional detection data can be selected from the N preliminary screening one-dimensional detection data to obtainNamely->A plurality of undetermined characteristic data sets, then at a preset numberAdding one successively on the basis of the quantity, and selecting 6 one-dimensional detection data from N pieces of preliminary screening one-dimensional detection data to obtain +.>Namely->The undetermined characteristic data set and 7 one-dimensional detection data are selected from N preliminary screening one-dimensional detection data to obtain +.>Namely->A set of pending characteristic data, and so on until +.>The set of undetermined characteristics, in other words, the one-dimensional test data is selected from the preliminary screening>The number Y of the one-dimensional detection data is selected from the preliminary screening one-dimensional detection data, and the value range of the number Y of the one-dimensional detection data is +. >。
Each obtained undetermined characteristic data set is subjected to clustering quantity with k value from 1 to presetAnd generating SSE images according to the SSE values obtained by corresponding to each k value after each clustering, so as to obtain SSE images corresponding to each to-be-determined characteristic data set. Wherein ∈ about the preset number of clusters>Since we finely classify tea leaves produced in the same batch, the expected k value is not too large in the case of mature production process, and thus according to the experiencePreferably preset number of clusters +.>It is easy to understand that in other embodiments the preset number of clusters +.>The specific value of (c) is set to other values.
Regarding the SSE image corresponding to each generated set of pending feature data:
1. if the SSE line graph changes gradually along with the change of the k value, the internal variance of each one-dimensional detection data in the undetermined characteristic data set is small, the difference limit between sample data is fuzzy, and the elbow point position cannot be determined.
2. If the SSE line graph has a relatively strong variation trend along with the variation of the k value, the variance of each one-dimensional detection data in the undetermined characteristic data set is large, and the distance measurement parameters among samples are relatively large, the k value of the clustering number of clusters is far greater than the preset clustering number The elbow point will appear but the k value at this time is obviously not the true cluster number.
3. If the SSE line graph is changed more severely in the initial stage along with the change of the k value, then gradually and gradually and obviously appears an elbow point, then each one-dimensional detection data in the undetermined characteristic data set at the moment is considered to be capable of effectively representing multi-dimensional data for clustering.
Constructing an optimal objective function according to the analysis content of the SSE line graph:
wherein ,representing the optimal objective function>Representing the slope of the line segment formed between the kth and the kth+1th SSE value on the resulting SSE line graph for the pending profile data set, +.>Representing the slope of the line segment formed between the kth-1 and kth SSE values on the resulting SSE line graph for the pending profile data set, +.>And->The values of the kth SSE value and the (k+1) SSE value on the SSE line diagram respectively represent the values of the to-be-determined characteristic data set, the to-be-determined characteristic data set corresponds to the abscissa k value on the SSE line diagram, and the specific clustering quantity in the process of clustering the to-be-determined characteristic data set to obtain the SSE value corresponding to the abscissa k value is represented as k>Representation of +.>Is also represented by the maximum value of the undetermined characteristic data set corresponding to the change value of the slope of the line segments on both sides of the elbow point on the obtained SSE line graph, " >Representing the average value obtained by averaging the absolute values of the slope differences between any two adjacent line segments on the SSE line graph corresponding to the feature data set to be determined, < + >>K-values representing the correspondence of the undetermined characteristic data set to the elbow points on the resulting SSE line graph,/-, respectively>Representing the preset number of clusters, +.>Representing waitingThe fixed characteristic data set corresponds to the mean square error of the slope of each line segment on the obtained SSE line graph, +.>Representation of +.>Is a maximum value of (a).
In the above-mentioned method, the step of,representing the change value of the slope of the adjacent broken line segments, thereby +.>Represents the average value of the slope change values of the respective adjacent line segments on the line graph, and at the same time because +.>Representing the change value of the slope of the line segment of the two-sided broken line at the elbow point on the line graph, so +.>The larger represents the more prominent the elbow point on the SSE line graph. Then adding penalty term in the optimal objective function, < ->Representing the corresponding k value at the elbow point, < >>Representing the number of preset clusters, i.e. the number of abscissa k values on the SSE line graph, and k values also representing the total number of clustering iterations during acquisition of the SSE line graph, then +.>That is, the position of the elbow point on the SSE line diagram in the total iteration number, that is, whether the elbow point is forward or backward on the SSE line diagram, because of the above, when the one-dimensional detection data is primarily screened to contain excessive one-dimensional detection data with larger internal variance, the optimal k value is determined according to the existing elbow method A large number of classification results will appear, and the position of the elbow point on the SSE line diagram will be obviously behind due to the large optimal k value, but the purpose of this embodiment is to prevent unreasonable condition of the large classification results caused by the excessive classification of tea leaves, so it is expected that in the to-be-determined characteristic data set formed by selecting a plurality of one-dimensional detection data from the preliminary screening one-dimensional detection data, the position of the elbow point on the SSE line diagram is located in front, becauseThe larger the representation the farther back the elbow point appears and thus the further away from the desire, so +.>As a penalty factor in the optimal objective function; in addition, the flatter the SSE line graph, the more fuzzy the boundary between each one-dimensional detection data representing the current pending feature data set, that is, the more fuzzy the boundary between the multi-dimensional detection data represented by the pending feature data set, the more effective the elbow points cannot be determined or the more likely the elbow points are present, so by calculating the mean square error of the slopes of adjacent line segments on the SSE line graph as another penalty factor, it is easy to understand that the flatter the line graph should be when the mean square error is smaller, the correspondingly greater the penalty should be, and thus the reciprocal of the mean square error of the slopes of adjacent line segments is taken >With existing penalty factorsMultiplying to form penalty term; meanwhile, in order to avoid unreasonable situation that 0 is taken out of the denominator on the penalty term due to the fact that the mean square error is 0, the denominator is supplemented with a constant 1 to obtain the final penalty term ∈ ->. When the elbow point in the penalty term appears more recently or the SSE line graph is flatter overall, then +.>The bigger the->The smaller will be.
Finally, by finding the maximum value of the optimal objective function, i.e. obtainingThe set of undetermined feature data corresponding to the convergence of the optimal objective function is the optimal feature data set, and each one-dimensional detection data in the set of undetermined feature data set is data obtained by performing secondary screening on the primary screening one-dimensional detection data, namely the optimal screening multi-dimensional detection data obtained after the multi-dimensional detection data formed by the initial one-dimensional detection data is subjected to dimension reduction in the embodiment.
And fourthly, re-carrying out k-means clustering on the multidimensional detection data in the optimal characteristic data set, and finishing tracing classification of the tea products according to the clustering result.
The data in the optimal characteristic data set obtained in the third step is actually the optimal multidimensional detection data after dimensionality reduction, at this time, the obtained multidimensional detection data after dimensionality reduction avoids the condition that the data is concentrated in one or more process flows of tea refining production, and also avoids the condition that the trace-source classification is unreasonable due to the fact that too many classification results are generated after clustering, and the characterization effect of the trace-source classification on the tea difference is optimal, so that k-means clustering is performed again based on the obtained optimal characteristic data set, and the most reasonable classification on tea products can be realized.
However, a basic multidimensional data difference metric, or distance metric model, was previously required to perform subsequent clustering calculations. And substituting one-dimensional detection data of all undetermined characteristic data sets into the Euclidean norm model, and outputting the result as a measurement model of the distance between the multidimensional data. The acquisition of the metric model is the prior art, so this embodiment is not repeated.
And after the measurement model of the distances between the multidimensional data is obtained, recalculating the distance measurement of the multidimensional detection data between different tea samples in the optimal characteristic data set, namely, recalculating k-means clustering on the optimal characteristic data set to obtain a clustering result. It is easy to understand that the clustering number when k-means clustering is performed again is the clustering number value of the elbow in the SSE line graph corresponding to the optimal characteristic data set.
According to the clustering result, the same tracing code of packaged tea leaves in the same cluster can be given after the tea leaves are packaged, the detection data recorded in the tracing code are the average value of all detection data of all tea leaf samples in the cluster, and the mean square error is used as an error fluctuation interval to record all the detection data together.
Filling quality detection data of corresponding items according to all production records disclosed to consumers, obtaining a tracing matrix after filling, converting the tracing matrix into tracing codes by using two-dimensional codes, bar codes and the like, and endowing the tracing codes with the corresponding packaged tea at each sample position according to the clustering result of the sample position.
Thus, the tea quality detection data recorded during the tracing of the production data according to the tracing code is more reliable, and the brand power of a production enterprise is improved.
System embodiment:
the embodiment provides a refined production traceability system of tea, which comprises a memory and a processor, wherein the memory and the processor are mutually communicated through an internal bus, and the processor can be a processing device such as a microprocessor MCU (micro control unit), a programmable logic device FPGA (field programmable gate array) and the like.
The processor may call the logic instructions in the memory to implement a tracing method for refined tea production, and in the method embodiment, the method is described in detail, so that details are not repeated here.
In summary, the invention performs dimension reduction screening on multi-dimensional detection data of all process flows of samples in the tea refining production process by utilizing the variance contribution rate screening idea of the principal component analysis method, wherein the accumulating mode of variance contribution rates during screening of characteristic data is provided according to the process flow sequence and the variance contribution rate sequence, so that the problem that the similarity calculation of the follow-up multi-dimensional data is not comprehensive due to the fact that detection data with larger variances during principal component analysis are concentrated in one or more process flows can be avoided. And then iterating one-dimensional detection data types in the initial characteristic data set to construct an objective function, obtaining a group of characteristic data with the most prominent elbow points, the earliest elbow points and the smallest smoothness of the SSE line graph in the SSE line graph, and taking the characteristic data as an optimal dimension reduction characteristic data set, wherein the optimal dimension reduction characteristic data set can enable k-means clustering results of the multidimensional data to be more accurate, the sample difference classification caused by process deviation in each process flow to be most comprehensive, and further, each type of tea clusters are distributed with the same tracing code, so that tracing information is more accurate, reasonable and reliable, and the sales quality of products is ensured.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the application and are intended to be included within the scope of the application.
Claims (5)
1. The tracing method for the refined production of the tea is characterized by comprising the following steps of:
dividing the same batch of tea leaves into a plurality of parts, obtaining detection values of each part of tea leaves under each detection index in the refining processing process, and forming a sequence of detection values of a certain detection index in each part of tea leaves as one-dimensional detection data of the detection index, so as to obtain a plurality of one-dimensional detection data with the same number as the number of items of the detection index;
calculating variance contribution rate of each one-dimensional detection data, sorting the variance contribution rate of each one-dimensional detection data under the same process flow from large to small, and accumulating the variance contribution rates with the same sorting order under each process flow according to the sequence of the process flow; in the accumulating process, firstly accumulating the variance contribution rate of the front of the ordering order, and then accumulating the variance contribution rate of the rear of the ordering order; stopping accumulation when the accumulated value of the variance contribution rate is not smaller than a preset accumulation threshold value, and taking the one-dimensional detection data corresponding to each accumulated variance contribution rate as preliminary screening one-dimensional detection data;
Selecting one-dimensional detection data from the preliminary screening one-dimensional detection data, and determining various one-dimensional detection data combinations which are set in number and are not identical in one-dimensional detection data in the selection process, wherein each combination is used as a to-be-determined characteristic data set; the value range of the set number is from the preset number to the total number of one-dimensional detection data in the preliminary screening one-dimensional detection data; carrying out k-means clustering for a plurality of times from one cluster number to a preset cluster number on each undetermined characteristic data set, determining the square sum SSE value of the distances between the cluster center point and the data points under each cluster number, and forming an SSE line graph of the current undetermined characteristic data set based on the SSE value obtained by the corresponding cluster number; taking a undetermined characteristic data set corresponding to the SSE line graph with large elbow point protrusion degree, large elbow point early-appearance degree and small smoothness degree as an optimal characteristic data set;
k-means clustering is carried out on the multidimensional detection data in the optimal characteristic data set again, and tracing classification of the tea products is completed according to the clustering result;
the method for determining the degree of elbow protrusion, the degree of elbow early-stage and the smoothness degree of the SSE line graph comprises the following steps:
Calculating a first ratio between the gradient difference degree of the adjacent line segments on two sides of the elbow point on the SSE line graph and the average value of the gradient difference degree of the adjacent line segments on two sides of each k value on the SSE line graph, wherein the greater the first ratio is, the greater the elbow point protrusion degree is;
calculating a second ratio between a k value corresponding to the elbow point on the SSE line graph and the preset clustering quantity, wherein the smaller the second ratio is, the greater the early-reaching degree of the elbow point is;
calculating the mean square error of the slopes of all the line segments on the SSE line graph, wherein the larger the mean square error of the slopes of all the line segments is, the smaller the smoothness of the SSE line graph is;
the method for determining the optimal characteristic data set comprises the following steps:
establishing an optimal objective function with respect to the degree of elbow prominence, the degree of elbow presences, and the degree of smoothness, the optimal objective function being proportional to the degree of elbow prominence and the degree of elbow presences and inversely proportional to the degree of smoothness;
determining the value of an optimal objective function of each undetermined characteristic data set, and taking the undetermined characteristic data set with the maximum value of the optimal objective function as the optimal characteristic data set;
the optimal objective function is:
wherein ,representing the optimal objective function>Representing the slope of the line segment formed between the kth and the kth+1th SSE value on the resulting SSE line graph for the pending profile data set, +. >Representing the slope of the line segment formed between the kth-1 and kth SSE values on the resulting SSE line graph for the pending profile data set, +.>And->The values of the kth SSE value and the (k+1) SSE value on the SSE line diagram respectively represent the values of the to-be-determined characteristic data set, the to-be-determined characteristic data set corresponds to the abscissa k value on the SSE line diagram, and the specific clustering quantity in the process of clustering the to-be-determined characteristic data set to obtain the SSE value corresponding to the abscissa k value is represented as k>Representation of +.>Is also represented by the maximum value of the undetermined characteristic data set corresponding to the change value of the slope of the line segments on both sides of the elbow point on the obtained SSE line graph, ">Representing the average value obtained by averaging the absolute values of the slope differences between any two adjacent line segments on the SSE line graph corresponding to the feature data set to be determined, < + >>K-values representing the correspondence of the undetermined characteristic data set to the elbow points on the resulting SSE line graph,/-, respectively>Representing the preset number of clusters, +.>Representing the mean square error of the slope of each line segment on the resulting SSE line graph for the set of undetermined feature data,representation of +.>Is a maximum value of (a).
2. The tracing method for refined tea production according to claim 1, wherein the method for acquiring one-dimensional detection data combination is as follows:
Y one-dimensional detection data are selected from the preliminary screening one-dimensional detection data to obtainA plurality of undetermined characteristic data sets, wherein Y is the set number, and the value range of Y is +.>，/>For said preset number,/->And the total number of the one-dimensional detection data in the one-dimensional detection data is screened for preliminarily.
3. The tracing method for refined tea production according to claim 1, wherein the accumulated value of variance contribution rate is:
wherein W represents the variance contribution ratio accumulated value of one-dimensional detection data, R represents the number of detection indexes under different process flows, namely the number of corresponding obtained one-dimensional detection data under different process flows, L represents the number of process flows,and the variance contribution rate of the r one-dimensional detection data after the variance contribution rates of all one-dimensional detection data in the z-th process flow are sequenced from large to small is represented.
4. The tracing method for refined tea production according to claim 1, wherein the method for re-performing k-means clustering on the multidimensional detection data in the optimal feature data set comprises the following steps:
and taking the k value of the elbow point on the SSE line graph corresponding to the optimal characteristic data set as the optimal clustering quantity, and carrying out k-means clustering on the optimal characteristic data set according to the optimal clustering quantity.
5. A tea refining production traceability system comprising a memory and a processor for executing instructions stored in the memory to implement a tea refining production traceability method according to any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310735477.3A CN116485418B (en) | 2023-06-21 | 2023-06-21 | Tracing method and system for tea refining production |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310735477.3A CN116485418B (en) | 2023-06-21 | 2023-06-21 | Tracing method and system for tea refining production |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116485418A CN116485418A (en) | 2023-07-25 |
| CN116485418B true CN116485418B (en) | 2023-09-05 |
Family
ID=87214151
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310735477.3A Active CN116485418B (en) | 2023-06-21 | 2023-06-21 | Tracing method and system for tea refining production |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116485418B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117078281B (en) * | 2023-10-17 | 2024-01-09 | 肥城恒丰塑业有限公司 | Anti-counterfeiting traceability management method and system for production of soundproof cloth |
| CN117094611B (en) * | 2023-10-20 | 2024-01-09 | 济南优谷生物技术有限公司 | Quality safety traceability management method and system for food processing |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778838A (en) * | 2016-11-29 | 2017-05-31 | 北京科技大学 | A kind of method for predicting air quality |
| CN108241925A (en) * | 2016-12-23 | 2018-07-03 | 重庆邮电大学 | A kind of discrete manufacture mechanical product quality source tracing method based on outlier detection |
| CN111671128A (en) * | 2020-06-05 | 2020-09-18 | 江苏中烟工业有限责任公司 | Processing attribute classification method for threshing and redrying tobacco leaves |
| CN112487720A (en) * | 2020-11-30 | 2021-03-12 | 重庆大学 | Method and system for quickly partitioning wind pressure coefficient based on K-means three-dimensional clustering algorithm and storage medium |
| CN112766440A (en) * | 2021-02-01 | 2021-05-07 | 中穗科技股份有限公司 | Tea production circulation traceability management method and system based on block chain |
| CN113257364A (en) * | 2021-05-26 | 2021-08-13 | 南开大学 | Single cell transcriptome sequencing data clustering method and system based on multi-objective evolution |
| CN115165950A (en) * | 2022-05-07 | 2022-10-11 | 安徽农业大学 | Method for identifying origin tracing of tea leaves through double-phase extraction NMR spectrum and application thereof |
| CN116205738A (en) * | 2023-03-09 | 2023-06-02 | 浙江科技学院 | Financial data selection optimization method based on deep clustering |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3965007A1 (en) * | 2020-09-04 | 2022-03-09 | Hitachi, Ltd. | Action recognition apparatus, learning apparatus, and action recognition method |
-
2023
- 2023-06-21 CN CN202310735477.3A patent/CN116485418B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778838A (en) * | 2016-11-29 | 2017-05-31 | 北京科技大学 | A kind of method for predicting air quality |
| CN108241925A (en) * | 2016-12-23 | 2018-07-03 | 重庆邮电大学 | A kind of discrete manufacture mechanical product quality source tracing method based on outlier detection |
| CN111671128A (en) * | 2020-06-05 | 2020-09-18 | 江苏中烟工业有限责任公司 | Processing attribute classification method for threshing and redrying tobacco leaves |
| CN112487720A (en) * | 2020-11-30 | 2021-03-12 | 重庆大学 | Method and system for quickly partitioning wind pressure coefficient based on K-means three-dimensional clustering algorithm and storage medium |
| CN112766440A (en) * | 2021-02-01 | 2021-05-07 | 中穗科技股份有限公司 | Tea production circulation traceability management method and system based on block chain |
| CN113257364A (en) * | 2021-05-26 | 2021-08-13 | 南开大学 | Single cell transcriptome sequencing data clustering method and system based on multi-objective evolution |
| CN115165950A (en) * | 2022-05-07 | 2022-10-11 | 安徽农业大学 | Method for identifying origin tracing of tea leaves through double-phase extraction NMR spectrum and application thereof |
| CN116205738A (en) * | 2023-03-09 | 2023-06-02 | 浙江科技学院 | Financial data selection optimization method based on deep clustering |
Non-Patent Citations (1)
| Title |
|---|
| 化学指纹图谱检测技术在茶叶领域的应用;郭小媛，董俊杰，叶影，叶俭慧，梁月荣;《茶叶》;第46卷(第1期);第1-7页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116485418A (en) | 2023-07-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116485418B (en) | Tracing method and system for tea refining production | |
| Ucat et al. | Postharvest grading classification of cavendish banana using deep learning and tensorflow | |
| EP1760634B1 (en) | Method, Apparatus and computer program for measurement of mitotic activity | |
| CN116246174B (en) | Sweet potato variety identification method based on image processing | |
| Sari et al. | Classification quality of tobacco leaves as cigarette raw material based on artificial neural networks | |
| CN113449792A (en) | Method for nondestructive rapid detection of food quality | |
| Akter et al. | Development of a computer vision based eggplant grading system | |
| CN114332492B (en) | Vegetation feature extraction and selection method based on long-time sequence | |
| Al Haque et al. | A computer vision system for bangladeshi local mango breed detection using convolutional neural network (cnn) models | |
| CN114549485A (en) | Stem detection method based on X-ray vision | |
| CN113222427A (en) | Fresh electric commercial commodity quality intelligent monitoring and management method based on multi-dimensional characteristic data analysis | |
| Oraño et al. | Analyzing the Accuracy of KNN-based Cacao Bean Grading System | |
| CN117196659B (en) | Product tracing method and system based on block chain technology | |
| Arora et al. | A machine vision based approach to Cashew Kernel grading for efficient industry grade application | |
| Susetyoko et al. | Characteristics of Accuracy Function on Multiclass Classification Based on Best, Average, and Worst (BAW) Subset of Random Forest Model | |
| Kirchgessner et al. | FruitPhenoBox–a device for rapid and automated fruit phenotyping of small sample sizes | |
| Hernández et al. | Unsupervised learning for ripeness estimation from grape seeds images | |
| CN117689999B (en) | Method and system for realizing TC4 tape coiling process optimization | |
| CN116664172B (en) | Food sales data mining analysis method based on big data | |
| Gourav | Rice Grain Quality Determination Using OTSU's Thresholding Method | |
| CN113218880B (en) | Food material detection method of refrigerator, refrigerator and storage medium | |
| CN116326654B (en) | Cooperative control method and system for units in primary tea making process | |
| Prabhakar et al. | Computer vision based technique for surface defect detection of apples | |
| CN113218141B (en) | Food material detection method for refrigerator, refrigerator and storage medium | |
| Thinh | Using Random Forest Algorithm to Grading Mango’s Quality Based on External Features Extracted from Captured Images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A method and system for tracing the origin of tea refining production Granted publication date: 20230905 Pledgee: Bank of China Limited Xiamen hi tech Park sub branch Pledgor: Fujian Jicha Biotechnology Co.,Ltd. Registration number: Y2024980002632 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right |