CN102589470A - Fuzzy-neural-network-based tea leaf appearance quality quantification method - Google Patents

Fuzzy-neural-network-based tea leaf appearance quality quantification method Download PDF

Info

Publication number
CN102589470A
CN102589470A CN2012100318829A CN201210031882A CN102589470A CN 102589470 A CN102589470 A CN 102589470A CN 2012100318829 A CN2012100318829 A CN 2012100318829A CN 201210031882 A CN201210031882 A CN 201210031882A CN 102589470 A CN102589470 A CN 102589470A
Authority
CN
China
Prior art keywords
fuzzy
layer
tealeaves
neural network
network
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.)
Pending
Application number
CN2012100318829A
Other languages
Chinese (zh)
Inventor
蒋艾青
岳鹏翔
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
DAMIN FOODSTUFF (ZHANGZHOU) Co Ltd
Original Assignee
DAMIN FOODSTUFF (ZHANGZHOU) Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by DAMIN FOODSTUFF (ZHANGZHOU) Co Ltd filed Critical DAMIN FOODSTUFF (ZHANGZHOU) Co Ltd
Priority to CN2012100318829A priority Critical patent/CN102589470A/en
Publication of CN102589470A publication Critical patent/CN102589470A/en
Pending legal-status Critical Current

Links

Abstract

The invention discloses a fuzzy-neural-network-based tea leaf appearance quality quantification method, which comprises the following steps of: (a) selecting a batch of representative samples, and carrying out sensory evaluation on the tea leaf samples by tea leaf tasters with national certificates so as to obtain the appearance grading values of the tea leaf samples; (b) obtaining the visible images of the appearances of tea leaves by adopting a computer vision technology, respectively extracting shape features and color features after carrying out pretreatment on the visible images, and carrying out principal component analysis on all extracted feature variables to obtain a group of uncorrelated new variables; (c) establishing a fuzzy-neural-network-based tea leaf appearance quality quantitative evaluation model, taking the front p principal component factors extracted in the step (b) as an input layer of a network, and taking the sensory evaluation grading value of the appearances of the tea leaves as a desired output of a fuzzy neural network model, wherein the fuzzy-neural-network-based tea leaf appearance quality quantitative evaluation model comprises an input layer, a fuzzy layer, a fuzzy rule calculation layer and an output layer; and (d) calculating the graded value of the appearance quality of unknown tea leaf samples by using the established fuzzy neural network model for tea leaf appearance quality.

Description

A kind of quantization method of the tealeaves exterior quality based on fuzzy neural network
Technical field
The present invention relates to a kind of objective quantification method of tealeaves exterior quality, specifically a kind of quantization method of the tealeaves exterior quality based on fuzzy neural network.
Background technology
Tea leaf quality is that various compositions cooperatively interact in the tealeaves, the concentrated expression after coordinated with each other, but the composition of tealeaves is very complicated, can not just can express through the quantification of some internal component; Therefore, at home and abroad, the quality of tea leaf quality, grade classification, value just determine mainly to evaluate through people's sense organ.Sensory review's method is easy relatively, and can differentiate and describe the flavor characteristic of tealeaves, but sensory review result is by the sense organ experience decision of commenting the tea teacher, has tangible randomness and uncertainty, and subjectivity is strong, and consistance is poor.Research adopts instrument to quantize the quality of tea leaves index, can effectively avoid the existing defective of artificial sense assessment method, improves accuracy, stability and consistance that tea leaf quality is estimated.
Tea leaf quality comprises 5 evaluation indexes at the bottom of tealeaves outward appearance, soup look, fragrance, flavour and the leaf, and wherein the tealeaves outward appearance is an evaluation index the most intuitively.The evaluation of tealeaves appearance index comprises that promptly face shaping and the color through watching tealeaves just can roughly be inferred this quality of tea leaves situation to tealeaves face shaping and two aspects of color; The outward appearance of utilizing sensory review's method to estimate tealeaves mainly watches 5 of the shape, color and luster, tender degree of tealeaves outward appearance, whole broken, cleanliness and color and lusters etc. to evaluate the factor.The appearance luster that some literature research adopt the colour examining technology to take quantitative analysis tealeaves; (tea science, 2005 the 1st phases, P37-42) the research and utilization computer was measured the method for tea color to Wang Wenjie; Adopt scanner to obtain the tealeaves picture, describe the tealeaves color through the Lab colour system.(Chinese agronomy circular, 1997 the 6th phases, P24-26) Application Research colour examining technology and multiple regression analysis method were set up the linear regression model (LRM) of congou tea appearance luster and black tea powder color and luster to Yan Jun.Lai Guoliang (Fujian tealeaves; 1999 the 2nd phases; P19-21) Application Research colour examining technology and multiple regression analysis method have been set up the appearance luster of roasted green tea and the linear regression model (LRM) of tea-leaf power color and luster, but colour difference meter is surveyed the visual angle diameter and had only 8mm; When measuring dried tea or leaf background color pool, its representativeness is affected.The tealeaves exterior quality not only comprises the profile color and luster, the shape of its tealeaves, tender degree, whole broken, these evaluations influences of evaluating the factor pair tea leaf quality of cleanliness are also very big, therefore adopt the colour examining technology to be difficult to realize the analysis of the shape facility of tealeaves profile.In recent years; There is literature research to utilize computer vision technique to analyze the exterior quality of tealeaves; Tealeaves is tiled; Obtain the visible images of tealeaves through computer vision technique, extract the shape that profile textural characteristics and color characteristic in the visible images can give expression to the tealeaves outward appearance, tender degree, whole broken, cleanliness and 5 characteristics of evaluating the factor of color and luster again." based on the texture analysis of multispectral image technology differentiate different green tea device (patent No.: ZL2007201100404.1) " propose to adopt the multispectral image technology to obtain the outward appearance textural characteristics of tealeaves.(Chinese journal of scientific instrument, 2006 the 12nd phases, P1704-1706) the research and utilization machine vision technique is discerned different types of tealeaves to Chen Quansheng.(agricultural mechanical journal, 2000 the 4th phases P67-70) are utilized computer vision quantitative description tea color to Cai Jianrong, select the HIS color system to describe the tealeaves color, analyze the tea leaf quality situation that different year is produced.Just utilize computer vision technique to realize the qualitative identification of tea leaf quality in these researchs, and the instrument quantitatively evaluating of the tea leaf quality of being unrealized.
Tea leaf quality is mainly accomplished through the organs of vision by tea-taster, and human body sensory organ mechanism is complicated, and the result who is drawn by the evaluation of sensory review's method has certain subjectivity and ambiguity; And the tealeaves internal component is complicated, and tea leaf quality is the result of various complex internal interaction between component, possibly be complicated nonlinear relationship between biochemical parameter and sensory evaluation scores in tealeaves.Zheng Yan (small-sized microcomputer system, 2004 the 7th phases) propose to adopt fuzzy-neural network method that 5 kinds of tea flavour signals are carried out qualitative identification, also and the quantitative analysis of the tealeaves sense of taste signal of being unrealized.By retrieval, utilize instrument to come quantitative analysis tealeaves exterior quality, and the subjectivity influence that in modeling process, how to reduce sensory review result yet there are no the correlative study report.
Summary of the invention
The objective of the invention is to overcome the tealeaves organoleptic quality and evaluate the defective of existence and the deficiency of above-mentioned prior art, set up a kind of method for quantitatively evaluating of the tealeaves outward appearance based on fuzzy neural network.This method is at first obtained the visible images of tealeaves with computer vision system; Extract the shape facility and the color characteristic of tealeaves visible images then; With the eigenwert of extracting as independent variable; Tealeaves outward appearance sensory review gets score value as dependent variable, sets up the fuzzy neural network evaluation model of tealeaves exterior quality.This method not only has good approaching property of numerical value and stability; And ambiguity, the subjectivity that can handle sensory review result well influence; Have good consistance with the evaluation result of the subjective evaluation method of continuing to use at present always, thereby improve accuracy and consistance that tea leaf quality is estimated.
For solving the problems of the technologies described above, the present invention realizes as follows:
(1) chooses a collection of representative sample, the outward appearance of each tea appearance is carried out the sensory review, provide the outward appearance score value of each tea appearance by tea-taster with national qualification certificate;
(2) adopt computer vision technique to obtain the visible images of tealeaves, behind the pre-service visible images, extract shape facility and color characteristic respectively; Characteristic variable to all extractions is carried out principal component analysis (PCA), obtains one group of mutual incoherent new variables;
(3) foundation is based on the quantitatively evaluating model of the tealeaves exterior quality of fuzzy neural network, and this model is divided into input layer, obfuscation layer, fuzzy rule computation layer and output layer; Preceding p the major component factor number that is extracted by step (2) be as the input layer of network, with tealeaves outward appearance sensory review to the desired output of score value as fuzzy neural network model; Adopt Gauss's membership function that input value is carried out obfuscation and obtain the fuzzy membership value, with this output as the obfuscation layer; Have the version of multilayer perceptron based on the fuzzy neural network of simplified model, its parameter adaptive adjusting method adopts the bp neural network algorithm;
(4) utilize the fuzzy neural network model of the tealeaves exterior quality set up, calculate unknown tea appearance exterior quality score value.
Good effect of the present invention:
The quantization method of a kind of tealeaves exterior quality based on fuzzy neural network according to the invention; Its neural net method is the function and thinking of simulation human brain structure; Through self study, certainly group know, non-linear parallel dispersion dynamics on the neural network of adaptation function, pattern information that can't languageization is handled; And fuzzy logic is according to the subordinate function of artificial definition and the rule of a series of parallel series, goes to handle the information of various ambiguities with reasoning from logic, is to represent and analyze Method and kit for uncertain, out of true information through the apish mode of thinking.With the input as fuzzy neural network model of the characteristic variable of outward appearance, outward appearance sensory review's score is as the output of network, sets up the contact between subjective, the objective evaluation standard, realizes utilizing the evaluation index of instrument quantitative analysis tealeaves.After through this fuzzy-neural network method tealeaves external appearance characteristic and outward appearance sensory review score being carried out Fuzzy processing, can eliminate the subjectivity influence of evaluating the result, adopt this method can reach following effect:
(1) fuzzy neural network model has solved interference problem, has eliminated the subjectivity influence that artificial sense is evaluated the method human factor, and it is to the prediction of exterior quality more accurate also to make.
(2) model based on neural network has better adaptive faculty and more is prone to realization.Fuzzy neural network in conjunction with expertise, can reduce the requirement to training dataset through introducing the fuzzy set notion, and computer realization is more prone to.Four layers of feedforward network structure that fuzzy neural network adopts both can realize complicated Nonlinear Mapping, can obtain bigger output area again, had improved adaptability.
Description of drawings
Below in conjunction with accompanying drawing and embodiment the present invention is done further explain:
Fig. 1 be utilize computer vision technique combine fuzzy-neural network method quantitative analysis tealeaves exterior quality method principle steps figure (wherein; What filled box comprised among the figure is the process of setting up of the quantitatively evaluating model of tealeaves exterior quality, and what dashed rectangle comprised is the forecasting process that unknown tealeaves exterior quality gets score value)
The structural drawing of Fig. 2 fuzzy neural network model
Embodiment
Below in conjunction with accompanying drawing and embodiment the present invention is explained further details:
As special embodiment of the present invention, shown in Figure 1ly, the present invention combines the principle steps figure of the method for fuzzy-neural network method quantitative analysis tealeaves exterior quality for utilizing computer vision technique, and method of the present invention comprises following steps:
Step 1: choose a collection of representational tealeaves sample (tealeaves that refers to same kind here; Such as green tea or black tea or oolong tea etc.); By the tea-taster who has the national qualification certificate more than 3, with reference to standard GB/T14487-2008, the form that adopts collective's scoring and password to evaluate; Under justice, just condition; Mark for the exterior quality of every batch of tea appearance by centesimal system, remove wrong, inconsistent or incomplete sensory evaluation scores, the mean value of all tea-tasters' scoring is as the exterior quality final score value of each tea appearance.
Step 2: adopt computer vision technique to obtain the outward appearance visible images of each tea appearance, behind the pre-service visible images, extract shape facility and color characteristic respectively, the concrete steps of this part are following:
(1) from each tealeaves sample, gets about 10g, evenly be tiled in the double dish that specification is Φ 10 * 1cm, adopt computer vision system to obtain the visible images of each tea appearance with the inquartation skewer;
(2) find the center pixel position of each width of cloth image, as the center, intercepting around it 200 * 200 field as object region;
(3) from object region, extract respectively 6 textural characteristics (being respectively: average gray level, standard variance, smoothness, third moment, consistance, entropy), 6 textural characteristics (being respectively: circumferential spectrum energy amplitude, radially spectrum energy amplitude, circumferential spectrum energy average, radially spectrum energy average, circumferential spectrum energy variance, spectrum energy variance radially), 28 textural characteristics (be respectively: gray scale moment of inertia, gray scale correlativity, gray scale energy, gray scale homogeney, corpus hemorrhagicum moment of inertia, corpus hemorrhagicum correlativity, corpus hemorrhagicum energy, corpus hemorrhagicum homogeney, green body moment of inertia, green body correlativity, green physical efficiency, green body homogeney, blue body moment of inertia, blue body correlativity, blue physical efficiency, blue body homogeney, tone are used to body square, tone correlativity, tone energy, tone homogeney, brightness moment of inertia, luminance correlation, luminance energy, brightness homogeney, saturation degree moment of inertia, saturation degree correlativity, saturation degree energy, saturation degree homogeney), 12 color characteristics based on gray level co-occurrence matrixes tolerance based on frequency spectrum tolerance based on statistical moment tolerance (be respectively: corpus hemorrhagicum average, green body average, blue body average, corpus hemorrhagicum standard deviation, green body standard deviation, blue body standard deviation, tone average, brightness average, saturation degree average, tonal criterion is poor, luminance standard is poor, the saturation degree standard deviation); Promptly extract 40 textural characteristics variablees and 12 color characteristic variablees, extract the characteristic variable of the characteristic variable of 52 outward appearances altogether as this tea appearance;
(4) extracted the characteristic of tealeaves outward appearance after; Because the dimension of characteristic is bigger; Directly sending into follow-up fuzzy neural network model foundation will be with the time complexity that increases training greatly; Also there are a lot of invalid redundancy features can influence model accuracy simultaneously in these characteristics; Even make network produce the over-fitting problem, so adopt principal component analytical method that original variable is carried out principal component analysis (PCA) again, obtain one group of new mutual incoherent variable (major component); Get the accumulation variance contribution ratio greater than 90% preceding P major component as follow-up data processing; Because when the variance of P major component accumulation contribution rate greater than 90% the time, the information that is comprised of this P major component enough can be expressed the information of original variable, wherein ratio is calculated as:
Step 3: set up the quantitatively evaluating model based on the tealeaves exterior quality of fuzzy neural network, this model is divided into input layer, obfuscation layer, fuzzy rule computation layer and output layer; Preceding p the major component factor number that is extracted by step 2 be as the input layer of network, with tealeaves outward appearance sensory review to the desired output of score value as fuzzy neural network model; Adopt Gauss's membership function that input value is carried out obfuscation and obtain the fuzzy membership value, with this output as the obfuscation layer; Have the version of multilayer perceptron based on the fuzzy neural network of simplified model, its parameter adaptive adjusting method adopts the bp neural network algorithm; The learning function of the output error of define grid model is: In the formula, y dBe the network desired output, Be the actual output of network, e is a desired output and the error of actual output, generates the correlation parameter of FUZZY NETWORK model through continuous study, reaches the error of defined up to network model, stops training.
Shown in Figure 2 is the structural drawing of fuzzy neural network model of the present invention, explains that in conjunction with Fig. 2 network model sets up process:
(1) input layer.Preceding P the major component that is calculated by step 2 step be as the input variable of network model, promptly input vector be x=[x1, x2, x3, x4 ..., xp], input layer is connected with input vector xi, makes Z j=x p, j=p, y j=Z j, x pP input for network model.
(2) obfuscation layer.At the obfuscation layer, adopt subordinate function that input value is carried out obfuscation, obtain the fuzzy membership value, the present invention adopts Gauss's membership function to calculate each input variable x pDegree of membership: y j=exp (Z j), c in the formula PmBeing the center of Gauss's subordinate function, promptly is m fuzzy subset's of p input correspondence average, b PmBeing the width of Gauss's subordinate function, promptly is m fuzzy subset's of p input correspondence standard deviation, and the anti-pass error term of obfuscation layer is:
The parameter of obfuscation, i.e. the Center Parameter c of Gauss's membership function and width parameter b have been confirmed by input layer to the neuronic connection weights of obfuscation layer.Center Parameter c and width parameter b adopt following algorithm to carry out:
Center Parameter c modified value is: c Ij ( n + 1 ) = c Ij ( n ) - η ∂ e ( n ) ∂ c Ij + α Δ c ;
Width parameter b modified value is: b Ij ( n + 1 ) = b Ij ( n ) - η ∂ e ( n ) ∂ b Ij ( n ) + α Δ b ;
In the formula, η is a learning rate, and α is a factor of momentum, Δ=c i(n)-c i(n-1), Δ b=b i(n)-b i(n-1).
(3) fuzzy rule layer.At the fuzzy rule layer degree of membership of obscuring layer output is carried out Fuzzy Calculation, adopt following formula to calculate: M is a number of fuzzy rules in formula.
(4) output layer: output layer adopts the output of following formula computational grid: y 0=Z 0, W in the formula pBe the weight of correspondence, y cOutput for fuzzy neural network.The output layer weight is repaiied item: W Ij ( n + 1 ) = W Ij ( n ) - η ∂ E p ( n ) ∂ W Ij ( n ) + α Δ W ,
η is a learning rate, and α is a factor of momentum, Δ W=W (n)-W (n-1).
Move above-mentioned (2), (3), (4) step repeatedly; When error during, adjustment parameter (the center c of Gauss's subordinate function, width b and weighted value W) more than or equal to setting threshold, up to error less than pre-set threshold; Stop training; The fuzzy neural network training is accomplished, and deposits the neural network that trains in the neural network storehouse of system, has promptly obtained the quantitative analysis model of tealeaves outward appearance official quality.
Step 4: utilize the network model of setting up to predict the value of drawing of the exterior quality of the tea appearance of seeking knowledge.For a collection of tea appearance of not carrying out sensory evaluation; To carry out quantitative analysis to the exterior quality of this batch tea appearance; Carry out to the step 4 step by the step 2 of above-mentioned implementation process; Be input in the fuzzy neural network through test utilizing step 2 to carry out preceding P the major component that principal component analysis (PCA) obtains, after the network operation, can obtain the quantized value that the exterior quality of each tea appearance is estimated.
Above-mentioned is embodiment of the present invention, but design concept of the present invention is not limited to this, everyly utilizes this design that the change of unsubstantiality is carried out in this invention, all belongs to the behavior of invading protection domain of the present invention.

Claims (2)

1. quantization method based on the tealeaves exterior quality of fuzzy neural network, it is characterized in that: step of the present invention does
A, choose a collection of representative sample, the profile of each tea appearance is carried out the sensory review, provide the profile score value of each tea appearance by tea-taster with national qualification certificate;
B, employing computer vision technique obtain the visible images of tealeaves profile, behind the pre-service visible images, extract shape facility and color characteristic respectively; Characteristic variable to all extractions is carried out principal component analysis (PCA), obtains one group of mutual incoherent new variables;
C, set up the quantitatively evaluating model based on the tealeaves profile quality of fuzzy neural network, this model is divided into input layer, obfuscation layer, fuzzy rule computation layer and output layer; Preceding p the major component factor number that is extracted by step b be as the input layer of network, with tealeaves profile sensory review to the desired output of score value as fuzzy neural network model;
D, utilize the fuzzy neural network model of the tealeaves profile quality set up, calculate unknown tea appearance the profile quality score value.
2. the quantization method of a kind of tealeaves exterior quality based on fuzzy neural network according to claim 1; It is characterized in that: step c adopts fuzzy-neural network method, sets up tealeaves profile qualitative characteristics and the profile sensory review gets the fuzzy neural network model between the score value; This model is divided into input layer, obfuscation layer, fuzzy rule computation layer and output layer, and wherein obscuring layer adopts Gauss's membership function to calculate the degree of membership of each input variable, and the parameter adaptive adjusting method of each layer adopts the BP neural network algorithm.
CN2012100318829A 2012-02-14 2012-02-14 Fuzzy-neural-network-based tea leaf appearance quality quantification method Pending CN102589470A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2012100318829A CN102589470A (en) 2012-02-14 2012-02-14 Fuzzy-neural-network-based tea leaf appearance quality quantification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2012100318829A CN102589470A (en) 2012-02-14 2012-02-14 Fuzzy-neural-network-based tea leaf appearance quality quantification method

Publications (1)

Publication Number Publication Date
CN102589470A true CN102589470A (en) 2012-07-18

Family

ID=46478463

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2012100318829A Pending CN102589470A (en) 2012-02-14 2012-02-14 Fuzzy-neural-network-based tea leaf appearance quality quantification method

Country Status (1)

Country Link
CN (1) CN102589470A (en)

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104008551A (en) * 2014-06-10 2014-08-27 华南农业大学 Liberobacter asiaticum detection method based on visible light images
CN104299010A (en) * 2014-09-23 2015-01-21 深圳大学 Image description method and system based on bag-of-words model
CN104569328A (en) * 2013-10-28 2015-04-29 云南天士力帝泊洱生物茶集团有限公司 Method for determining taste quality and aroma quality of Pu 'er sunshine green tea
CN105045091A (en) * 2015-07-14 2015-11-11 河海大学常州校区 Dredging process intelligent decision analysis method based on fuzzy neural control system
CN105403507A (en) * 2014-09-10 2016-03-16 中国农业科学院茶叶研究所 Mobile-terminal-based Longjing tea identification and judgment system and method
CN106127226A (en) * 2016-06-14 2016-11-16 河南工业大学 Grain grain and the flexible grain quality detection method of grain grain test sample
CN108287161A (en) * 2017-01-09 2018-07-17 中国计量大学 A kind of Intelligent tea organoleptic evaluation method and system
CN110320173A (en) * 2019-06-14 2019-10-11 湖北省农业科学院果树茶叶研究所 The method for rapidly judging of machine fresh tea picking mee tea vehicle tinctorial pattern product grade based on particle swarm optimization algorithm
CN110361334A (en) * 2019-06-14 2019-10-22 湖北省农业科学院果树茶叶研究所 The method for adopting mee tea vehicle tinctorial pattern product grade using general regression structure non-destructive prediction machine
CN110545373A (en) * 2018-05-28 2019-12-06 中兴通讯股份有限公司 Spatial environment sensing method and device
CN110705655A (en) * 2019-11-05 2020-01-17 云南省烟草农业科学研究院 Tobacco leaf classification method based on coupling of spectrum and machine vision
CN113358153A (en) * 2021-05-28 2021-09-07 杨春燕 Tea frying mechanical rotation detection system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101059425A (en) * 2007-05-29 2007-10-24 浙江大学 Method and device for identifying different variety green tea based on multiple spectrum image texture analysis
CN102013021A (en) * 2010-08-19 2011-04-13 汪建 Tea tender shoot segmentation and identification method based on color and region growth
CN102222164A (en) * 2011-05-30 2011-10-19 中国标准化研究院 Food sensory quality evaluation method and system thereof

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101059425A (en) * 2007-05-29 2007-10-24 浙江大学 Method and device for identifying different variety green tea based on multiple spectrum image texture analysis
CN102013021A (en) * 2010-08-19 2011-04-13 汪建 Tea tender shoot segmentation and identification method based on color and region growth
CN102222164A (en) * 2011-05-30 2011-10-19 中国标准化研究院 Food sensory quality evaluation method and system thereof

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
李姣: "茶叶品质的计算机视觉分级技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》, no. 05, 15 May 2007 (2007-05-15), pages 138 - 488 *
郑岩等: "基于模糊神经网络方法实现茶味信号识别的研究", 《小型微型计算机系统》, vol. 25, no. 7, 31 July 2004 (2004-07-31) *

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104569328A (en) * 2013-10-28 2015-04-29 云南天士力帝泊洱生物茶集团有限公司 Method for determining taste quality and aroma quality of Pu 'er sunshine green tea
CN104008551B (en) * 2014-06-10 2017-06-13 华南农业大学 A kind of Citrus Huanglongbing pathogen detection method based on visible images
CN104008551A (en) * 2014-06-10 2014-08-27 华南农业大学 Liberobacter asiaticum detection method based on visible light images
CN105403507A (en) * 2014-09-10 2016-03-16 中国农业科学院茶叶研究所 Mobile-terminal-based Longjing tea identification and judgment system and method
CN104299010A (en) * 2014-09-23 2015-01-21 深圳大学 Image description method and system based on bag-of-words model
CN104299010B (en) * 2014-09-23 2017-11-10 深圳大学 A kind of Image Description Methods and system based on bag of words
CN105045091A (en) * 2015-07-14 2015-11-11 河海大学常州校区 Dredging process intelligent decision analysis method based on fuzzy neural control system
CN106127226A (en) * 2016-06-14 2016-11-16 河南工业大学 Grain grain and the flexible grain quality detection method of grain grain test sample
CN106127226B (en) * 2016-06-14 2019-09-03 河南工业大学 The flexible grain quality detection method of grain grain and grain grain test sample
CN108287161A (en) * 2017-01-09 2018-07-17 中国计量大学 A kind of Intelligent tea organoleptic evaluation method and system
CN110545373A (en) * 2018-05-28 2019-12-06 中兴通讯股份有限公司 Spatial environment sensing method and device
CN110545373B (en) * 2018-05-28 2021-12-28 中兴通讯股份有限公司 Spatial environment sensing method and device
CN110320173A (en) * 2019-06-14 2019-10-11 湖北省农业科学院果树茶叶研究所 The method for rapidly judging of machine fresh tea picking mee tea vehicle tinctorial pattern product grade based on particle swarm optimization algorithm
CN110361334A (en) * 2019-06-14 2019-10-22 湖北省农业科学院果树茶叶研究所 The method for adopting mee tea vehicle tinctorial pattern product grade using general regression structure non-destructive prediction machine
CN110705655A (en) * 2019-11-05 2020-01-17 云南省烟草农业科学研究院 Tobacco leaf classification method based on coupling of spectrum and machine vision
CN113358153A (en) * 2021-05-28 2021-09-07 杨春燕 Tea frying mechanical rotation detection system

Similar Documents

Publication Publication Date Title
CN102589470A (en) Fuzzy-neural-network-based tea leaf appearance quality quantification method
CN105100789B (en) A kind of method for evaluating video quality
CN106991666B (en) A kind of disease geo-radar image recognition methods suitable for more size pictorial informations
CN102421007B (en) Image quality evaluating method based on multi-scale structure similarity weighted aggregate
CN106096561A (en) Infrared pedestrian detection method based on image block degree of depth learning characteristic
CN107292343A (en) A kind of Classification of hyperspectral remote sensing image method based on six layers of convolutional neural networks and spectral space information consolidation
CN107463919A (en) A kind of method that human facial expression recognition is carried out based on depth 3D convolutional neural networks
CN104459668A (en) Radar target recognition method based on deep learning network
CN104268590B (en) The blind image quality evaluating method returned based on complementary combination feature and multiphase
CN102539326A (en) Method for carrying out quantitative evaluation on soup hue quality of tea
CN109145992A (en) Cooperation generates confrontation network and sky composes united hyperspectral image classification method
CN104680541B (en) Remote Sensing Image Quality evaluation method based on phase equalization
CN103761742B (en) A kind of high-spectrum remote sensing sparse solution mixing method based on homogeneity index
Shah Rizam et al. Non-destructive watermelon ripeness determination using image processing and artificial neural network (ANN)
CN106127234B (en) Non-reference picture quality appraisement method based on characteristics dictionary
CN108776831A (en) A kind of complex industrial process Data Modeling Method based on dynamic convolutional neural networks
CN108596108A (en) Method for detecting change of remote sensing image of taking photo by plane based on the study of triple semantic relation
CN104103033A (en) Image real-time processing method
Jiang et al. Application of back propagation neural network in the classification of high resolution remote sensing image: take remote sensing image of Beijing for instance
CN102297822A (en) Method for predicting ash content of coal particles by utilizing image analysis
CN108875794A (en) A kind of image visibility detection method based on transfer learning
CN111783555A (en) Deep learning satellite data cloud detection algorithm using unified samples
CN107844870A (en) Heavy metal content in soil Forecasting Methodology based on Elman neural network models
CN106920231A (en) A kind of remote sensing image clouds appraisal procedure based on full-colour image statistical nature
CN109683161A (en) A method of the inverse synthetic aperture radar imaging based on depth ADMM network

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20120718