CN111274860B - Recognition method for online automatic tobacco grade sorting based on machine vision - Google Patents
Recognition method for online automatic tobacco grade sorting based on machine vision Download PDFInfo
- Publication number
- CN111274860B CN111274860B CN201911087967.7A CN201911087967A CN111274860B CN 111274860 B CN111274860 B CN 111274860B CN 201911087967 A CN201911087967 A CN 201911087967A CN 111274860 B CN111274860 B CN 111274860B
- Authority
- CN
- China
- Prior art keywords
- tobacco
- tobacco leaf
- machine vision
- leaf
- features
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Probability & Statistics with Applications (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Geometry (AREA)
- Manufacture Of Tobacco Products (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of tobacco leaf grade separation and machine vision, and discloses an online automatic tobacco leaf grade separation identification method based on machine vision, which comprises the following steps: a) Tobacco pretreatment is carried out; b) Collecting tobacco leaf images by using a machine vision sensor, and preprocessing the tobacco leaf images; c) Building a convolutional neural network model to obtain high-level tobacco leaf characteristics; d) Extracting tobacco leaf characteristics, carrying out characteristic fusion on the tobacco leaf characteristics and high-level tobacco leaf characteristics, and establishing a fusion network formed by full-connection layers; e) And aggregating the output of the last full-connection layer into the total characteristics, and inputting the global characteristics into a classifier to obtain the tobacco leaf grade sorting result. According to the invention, the tobacco leaf image is acquired by utilizing the machine vision technology, the color correction is carried out on the tobacco leaf image, the convolutional neural network model is constructed to obtain high-level tobacco leaf characteristics, the tobacco leaf characteristics are fused, the fusion network is adopted to carry out tobacco leaf grade identification, and the identification accuracy is high.
Description
Technical Field
The invention relates to the technical field of tobacco leaf grade sorting and machine vision, in particular to an identification method for online automatic tobacco leaf grade sorting based on machine vision.
Background
The machine vision technology converts a shot target into an image signal through an image shooting device, transmits the image signal to a special image processing system to obtain form information of the shot target, analyzes the content in an image according to pixel distribution, brightness, color and other information, obtains preset information and a preset result, and further controls the action of on-site equipment according to the judging result. The purchased tobacco leaves are required to be classified and sorted secondarily in the early stage of the tobacco redrying link, the secondarily classified and sorted tobacco leaves are boxed and stored according to the classification, and the tobacco leaves are used for later redrying procedures, the existing tobacco leaf classification is manually measured on line, the randomness is particularly high, the accuracy is low, the factor of the tobacco leaf sorting result is relatively large under the condition that no sensor is used for assistance, and the quality of the subsequent treatment of the tobacco leaves is not guaranteed.
For example, a "a tobacco leaf classification baking process" disclosed in chinese patent literature, publication number CN104161298A, the invention classifies fresh tobacco leaves according to 18 categories, the 18 categories being respectively lower leaf-water content, lower leaf-water content normal, lower leaf-water content low, lower leaf-root disease, lower leaf-overnutrition, lower leaf-premature de-fertilisation, middle leaf-water content high, middle leaf-water content normal, middle leaf-water content low, middle leaf-root disease, middle leaf-overnutrition, middle leaf-premature de-fertilisation, upper leaf-water content high, upper leaf-water content normal, upper leaf-water content low, upper leaf-root disease, upper leaf-overnutrition, upper leaf-premature de-fertilisation; each category corresponds to a baking temperature system; and storing each baking temperature system in a temperature controller for tobacco growers to select. According to the invention, the tobacco leaf grade separation is performed manually, so that the influence of human factors is large, and the accuracy of the tobacco leaf grade separation is relatively low.
Disclosure of Invention
The invention provides an online automatic tobacco grade sorting identification method based on machine vision, which aims to solve the problems of large influence of human factors and low tobacco grade sorting accuracy in the tobacco grade sorting process. According to the invention, the tobacco leaf image is acquired by utilizing the machine vision technology, the tobacco leaf feature fusion is carried out, the global feature is constructed, the tobacco leaf grade identification is carried out by adopting a method based on a convolutional neural network model, and the identification accuracy is high.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
an identification method of online automatic tobacco grade sorting based on machine vision comprises the following steps:
a) Tobacco pretreatment is carried out, tobacco leaves are spread and spread, and the tobacco leaves are sequentially and dispersedly arranged;
b) The method comprises the steps of collecting tobacco leaf images by using a machine vision sensor, preprocessing the tobacco leaf images, and taking the preprocessed tobacco leaf images as tobacco leaf samples;
c) Building a convolutional neural network model, and training the convolutional neural network model by using tobacco samples to obtain high-level tobacco leaf characteristics;
d) Extracting tobacco leaf characteristics, carrying out characteristic fusion on the tobacco leaf characteristics and high-level tobacco leaf characteristics, and establishing a fusion network formed by all connection layers, wherein the last all connection layer of the fusion network is connected with a classifier;
e) And aggregating the output of the last full-connection layer into the total characteristics, and inputting the global characteristics into a classifier to obtain the tobacco leaf grade sorting result.
The method comprises the steps of carrying out high-level feature learning on tobacco leaves by adopting a convolutional neural network model, and then realizing feature fusion of the tobacco leaves by adopting a fusion network formed by full-connection layers, so as to construct global features of the tobacco leaves for sorting tobacco leaf grades.
Further, the preprocessing of the tobacco leaf image in the step B) comprises the step of adopting a color correction algorithm to carry out color correction on the tobacco leaf image to obtain a corrected tobacco leaf image I new =am×i, I represents the tobacco leaf image before correction, a is the color conversion matrix, and M is the color conversion model.
The color of the collected image has a great relation with the collection environment, and the color of the pictures obtained by the same sample under different collection environments is different, so that the comparison among the samples is influenced. In the actual photographing process, the machine vision sensor photographs tobacco leaves and is interfered by a sensor CCD or CMOS, a light source, external illumination and the like, so that color correction is required to be carried out on the collected samples, and the tobacco leaf image is more approximate to a true value.
Further, the color correction algorithm is any one of a polynomial regression method, a back propagation network method, and a support vector regression method.
The color correction algorithm adopts any one of a polynomial regression method, a backward propagation network method and a support vector regression method, and the three color correction algorithms are all based on a supervision idea, namely, the color correction is carried out on the picture through a standard color plate, and parameters are required to be set according to actual problems.
Further, the convolutional neural network model in the step C) comprises n convolutional layers, m pooling layers and k full-connection layers, all pooling layers adopt average pooling, and the last full-connection layer is connected with the classifier.
The last full-connection layer of the convolutional neural network model layer is connected with a classifier, and a tobacco leaf grade sorting result is output by the classifier.
Further, in the step C), the method further includes calculating a total sample loss function, and updating a weight value of the convolutional neural network model; by calculation ofObtaining a loss function->A true value representing the j-th tobacco sample level, S j The j-th tobacco sample grade predicted value output by the classifier of the convolutional neural network model is represented, and l represents the number of tobacco grades; by calculating->Obtaining a total sample loss function L i (S, y) represents the loss function of the ith tobacco sample, and r represents the total number of tobacco samples.
The loss function refers to the error of one sample, and the invention calculates the loss function of a single sampleAnd then the loss functions of all the samples are superimposed to obtain the total sample loss function
Further, in step D), the tobacco leaf features include color features, geometric features, overall texture features and tobacco leaf vein texture features,
obtaining a color histogram of the tobacco sample by utilizing the tobacco sample, calculating the mean value, standard deviation, kurtosis and skewness of the color histogram, and taking the mean value, standard deviation, kurtosis and skewness of the color histogram as color characteristics;
the method comprises the steps of obtaining a gray level co-occurrence matrix of a tobacco sample by utilizing the tobacco sample, calculating ASM energy, contrast, inverse difference moment and entropy of the gray level co-occurrence matrix, and taking the ASM energy, contrast, inverse difference moment and entropy of the gray level co-occurrence matrix as integral texture features;
obtaining RGB images of tobacco samples, converting the RGB images into gray level images, carrying out Gaussian smoothing treatment on the gray level images, calculating second moments of the gray level images to obtain leaf vein lines, matching all leaf vein lines through a leaf vein image template, and counting the number, total length, maximum value and minimum value of the leaf vein lines as tobacco vein texture features;
the eccentricity, the slenderness, the rectangularity and the compactness of the minimum outsourcing ellipse of the tobacco leaf are taken as geometric characteristics.
The values in the histogram are all counted, and the mean value, standard deviation, kurtosis and skewness of the histogram reflect the statistical distribution characteristics of the histogram. The gray level image is subjected to Gaussian smoothing, the second moment of the gray level image is calculated to obtain leaf vein lines, then templates of leaf vein images are used for matching to obtain all leaf vein lines, and then the number, the total length, the maximum value and the minimum value of the leaf vein lines of tobacco leaves are counted to be used as the vein texture characteristics of the tobacco leaves. The gray level co-occurrence matrix is a common method for describing textures by researching the spatial correlation characteristics of gray level, ASM energy is the sum of squares of the element values of the gray level co-occurrence matrix, and a calculation formula is thatG (i, j) represents the element value of the ith row and the jth column, h is the total number of rows of the gray level co-occurrence matrix, G is the total number of columns of the gray level co-occurrence matrix,ASM energy reflects the image gray level distribution uniformity and texture thickness. The contrast directly reflects the contrast of the brightness of a certain pixel value and its field pixel value. The inverse moment reflects the homogeneity of the image texture, measures the local change of the image texture, and indicates that the image texture is lack of change in different areas and is very uniform in local when the inverse moment is large. Entropy is a measure of the amount of information that an image has, texture information also belongs to the information of an image, and is a measure of randomness, representing the degree of non-uniformity or complexity of textures in an image. In the geometric characteristics, the thin length is the ratio of the length and the width of the minimum circumscribed rectangle, the rectangle degree is the ratio of the tobacco leaf area to the minimum circumscribed rectangle area of the tobacco leaf, and the compactness degree is the ratio of the tobacco leaf area to the tobacco leaf circumference.
Further, the step D) further comprises the step of reducing the dimension of the tobacco leaf characteristics by utilizing a PCA algorithm to obtain p tobacco leaf characteristics with the largest correlation with tobacco leaf grade sorting.
The PCA algorithm, principal component analysis, is a dimension reduction algorithm that converts high-dimensional data into low-dimensional data with minimal loss. The extracted tobacco features are numerous, if the tobacco features are all used for model training, overfitting is easy to cause, so that the PCA algorithm is adopted to reduce the number of features, the generalization capability of the model is stronger, and the overfitting is reduced.
Further, in the step D), the tobacco leaf characteristics and the high-level tobacco leaf characteristics are subjected to characteristic fusion, and the method comprises the following steps: d1 Fixing the weight value of the trained convolutional neural network model;
d2 Removing the classifier of the convolutional neural network model;
d3 Combining the output of the last full-connection layer of the convolutional neural network model with the classifier removed with tobacco leaf features to form a row to obtain a total feature vector, and taking the total feature vector as the input of the fusion network.
The invention adopts a fusion network formed by full-connection layers to carry out fusion learning on the high-level tobacco leaf characteristics and tobacco leaf characteristics learned by the convolutional neural network model.
Further, in step D), the fusion network includes an input data layer and a full connection layer, the total feature vector is used as the input data layer of the fusion network, a random inactivation optimization method is adopted between the full connection layers, and a classifier of the fusion network adopts a linear support vector machine classifier or a softmax classifier.
Random inactivation (dropout) is a method for optimizing an artificial neural network with a deep structure, and in the learning process, the interdependencies among nodes are reduced by randomly zeroing partial weights or outputs, so that regularization of the neural network is realized, the generalization capability of a model is improved, and the structural risk of the network is reduced.
Further, the image in the step B) is an RGB image or an HSV image.
The machine vision sensor can quantitatively measure the color, and the CCD or the CMOS respectively senses the visible light of different frequency spectrums through filtering to form an RGB color space and generate an RGB image. RGB color space is a type of machine-suitable description, however, not suitable for the perception of the human eye. The invention selects the original RGB image collected by the machine vision sensor or converts the original RGB image into the HSV image which is more consistent with subjective feeling of people as a color feature space, thereby extracting the tobacco leaf features.
Therefore, the invention has the following beneficial effects: according to the invention, a machine vision technology is utilized to collect tobacco leaf images, color correction is carried out on the tobacco leaf images, a convolutional neural network model is constructed to obtain high-level tobacco leaf characteristics, color characteristics, texture characteristics and geometric characteristics of tobacco leaves are obtained, tobacco leaf characteristic fusion is carried out, global characteristics of tobacco leaves are constructed, and a fusion network is adopted to carry out tobacco leaf grade identification, so that identification accuracy is high.
Drawings
Fig. 1 is a flow chart of a first embodiment of the present invention.
Fig. 2 is a schematic flow chart of color correction of a tobacco leaf image according to an embodiment of the present invention.
FIG. 3 is a feature fusion diagram of an embodiment of the present invention.
Fig. 4 is a tobacco context diagram of a first embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the drawings and detailed description.
In a first embodiment, a machine vision-based identification method for online automatic tobacco leaf grade sorting, as shown in fig. 1, includes the steps of:
a) Tobacco pretreatment is carried out, tobacco leaves are spread and spread, and the tobacco leaves are sequentially and dispersedly arranged;
b) The method comprises the steps of collecting a tobacco leaf image by using a machine vision sensor, preprocessing the tobacco leaf image by using an RGB image as the tobacco leaf image, and carrying out color correction on the tobacco leaf image by using a polynomial regression color correction algorithm to obtain a corrected tobacco leaf image I as shown in figure 2 new =am×i, I represents the tobacco leaf image before correction, a is the color conversion matrix, and M is the color conversion model. The machine vision sensor can quantitatively measure the color, and the CCD or the CMOS respectively senses the visible light of different frequency spectrums through filtering to form an RGB color space and generate an RGB image. RGB color space is a type of machine-suitable description, however, not suitable for the perception of the human eye. In agreement with human eye observation is the HIS or HSV color space, XYZ color space being able to construct a device independent color space. The RGB color space is converted into a HIS, HSV, or XYZ color space by color space conversion.
Taking the pretreated tobacco leaf image as a tobacco leaf sample.
C) And establishing a convolutional neural network model, wherein the convolutional neural network model comprises n convolutional layers, m pooling layers and k full-connection layers, all pooling layers adopt average pooling, and the last full-connection layer is connected with a classifier. Training the convolutional neural network model by using tobacco samples to obtain high-level tobacco leaf characteristics;
in the step C), the method also comprises the steps of calculating a total sample loss function and updating the weight value of the convolutional neural network model; by calculation ofObtaining a loss function->A true value representing the j-th tobacco sample level, S j Representing convolution godThe j-th tobacco sample grade predicted value output by the classifier of the network model, wherein l represents the number of tobacco grades; by calculating->Obtaining a total sample loss function L i (S, y) represents the loss function of the ith tobacco sample, and r represents the total number of tobacco samples.
D) Extracting tobacco leaf features including color features, geometric features, overall texture features and tobacco leaf vein texture features,
obtaining a color histogram of the tobacco sample by utilizing the tobacco sample, calculating the mean value, standard deviation, kurtosis and skewness of the color histogram, and taking the mean value, standard deviation, kurtosis and skewness of the color histogram as color characteristics;
the method comprises the steps of obtaining a gray level co-occurrence matrix of a tobacco sample by utilizing the tobacco sample, calculating ASM energy, contrast, inverse difference moment and entropy of the gray level co-occurrence matrix, and taking the ASM energy, contrast, inverse difference moment and entropy of the gray level co-occurrence matrix as integral texture features;
the RGB image of the tobacco sample is obtained, the RGB image is converted into a gray level image, the gray level image is subjected to Gaussian smoothing, the second moment of the gray level image is calculated to obtain leaf vein lines, all the leaf vein lines are matched through a leaf vein image template, and the number, the total length, the maximum value and the minimum value of the leaf vein lines are counted to be used as the vein texture characteristics of the tobacco. As shown in fig. 4, is a context diagram of tobacco leaves.
The eccentricity, the slenderness, the rectangularity and the compactness of the minimum outsourcing ellipse of the tobacco leaf are taken as geometric characteristics.
And reducing the dimension of the tobacco leaf characteristics by using a PCA algorithm to obtain p tobacco leaf characteristics with the largest correlation with tobacco leaf grade separation.
As shown in fig. 3, feature fusion is performed on the feature of the tobacco leaves after dimension reduction and the feature of the tobacco leaves with high levels, and the method comprises the following steps: d1 Fixing the weight value of the trained convolutional neural network model;
d2 Removing the classifier of the convolutional neural network model;
d3 Combining the output of the last full-connection layer of the convolutional neural network model with the classifier removed with tobacco leaf features to form a row to obtain a total feature vector, and taking the total feature vector as the input of the fusion network.
And establishing a fusion network formed by full connection layers, wherein the fusion network comprises an input data layer and a full connection layer, the total feature vector is used as the input data layer of the fusion network, a random inactivation optimization method is adopted among the full connection layers, a classifier of the fusion network adopts a softmax classifier, and the last full connection layer of the fusion network is connected with the classifier.
E) And aggregating the output of the last full-connection layer into the total characteristics, and inputting the global characteristics into a classifier to obtain the tobacco leaf grade sorting result.
According to the invention, a machine vision technology is utilized to collect tobacco leaf images, color correction is carried out on the tobacco leaf images, a convolutional neural network model is constructed to obtain high-level tobacco leaf characteristics, color characteristics, texture characteristics and geometric characteristics of tobacco leaves are obtained, tobacco leaf characteristic fusion is carried out, global characteristics of tobacco leaves are constructed, and a fusion network is adopted to carry out tobacco leaf grade identification, so that identification accuracy is high.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (10)
1. The recognition method for online automatic tobacco grade sorting based on machine vision is characterized by comprising the following steps:
a) Tobacco pretreatment is carried out, tobacco leaves are spread and spread, and the tobacco leaves are sequentially and dispersedly arranged;
b) The method comprises the steps of collecting tobacco leaf images by using a machine vision sensor, preprocessing the tobacco leaf images, and taking the preprocessed tobacco leaf images as tobacco leaf samples;
c) Building a convolutional neural network model, training the convolutional neural network model by using tobacco samples to obtain high-level tobacco leaf characteristics, calculating a total sample loss function, and updating the weight value of the convolutional neural network model;
d) Extracting tobacco leaf characteristics, carrying out characteristic fusion on the tobacco leaf characteristics and high-level tobacco leaf characteristics, and establishing a fusion network formed by all connection layers, wherein the last all connection layer of the fusion network is connected with a classifier;
the extracted tobacco leaf features comprise color features, geometric features, integral texture features and tobacco leaf vein features;
e) And aggregating the output of the last full-connection layer into the total characteristics, and inputting the global characteristics into a classifier to obtain the tobacco leaf grade sorting result.
2. The machine vision-based on-line automatic tobacco grade sorting identification method as claimed in claim 1, wherein the preprocessing of the tobacco image in step B) includes color correction of the tobacco image by a color correction algorithm to obtain a corrected tobacco image I new =am×i, I represents the tobacco leaf image before correction, a is the color conversion matrix, and M is the color conversion model.
3. The machine vision-based identification method for online automatic tobacco leaf grade sorting according to claim 2, wherein the color correction algorithm is any one of a polynomial regression method, a backward propagation network method and a support vector regression method.
4. The machine vision-based on-line automatic tobacco leaf grade sorting identification method according to claim 3, wherein in the step C), the convolutional neural network model comprises n convolutional layers, m pooling layers and k full-connection layers, all pooling layers adopt average pooling, and the last full-connection layer is connected with a classifier.
5. The machine vision based on-line automatic tobacco leaf grade sorting identification method of claim 4, wherein in step C), further comprising calculating a total sample loss function, updating convolutional nervesA weight value of the network model; by calculation ofObtaining a loss function->A true value representing the j-th tobacco sample level, S j The j-th tobacco sample grade predicted value output by the classifier of the convolutional neural network model is represented, and l represents the number of tobacco grades; by calculating->Obtaining a total sample loss function L i (S, y) represents the loss function of the ith tobacco sample, and r represents the total number of tobacco samples.
6. The method for machine vision based on-line automatic tobacco leaf grade sorting identification of claim 5, wherein in step D), the extracted tobacco leaf features include color features, geometric features, overall texture features, and tobacco leaf venation texture features,
obtaining a color histogram of the tobacco sample by utilizing the tobacco sample, calculating the mean value, standard deviation, kurtosis and skewness of the color histogram, and taking the mean value, standard deviation, kurtosis and skewness of the color histogram as color characteristics;
the method comprises the steps of obtaining a gray level co-occurrence matrix of a tobacco sample by utilizing the tobacco sample, calculating ASM energy, contrast, inverse difference moment and entropy of the gray level co-occurrence matrix, and taking the ASM energy, contrast, inverse difference moment and entropy of the gray level co-occurrence matrix as integral texture features;
obtaining RGB images of tobacco samples, converting the RGB images into gray level images, carrying out Gaussian smoothing treatment on the gray level images, calculating second moments of the gray level images to obtain leaf vein lines, matching all leaf vein lines through a leaf vein image template, and counting the number, total length, maximum value and minimum value of the leaf vein lines as tobacco vein texture features;
the eccentricity, the slenderness, the rectangularity and the compactness of the minimum outsourcing ellipse of the tobacco leaf are taken as geometric characteristics.
7. The machine vision-based identification method for online automatic tobacco leaf grade sorting according to claim 5 or 6, wherein the step D) further comprises the step of reducing dimensions of the extracted tobacco leaf features by using a PCA algorithm to obtain p tobacco leaf features with the largest correlation with tobacco leaf grade sorting.
8. The machine vision-based identification method for online automatic tobacco leaf grade sorting according to claim 7, wherein the feature fusion of the tobacco leaf features and the high-level tobacco leaf features in the step D) comprises the steps of:
d1 Fixing the weight value of the trained convolutional neural network model;
d2 Removing the classifier of the convolutional neural network model;
d3 Combining the output of the last full-connection layer of the convolutional neural network model with the classifier removed and the extracted tobacco leaf features into a row to obtain a total feature vector, and taking the total feature vector as the input of a fusion network.
9. The machine vision-based on-line automatic tobacco leaf grade sorting identification method according to claim 8, wherein in the step D), the fusion network comprises an input data layer and a full-connection layer, the total feature vector is used as the input data layer of the fusion network, a random inactivation optimization method is adopted between the full-connection layer, and a linear support vector machine classifier or a softmax classifier is adopted as a classifier of the fusion network.
10. A machine vision based on-line automatic tobacco grade sorting identification method as claimed in claim 8 or 9 wherein the image in step B) is an RGB image or an HSV image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911087967.7A CN111274860B (en) | 2019-11-08 | 2019-11-08 | Recognition method for online automatic tobacco grade sorting based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911087967.7A CN111274860B (en) | 2019-11-08 | 2019-11-08 | Recognition method for online automatic tobacco grade sorting based on machine vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111274860A CN111274860A (en) | 2020-06-12 |
| CN111274860B true CN111274860B (en) | 2023-08-22 |
Family
ID=70998684
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911087967.7A Active CN111274860B (en) | 2019-11-08 | 2019-11-08 | Recognition method for online automatic tobacco grade sorting based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111274860B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112163527B (en) * | 2020-09-29 | 2022-06-14 | 华中科技大学 | Fusion model-based tobacco leaf baking state identification method, device and system |
| CN112529838A (en) * | 2020-11-05 | 2021-03-19 | 云南省烟草农业科学研究院 | Tobacco leaf maturity online judging method based on image processing technology |
| CN112539785B (en) * | 2020-12-11 | 2022-09-09 | 云南中烟工业有限责任公司 | Tobacco grade identification system and method based on multi-dimensional characteristic information |
| CN112704259B (en) * | 2020-12-11 | 2022-09-13 | 杭州安脉盛智能技术有限公司 | Threshing quality evaluation method and system for dense tobacco leaf recognition |
| CN112949663B (en) * | 2021-01-28 | 2024-01-23 | 贵州省烟草科学研究院 | Tobacco leaf baking control system and method based on image recognition |
| CN113919442B (en) * | 2021-02-24 | 2022-05-27 | 北京优创新港科技股份有限公司 | Tobacco maturity state identification method based on convolutional neural network |
| CN113222062A (en) * | 2021-05-31 | 2021-08-06 | 中国烟草总公司郑州烟草研究院 | Method, device and computer readable medium for tobacco leaf classification |
| CN113780306B (en) * | 2021-08-11 | 2024-04-09 | 北京工业大学 | Deep convolutional neural network-based waste mobile phone color recognition method |
| CN113781423B (en) * | 2021-09-02 | 2024-04-26 | 河南启迪睿视智能科技有限公司 | Tobacco leaf state detection method |
| CN114862858B (en) * | 2022-07-08 | 2022-11-11 | 湖北省烟草科学研究院 | Cigar harvesting maturity identification method and system based on ensemble learning |
| CN114972896A (en) * | 2022-08-03 | 2022-08-30 | 江苏布罗信息技术有限公司 | Automatic intelligent tobacco leaf identification method and tobacco leaf subpackaging system |
| CN115953384B (en) * | 2023-01-10 | 2024-02-02 | 杭州首域万物互联科技有限公司 | Online detection and prediction method for morphological parameters of tobacco leaves |
| CN116862456B (en) * | 2023-09-04 | 2023-12-19 | 山东一方制药有限公司 | Traditional Chinese medicine production monitoring control system and method based on image processing |
| CN117253024B (en) * | 2023-11-17 | 2024-02-06 | 山东海晟盐业有限公司 | Industrial salt quality inspection control method and system based on machine vision |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5303720A (en) * | 1989-05-22 | 1994-04-19 | R. J. Reynolds Tobacco Company | Smoking article with improved insulating material |
| CN106372648A (en) * | 2016-10-20 | 2017-02-01 | 中国海洋大学 | Multi-feature-fusion-convolutional-neural-network-based plankton image classification method |
| CN108805867A (en) * | 2018-05-25 | 2018-11-13 | 北京百度网讯科技有限公司 | Method and apparatus for exporting tobacco leaf degree information |
| CN109871885A (en) * | 2019-01-28 | 2019-06-11 | 南京林业大学 | A kind of plants identification method based on deep learning and Plant Taxonomy |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106326899A (en) * | 2016-08-18 | 2017-01-11 | 郑州大学 | Tobacco leaf grading method based on hyperspectral image and deep learning algorithm |
-
2019
- 2019-11-08 CN CN201911087967.7A patent/CN111274860B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5303720A (en) * | 1989-05-22 | 1994-04-19 | R. J. Reynolds Tobacco Company | Smoking article with improved insulating material |
| CN106372648A (en) * | 2016-10-20 | 2017-02-01 | 中国海洋大学 | Multi-feature-fusion-convolutional-neural-network-based plankton image classification method |
| CN108805867A (en) * | 2018-05-25 | 2018-11-13 | 北京百度网讯科技有限公司 | Method and apparatus for exporting tobacco leaf degree information |
| CN109871885A (en) * | 2019-01-28 | 2019-06-11 | 南京林业大学 | A kind of plants identification method based on deep learning and Plant Taxonomy |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111274860A (en) | 2020-06-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111274860B (en) | Recognition method for online automatic tobacco grade sorting based on machine vision | |
| CN109308697B (en) | Leaf disease identification method based on machine learning algorithm | |
| CN106295124B (en) | The method of a variety of image detecting technique comprehensive analysis gene subgraph likelihood probability amounts | |
| CN110479636B (en) | Method and device for automatically sorting tobacco leaves based on neural network | |
| CN108875821A (en) | The training method and device of disaggregated model, mobile terminal, readable storage medium storing program for executing | |
| CN101059425A (en) | Method and device for identifying different variety green tea based on multiple spectrum image texture analysis | |
| KR20200084940A (en) | CNN(Convolutional Neural Network) based pest and damage fruit classification device and method | |
| CN108734283A (en) | Nerve network system | |
| CN106874929B (en) | Pearl classification method based on deep learning | |
| CN109222914A (en) | The information collection of Chinese medicine facial diagnosis and intelligent processing method based on deep learning and multi-optical spectrum imaging technology | |
| CN110849828A (en) | Saffron crocus classification method based on hyperspectral image technology | |
| CN114359727A (en) | Tea disease identification method and system based on lightweight optimization Yolo v4 | |
| CN113077450B (en) | Cherry grading detection method and system based on deep convolutional neural network | |
| CN115994907B (en) | Intelligent processing system and method for comprehensive information of food detection mechanism | |
| CN109034184A (en) | A kind of grading ring detection recognition method based on deep learning | |
| CN116849612B (en) | Multispectral tongue picture image acquisition and analysis system | |
| CN117010971B (en) | Intelligent health risk providing method and system based on portrait identification | |
| CN117557840B (en) | Fundus lesion grading method based on small sample learning | |
| CN103235943A (en) | Principal component analysis-based (PCA-based) three-dimensional (3D) face recognition system | |
| CN117496325A (en) | Rice pest detection method and device based on improved YOLOv5s model | |
| CN115494009A (en) | Intelligent discrimination method and system for tobacco leaf maturity | |
| CN111601418A (en) | Color temperature adjusting method and device, storage medium and processor | |
| Jian et al. | Identification and grading of tea using computer vision | |
| CN116758589B (en) | Cattle face recognition method for processing gesture and visual angle correction | |
| CN118072251B (en) | Tobacco pest identification method, medium and system |
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 |