CN110555487B - Fresh tea leaf identification and classification method and system based on convolutional neural network - Google Patents
Fresh tea leaf identification and classification method and system based on convolutional neural network Download PDFInfo
- Publication number
- CN110555487B CN110555487B CN201910867669.3A CN201910867669A CN110555487B CN 110555487 B CN110555487 B CN 110555487B CN 201910867669 A CN201910867669 A CN 201910867669A CN 110555487 B CN110555487 B CN 110555487B
- Authority
- CN
- China
- Prior art keywords
- fresh tea
- processing
- neural network
- tea leaves
- convolutional neural
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- 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
-
- 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)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a fresh tea leaf identification and classification method and system based on a convolutional neural network, aiming at the requirement of accurate sorting of fresh tea leaves, the fresh tea leaf identification and classification method and system based on the convolutional neural network simulate an artificial identification mode through the convolutional neural network, and automatically learn the difference among different types of fresh tea leaves from complex data, so that the fresh tea leaf identification and classification method and system based on the convolutional neural network are suitable for the classification and identification of all types of fresh tea leaves; in addition, the fresh tea leaf identification and classification method and system can well identify the fresh tea leaf samples of each grade, the identification accuracy rate is usually not lower than 90%, the grade separation of different kinds of fresh tea leaves can be realized, and the identification accuracy rate of the grades of the fresh tea leaves can be further improved by optimizing the architecture and the structural parameters of the convolutional neural network.
Description
Technical Field
The invention relates to the technical field of tea production and processing, in particular to a fresh tea identification and classification method and system based on a convolutional neural network.
Background
At present, most of procedures in the production and processing processes of tea have been realized flow line mechanization, and a manufacturer can automatically complete different production steps of tea frying, drying, packaging and the like only by placing collected fresh tea in a corresponding flow production line. However, since the grade of the fresh tea leaves affects the grade of the tea leaves finally obtained, in order to realize accurate grading of the tea leaves, in the actual production process, the fresh tea leaves need to be identified and classified in advance to distinguish the fresh tea leaves with different leaf bud forms and leaf stalk forms.
Although the fresh tea leaf sorting technology enters an intelligent identification sorting stage, the fresh tea leaf sorting technology is mainly based on the geometric shape and color texture of a fresh tea leaf image and combines a neural network technology to more accurately sort out fresh tea leaves of various grades, the sorting algorithm of the neural network needs to be adaptively adjusted according to the fresh tea leaf varieties due to the large difference of the shapes and the textures of the fresh tea leaves of different varieties, and the intelligent fresh tea leaf sorting technology is poor in universality.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a fresh tea leaf identification and classification method and system based on a convolutional neural network, aiming at the requirement of accurate sorting of fresh tea leaves, the fresh tea leaf identification and classification method and system based on the convolutional neural network simulates an artificial identification mode through the convolutional neural network and automatically learns the difference between different types of fresh tea leaves from complex data, so that the fresh tea leaf identification and classification method and system are suitable for the classification and identification of all types of fresh tea leaves, and the fresh tea leaf identification and classification method and system also utilize a local connection and weight sharing mode between a convolutional layer and a pooling layer in the convolutional neural network to improve the optimized training degree of the convolutional neural network, thereby enabling the convolutional neural network to realize faster convergence and obtain higher learning efficiency in the training process; in addition, the fresh tea leaf identification and classification method and the fresh tea leaf identification and classification system can well identify the fresh tea leaf samples of each grade, the identification accuracy rate is usually not lower than 90%, the grade separation of different kinds of fresh tea leaves can be realized, and the identification accuracy rate of the grades of the fresh tea leaves can be further improved by optimizing the architecture and the structural parameters of the convolutional neural network.
The invention provides a fresh tea leaf identification and classification method based on a convolutional neural network, which is characterized by comprising the following steps of:
the method comprises the following steps of (S1), carrying out mechanical sorting treatment on fresh tea leaves so that the fresh tea leaves meet preset distribution conditions;
step (S2), collecting the image corresponding to the fresh tea meeting the preset distribution condition, and performing pre-analysis calculation processing on the image;
a step (S3) of performing deep learning processing on the image subjected to the pre-analysis calculation processing based on the convolutional neural network so as to obtain a type grade identification result of the fresh tea leaves;
further, the step (S1) of mechanically sorting the fresh tea leaves so that the fresh tea leaves satisfy a preset distribution condition specifically includes,
step (S101), after the fresh tea leaves are stirred and separated, the fresh tea leaves are subjected to rotary centrifugal treatment relative to a preset laying surface, so that the fresh tea leaves are flatly laid on the preset laying surface;
a step (S102) of acquiring a distribution image of the fresh tea leaves laid on the preset laying surface, and calculating and extracting the overlapping degree and/or the winding degree of the different corresponding fresh tea leaves in the distribution image;
step (S103), judging whether the fresh tea leaves meet the preset distribution condition according to the overlapping degree and/or the winding degree of the different fresh tea leaves, wherein if the overlapping degree and/or the winding degree are respectively smaller than the corresponding overlapping degree threshold value and/or the corresponding winding degree threshold value, the step (S2) is entered, otherwise, the step (S101) and the step (S102) are sequentially and repeatedly executed until the overlapping degree and/or the winding degree are respectively smaller than the corresponding overlapping degree threshold value and/or the corresponding winding degree threshold value;
further, in the step (S2), the collecting of the image corresponding to the fresh tea leaves satisfying the preset distribution condition and the pre-analysis calculation processing of the image specifically include,
step (S201), collecting a plurality of images of the fresh tea leaves at different azimuth angles, and carrying out normalization processing on the images at the different azimuth angles to obtain a plurality of normalized images;
a step (S202) of performing background noise suppression processing and RGB color discrimination processing on the plurality of normalized images to obtain a plurality of color corrected images;
a step (S203) of extracting corresponding HIS components from the plurality of color-corrected images respectively, and performing pixel segmentation processing and size conversion processing on the plurality of color-corrected images according to the HIS components to obtain input image information corresponding to the convolutional neural network;
further, in the step (S3), performing a deep learning process on the image subjected to the pre-analysis calculation process based on the convolutional neural network, thereby obtaining a type grade recognition result on the fresh tea leaves specifically includes,
step (S301), inputting image information obtained by the pre-analysis calculation processing of the image into the convolutional neural network, and performing secondary image feature extraction processing on the image information through the convolutional neural network;
step (S302), constructing a European spatial pixel distribution function related to the fresh tea leaves based on the image characteristic parameters obtained by the secondary image characteristic extraction processing;
step (S303), calculating at least one of the distribution state, the size distribution state and the condition distribution state of the leaf bud of the fresh tea leaves based on the Euclidean space pixel distribution function, and taking the distribution state as the type grade identification result;
further, before the step (S301), constructing a fresh tea image information training set and performing optimization training processing on the convolutional neural network based on the fresh tea image information training set, so that the convolutional neural network satisfies a preset learning rate condition;
alternatively, the first and second electrodes may be,
in the step (S301), the performing secondary image feature extraction processing on the image information by the convolutional neural network specifically includes,
a step (S3011) of extracting a convolution feature value concerning the image information by performing convolution processing on the image information by a convolution layer of the convolutional neural network;
a step (S3012) of performing local correlation operation processing on the convolution feature values through a pooling layer of the convolutional neural network to obtain the image feature parameters;
alternatively, the first and second electrodes may be,
in the step (S303), at least one of a leaf bud distribution status, a size distribution status and a facies distribution status of the fresh tea leaves is obtained by calculation based on the euclidean space pixel distribution function, which is specifically included as the type class identification result,
and calculating the distribution state of at least one of single bud, one bud and one leaf, one bud and two leaves, one bud and three leaves, bud size, leaf size, broken leaves and tea stems corresponding to the fresh tea leaves based on the European-style spatial pixel distribution function, and taking the distribution state as the type grade identification result.
The invention also provides a fresh tea leaf identification and classification system based on the convolutional neural network, which is characterized in that:
the fresh tea leaf identification and classification system based on the convolutional neural network comprises a mechanical sorting module, an image preprocessing module and a convolutional neural network processing module; wherein the content of the first and second substances,
the mechanical sorting module is used for carrying out mechanical sorting treatment on the fresh tea leaves so as to enable the fresh tea leaves to meet preset distribution conditions;
the image preprocessing module is used for acquiring images corresponding to the fresh tea leaves meeting the preset distribution conditions and performing pre-analysis calculation processing on the images;
the convolutional neural network processing module is used for carrying out deep learning processing on the image subjected to the pre-analysis calculation processing so as to obtain a type grade identification result of the fresh tea leaves;
further, the mechanical sorting module comprises a stirring and separating submodule, a rotating and centrifuging submodule, a distributed image processing submodule and a sorting control submodule; wherein the content of the first and second substances,
the stirring and separating sub-module is used for stirring and separating the fresh tea leaves;
the rotary centrifugal submodule is used for carrying out rotary centrifugal processing on the fresh tea leaves on a preset laying surface so as to enable the fresh tea leaves to be flatly laid on the preset laying surface;
the distribution image processing submodule is used for calculating and processing a distribution image of the fresh tea leaves laid on the preset laying surface so as to extract the overlapping degree and/or the winding degree of the different corresponding fresh tea leaves in the distribution image;
the sorting control sub-module is used for controlling the respective working states of the separation sub-module, the rotary centrifugation sub-module and the distributed image processing sub-module according to the relation between the overlapping degree and/or the winding degree and the corresponding overlapping degree threshold value and/or winding degree threshold value;
further, the image preprocessing module comprises a normalization processing sub-module, a noise suppression sub-module, a color distinguishing processing sub-module, an HIS component extraction sub-module and a pixel segmentation and size transformation sub-module; wherein the content of the first and second substances,
the normalization processing submodule is used for carrying out normalization processing on a plurality of images related to different azimuth angles of the fresh tea leaves so as to obtain a plurality of normalized images;
the noise suppression submodule and the color distinguishing processing submodule are used for respectively carrying out background noise suppression processing and RGB color distinguishing processing on the plurality of normalized images so as to obtain a plurality of color corrected images;
the HIS component extraction submodule is used for respectively extracting corresponding HIS components from the color corrected images;
the pixel segmentation and size transformation submodule is used for carrying out pixel segmentation processing and size transformation processing on the color correction images according to the HIS components so as to obtain input image information corresponding to the convolutional neural network;
further, the convolutional neural network processing module comprises a secondary image feature extraction processing submodule, a pixel distribution function construction submodule and a fresh tea type grade identification submodule; wherein the content of the first and second substances,
the secondary image feature extraction processing submodule is used for inputting image information obtained after the image is subjected to the pre-analysis calculation processing into the convolutional neural network and performing secondary image feature extraction processing on the image information through the convolutional neural network;
the pixel distribution function construction submodule is used for constructing an European space pixel distribution function related to the fresh tea according to the image characteristic parameters obtained through the secondary image characteristic extraction processing;
the fresh tea type grade identifier module is used for resolving at least one of a leaf bud distribution state, a size distribution state and a grade distribution state of the fresh tea according to the Euclidean space pixel distribution function, and taking the at least one of the leaf bud distribution state, the size distribution state and the grade distribution state as the type grade identification result;
further, the secondary image feature extraction processing submodule comprises a convolution operation unit and a local correlation operation unit; wherein the content of the first and second substances,
the convolution operation unit is used for performing convolution operation processing on the image information through a convolution layer of the convolution neural network so as to extract a convolution characteristic value related to the image information;
the local correlation operation unit is used for performing local correlation operation processing on the convolution characteristic value through a pooling layer of the convolution neural network so as to obtain the image characteristic parameter;
alternatively, the first and second electrodes may be,
the fresh tea type grade identifier module specifically calculates and obtains a distribution state of at least one of a single bud, a bud and a leaf, a bud and two leaves, a bud and three leaves, a bud size, a leaf size, a broken leaf and a tea stalk corresponding to the fresh tea according to the European space pixel distribution function, and the distribution state is used as the type grade identification result.
Compared with the prior art, the fresh tea leaf identification and classification method and system based on the convolutional neural network aim at the requirement of accurate sorting of fresh tea leaves, an artificial identification mode is simulated through the convolutional neural network, and differences among different types of fresh tea leaves are learned automatically from complex data, so that the fresh tea leaf identification and classification method and system are suitable for classification and identification of all types of fresh tea leaves, the optimum training degree of the convolutional neural network is improved by utilizing a local connection and weight sharing mode between a convolutional layer and a pooling layer in the convolutional neural network, and the convolutional neural network is enabled to achieve faster convergence and obtain higher learning efficiency in the training process; in addition, the fresh tea leaf identification and classification method and the fresh tea leaf identification and classification system can well identify the fresh tea leaf samples of each grade, the identification accuracy rate is usually not lower than 90%, the grade separation of different kinds of fresh tea leaves can be realized, and the identification accuracy rate of the grades of the fresh tea leaves can be further improved by optimizing the architecture and the structural parameters of the convolutional neural network.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a fresh tea leaf identification and classification method based on a convolutional neural network provided by the present invention.
Fig. 2 is a schematic structural diagram of a fresh tea leaf identification and classification system based on a convolutional neural network provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, a schematic flow chart of a fresh tea leaf identification and classification method based on a convolutional neural network according to an embodiment of the present invention is shown. The fresh tea leaf identification and classification method based on the convolutional neural network comprises the following steps:
and (S1) mechanically sorting the fresh tea leaves so that the fresh tea leaves meet preset distribution conditions.
Preferably, in the step (S1), the mechanically sorting the fresh tea leaves so that the fresh tea leaves satisfy the preset distribution condition specifically includes,
step (S101), after the fresh tea leaves are stirred and separated, the fresh tea leaves are subjected to rotary centrifugal treatment relative to a preset laying surface, so that the fresh tea leaves are flatly laid on the preset laying surface;
step (S102), collecting a distribution image of the fresh tea leaves laid on the preset laying surface, and calculating and extracting the overlapping degree and/or the winding degree of the different fresh tea leaves in the distribution image;
step (S103), judging whether the fresh tea leaves meet the preset distribution condition according to the overlapping degree and/or the winding degree of the different fresh tea leaves, wherein if the overlapping degree and/or the winding degree are respectively smaller than the corresponding overlapping degree threshold value and/or winding degree threshold value, the step (S2) is entered, otherwise, the step (S101) and the step (S102) are sequentially and repeatedly executed until the overlapping degree and/or the winding degree are respectively smaller than the corresponding overlapping degree threshold value and/or winding degree threshold value.
And (S2) acquiring an image corresponding to the fresh tea meeting the preset distribution condition, and performing pre-analysis calculation processing on the image.
Preferably, in the step (S2), the collecting the image corresponding to the fresh tea leaves satisfying the preset distribution condition, and the performing the pre-analysis calculation processing on the image specifically includes,
step (S201), collecting a plurality of images of the fresh tea leaves at different azimuth angles, and carrying out normalization processing on the images at the different azimuth angles to obtain a plurality of normalized images;
a step (S202) of performing background noise suppression processing and RGB color discrimination processing on the plurality of normalized images to obtain a plurality of color corrected images;
and (S203) extracting corresponding HIS components from the plurality of color corrected images respectively, and performing pixel segmentation processing and size conversion processing on the plurality of color corrected images according to the HIS components so as to obtain input image information corresponding to the convolutional neural network.
And (S3) performing deep learning processing on the image subjected to the pre-analysis calculation processing based on the convolutional neural network so as to obtain a type grade identification result of the fresh tea leaves.
Preferably, in the step (S3), performing deep learning processing on the image subjected to the pre-analysis calculation processing based on the convolutional neural network, thereby obtaining the type grade recognition result on the fresh tea leaves specifically includes,
step (S301), inputting the image information obtained by the pre-analysis calculation processing of the image into the convolutional neural network, and performing secondary image feature extraction processing on the image information through the convolutional neural network;
step (S302), constructing a European spatial pixel distribution function about the fresh tea leaves based on the image characteristic parameters obtained by the secondary image characteristic extraction processing;
and (S303) calculating at least one of the distribution state, the size distribution state and the condition distribution state of the leaf buds of the fresh tea leaves based on the Euclidean spatial pixel distribution function, and taking the calculated result as the type grade identification result.
Preferably, before the step (S301), constructing a fresh tea image information training set and performing optimization training processing on the convolutional neural network based on the fresh tea image information training set, so that the convolutional neural network satisfies a preset learning rate condition;
preferably, in the step (S301), the performing secondary image feature extraction processing on the image information by the convolutional neural network specifically includes,
a step (S3011) of extracting a convolution feature value relating to the image information by performing convolution processing on the image information by a convolution layer of the convolutional neural network;
a step (S3012) of performing local correlation operation processing on the convolution feature value through a pooling layer of the convolutional neural network to obtain the image feature parameter;
preferably, in the step (S303), at least one of a leaf bud distribution status, a size distribution status and a facies distribution status of the fresh tea leaves is obtained by calculation based on the euclidean space pixel distribution function, which specifically includes as the type class identification result,
and calculating the distribution state of at least one of single bud, one bud and one leaf, one bud and two leaves, one bud and three leaves, bud size, leaf size, broken leaves and tea stems corresponding to the fresh tea leaves based on the European space pixel distribution function, and taking the distribution state as the type grade identification result.
Fig. 2 is a schematic structural diagram of a fresh tea leaf identification and classification system based on a convolutional neural network according to an embodiment of the present invention. The fresh tea leaf identification and classification system based on the convolutional neural network comprises a mechanical sorting module, an image preprocessing module and a convolutional neural network processing module; wherein the content of the first and second substances,
the mechanical sorting module is used for carrying out mechanical sorting treatment on the fresh tea leaves so as to enable the fresh tea leaves to meet the preset distribution condition;
the image preprocessing module is used for acquiring an image corresponding to the fresh tea meeting the preset distribution condition and performing pre-analysis calculation processing on the image;
the convolution neural network processing module is used for carrying out deep learning processing on the image subjected to the pre-analysis calculation processing so as to obtain a type grade identification result of the fresh tea leaves.
Preferably, the mechanical sorting module comprises a stirring and separating submodule, a rotating and centrifuging submodule, a distributed image processing submodule and a sorting control submodule;
preferably, the stirring and separating sub-module is used for stirring and separating the fresh tea leaves;
preferably, the rotary centrifugal submodule is used for carrying out rotary centrifugal treatment on the fresh tea leaves on a preset laying surface, so that the fresh tea leaves are laid on the preset laying surface;
preferably, the distribution image processing sub-module is configured to perform calculation processing on the distribution image of the fresh tea leaves laid on the preset laying surface, so as to extract the overlapping degree and/or the winding degree of different fresh tea leaves in the distribution image;
preferably, the sorting control sub-module is configured to control respective working states of the separation sub-module, the rotation centrifugation sub-module and the distributed image processing sub-module according to a relationship between the overlapping degree and/or the winding degree and a corresponding overlapping degree threshold value and/or winding degree threshold value;
preferably, the image preprocessing module comprises a normalization processing submodule, a noise suppression submodule, a color distinguishing processing submodule, an HIS component extraction submodule and a pixel segmentation and size transformation submodule;
preferably, the normalization processing sub-module is used for performing normalization processing on a plurality of images related to different azimuth angles of the fresh tea leaves so as to obtain a plurality of normalized images;
preferably, the noise suppression submodule and the color discrimination processing submodule are configured to perform background noise suppression processing and RGB color discrimination processing on the plurality of normalized images, respectively, so as to obtain a plurality of color-corrected images;
preferably, the HIS component extraction sub-module is configured to extract corresponding HIS components from the several color-corrected images, respectively;
preferably, the pixel segmentation and size transformation submodule is configured to perform pixel segmentation processing and size transformation processing on the color-corrected images according to the HIS component, so as to obtain input image information corresponding to the convolutional neural network;
preferably, the convolutional neural network processing module comprises a secondary image feature extraction processing submodule, a pixel distribution function construction submodule and a fresh tea type grade identification submodule;
preferably, the secondary image feature extraction processing sub-module is configured to input image information obtained by performing the pre-analysis calculation processing on the image into the convolutional neural network, and perform secondary image feature extraction processing on the image information through the convolutional neural network;
preferably, the pixel distribution function construction sub-module is used for constructing a European spatial pixel distribution function related to the fresh tea according to the image characteristic parameters obtained by the secondary image characteristic extraction processing;
preferably, the fresh tea type grade identifier module is configured to obtain at least one of a leaf bud distribution state, a size distribution state and a phase distribution state of the fresh tea by calculation according to the european spatial pixel distribution function, and use the at least one of the leaf bud distribution state, the size distribution state and the phase distribution state as a type grade identification result;
preferably, the secondary image feature extraction processing submodule includes a convolution operation unit and a local correlation operation unit;
preferably, the convolution operation unit is configured to perform convolution operation processing on the image information through convolution layers of the convolutional neural network, so as to extract a convolution feature value related to the image information;
preferably, the local correlation operation unit is configured to perform local correlation operation processing on the convolution feature value through a pooling layer of the convolutional neural network, so as to obtain the image feature parameter;
preferably, the fresh tea type grade identifier module specifically calculates a distribution state of at least one of a single bud, a bud-leaf-two-leaf, a bud-three-leaf, a bud size, a leaf size, a broken leaf and a tea stem corresponding to the fresh tea according to the european spatial pixel distribution function, and uses the distribution state as the type grade identification result.
From the content of the embodiment, the fresh tea leaf identification and classification method and system based on the convolutional neural network aim at the requirement of accurate sorting of fresh tea leaves, simulate an artificial identification mode through the convolutional neural network, and automatically learn the difference between different types of fresh tea leaves from complex data, so that the fresh tea leaf identification and classification method and system are suitable for the classification and identification of all types of fresh tea leaves, and further utilize a local connection and weight sharing mode between a convolutional layer and a pooling layer in the convolutional neural network to improve the optimized training degree of the convolutional neural network, so that the convolutional neural network realizes faster convergence and obtains higher learning efficiency in the training process; in addition, the fresh tea leaf identification and classification method and the fresh tea leaf identification and classification system can well identify the fresh tea leaf samples of each grade, the identification accuracy rate is usually not lower than 90%, the grade separation of different kinds of fresh tea leaves can be realized, and the identification accuracy rate of the grades of the fresh tea leaves can be further improved by optimizing the architecture and the structural parameters of the convolutional neural network.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (4)
1. A fresh tea leaf identification and classification method based on a convolutional neural network is characterized by comprising the following steps:
the method comprises the following steps of (S1), carrying out mechanical sorting treatment on fresh tea leaves so that the fresh tea leaves meet preset distribution conditions;
step (S2), collecting the image corresponding to the fresh tea meeting the preset distribution condition, and performing pre-analysis calculation processing on the image;
a step (S3) of performing deep learning processing on the image subjected to the pre-analysis calculation processing based on the convolutional neural network, thereby obtaining a type grade identification result about the fresh tea leaves;
in the step (S1), the mechanically sorting fresh tea leaves so that the fresh tea leaves satisfy a preset distribution condition specifically includes,
a step (S101) of stirring and separating the fresh tea leaves, and then subjecting the fresh tea leaves to a rotary centrifugal treatment with respect to a preset laying surface, so that the fresh tea leaves are laid on the preset laying surface;
a step (S102) of acquiring a distribution image of the fresh tea leaves laid on the preset laying surface, and calculating and extracting the overlapping degree and/or the winding degree of the different corresponding fresh tea leaves in the distribution image;
step (S103), judging whether the fresh tea leaves meet the preset distribution condition according to the overlapping degree and/or the winding degree of the different fresh tea leaves, wherein if the overlapping degree and/or the winding degree are respectively smaller than the corresponding overlapping degree threshold value and/or the corresponding winding degree threshold value, the step (S2) is entered, otherwise, the step (S101) and the step (S102) are sequentially and repeatedly executed until the overlapping degree and/or the winding degree are respectively smaller than the corresponding overlapping degree threshold value and/or the corresponding winding degree threshold value;
in the step (S2), the collecting of the image corresponding to the fresh tea leaves satisfying the preset distribution condition and the pre-analyzing and calculating process of the image specifically include,
step (S201), collecting a plurality of images of the fresh tea leaves at different azimuth angles, and carrying out normalization processing on the images at the different azimuth angles to obtain a plurality of normalized images;
a step (S202) of performing background noise suppression processing and RGB color discrimination processing on the plurality of normalized images to obtain a plurality of color corrected images;
a step (S203) of extracting corresponding HIS components from the plurality of color-corrected images respectively, and performing pixel segmentation processing and size conversion processing on the plurality of color-corrected images according to the HIS components to obtain input image information corresponding to the convolutional neural network;
in the step (S3), performing deep learning processing on the image subjected to the pre-analysis calculation processing based on the convolutional neural network to obtain a type grade recognition result on the fresh tea leaves specifically includes,
step (S301), inputting image information obtained by the pre-analysis calculation processing of the image into the convolutional neural network, and performing secondary image feature extraction processing on the image information through the convolutional neural network;
step (S302), constructing a European spatial pixel distribution function related to the fresh tea leaves based on the image characteristic parameters obtained by the secondary image characteristic extraction processing;
and (S303) calculating at least one of a leaf bud distribution state, a size distribution state and a condition distribution state of the fresh tea leaves based on the Euclidean space pixel distribution function to serve as the type grade identification result.
2. The fresh tea leaf identification and classification method based on the convolutional neural network as claimed in claim 1, wherein:
before the step (S301), constructing a fresh tea image information training set and performing optimization training processing on the convolutional neural network based on the fresh tea image information training set, so that the convolutional neural network meets a preset learning rate condition;
alternatively, the first and second electrodes may be,
in the step (S301), the performing secondary image feature extraction processing on the image information by the convolutional neural network specifically includes,
a step (S3011) of extracting a convolution feature value concerning the image information by performing convolution processing on the image information by a convolution layer of the convolutional neural network;
a step (S3012) of performing local correlation operation processing on the convolution feature values through a pooling layer of the convolutional neural network to obtain the image feature parameters;
alternatively, the first and second electrodes may be,
in the step (S303), at least one of a leaf bud distribution status, a size distribution status and a facies distribution status of the fresh tea leaves is obtained by calculation based on the euclidean space pixel distribution function, which is specifically included as the type class identification result,
and calculating the distribution state of at least one of single bud, one bud and one leaf, one bud and two leaves, one bud and three leaves, bud size, leaf size, broken leaves and tea stems corresponding to the fresh tea leaves based on the European space pixel distribution function, and taking the distribution state as the type grade identification result.
3. The utility model provides a bright tealeaves discernment classification system based on convolutional neural network which characterized in that:
the fresh tea leaf identification and classification system based on the convolutional neural network comprises a mechanical sorting module, an image preprocessing module and a convolutional neural network processing module; wherein, the first and the second end of the pipe are connected with each other,
the mechanical sorting module is used for carrying out mechanical sorting treatment on the fresh tea leaves so as to enable the fresh tea leaves to meet preset distribution conditions;
the image preprocessing module is used for acquiring images corresponding to the fresh tea leaves meeting the preset distribution conditions and performing pre-analysis calculation processing on the images;
the convolutional neural network processing module is used for carrying out deep learning processing on the image subjected to the pre-analysis calculation processing so as to obtain a type grade identification result of the fresh tea leaves;
the mechanical sorting module comprises a stirring and separating submodule, a rotating and centrifuging submodule, a distributed image processing submodule and a sorting control submodule; wherein the content of the first and second substances,
the stirring and separating sub-module is used for stirring and separating the fresh tea;
the rotary centrifugal submodule is used for carrying out rotary centrifugal processing on the fresh tea leaves on a preset laying surface so as to enable the fresh tea leaves to be flatly laid on the preset laying surface;
the distribution image processing submodule is used for calculating and processing a distribution image of the fresh tea leaves laid on the preset laying surface so as to extract the overlapping degree and/or the winding degree of the different corresponding fresh tea leaves in the distribution image;
the sorting control sub-module is used for controlling the respective working states of the separation sub-module, the rotary centrifugation sub-module and the distributed image processing sub-module according to the relation between the overlapping degree and/or the winding degree and the corresponding overlapping degree threshold value and/or winding degree threshold value;
the image preprocessing module comprises a normalization processing submodule, a noise suppression submodule, a color distinguishing processing submodule, an HIS component extraction submodule and a pixel segmentation and size transformation submodule; wherein the content of the first and second substances,
the normalization processing submodule is used for carrying out normalization processing on a plurality of images related to different azimuth angles of the fresh tea leaves so as to obtain a plurality of normalized images;
the noise suppression submodule and the color distinguishing processing submodule are used for respectively carrying out background noise suppression processing and RGB color distinguishing processing on the plurality of normalized images so as to obtain a plurality of color corrected images;
the HIS component extraction submodule is used for respectively extracting corresponding HIS components from the color corrected images;
the pixel segmentation and size transformation submodule is used for carrying out pixel segmentation processing and size transformation processing on the color correction images according to the HIS components so as to obtain input image information corresponding to the convolutional neural network;
the convolutional neural network processing module comprises a secondary image feature extraction processing submodule, a pixel distribution function construction submodule and a fresh tea type grade identification submodule; wherein the content of the first and second substances,
the secondary image feature extraction processing submodule is used for inputting image information obtained after the image is subjected to the pre-analysis calculation processing into the convolutional neural network and performing secondary image feature extraction processing on the image information through the convolutional neural network;
the pixel distribution function construction submodule is used for constructing an European space pixel distribution function related to the fresh tea according to the image characteristic parameters obtained through the secondary image characteristic extraction processing;
the fresh tea type grade identification submodule is used for resolving and obtaining at least one of a leaf bud distribution state, a size distribution state and a grade distribution state of the fresh tea according to the Euclidean space pixel distribution function, and the at least one of the leaf bud distribution state, the size distribution state and the grade distribution state is used as the type grade identification result.
4. The convolutional neural network-based fresh tea leaf identification and classification system as claimed in claim 3, wherein:
the secondary image feature extraction processing submodule comprises a convolution operation unit and a local correlation operation unit; wherein the content of the first and second substances,
the convolution operation unit is used for performing convolution operation processing on the image information through a convolution layer of the convolution neural network so as to extract a convolution characteristic value related to the image information;
the local correlation operation unit is used for performing local correlation operation processing on the convolution characteristic value through a pooling layer of the convolution neural network so as to obtain the image characteristic parameter;
alternatively, the first and second electrodes may be,
the fresh tea type grade identifier module specifically calculates and obtains a distribution state of at least one of a single bud, a bud and a leaf, a bud and two leaves, a bud and three leaves, a bud size, a leaf size, a broken leaf and a tea stalk corresponding to the fresh tea according to the European space pixel distribution function, and the distribution state is used as the type grade identification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910867669.3A CN110555487B (en) | 2019-09-14 | 2019-09-14 | Fresh tea leaf identification and classification method and system based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910867669.3A CN110555487B (en) | 2019-09-14 | 2019-09-14 | Fresh tea leaf identification and classification method and system based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110555487A CN110555487A (en) | 2019-12-10 |
| CN110555487B true CN110555487B (en) | 2023-03-24 |
Family
ID=68740308
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910867669.3A Active CN110555487B (en) | 2019-09-14 | 2019-09-14 | Fresh tea leaf identification and classification method and system based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110555487B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111414971B (en) * | 2020-03-27 | 2023-05-02 | 南京工业大学 | Convolutional neural network-based finished tea variety and grade identification method |
| CN111389745B (en) * | 2020-04-07 | 2022-05-03 | 黎明职业大学 | Intelligent sorting method based on machine vision |
| CN112883842B (en) * | 2021-02-02 | 2022-12-16 | 四川省机械研究设计院(集团)有限公司 | Motorcycle engine assembling method and system based on mutual matching of parts and light source |
| CN112906833B (en) * | 2021-05-08 | 2021-08-17 | 武汉大学 | Plasma energy deposition image identification method based on full convolution neural network |
| CN113358153B (en) * | 2021-05-28 | 2022-06-24 | 揭阳市悦心茶业有限公司 | Tea frying mechanical rotation detection system |
| CN114985303B (en) * | 2022-05-24 | 2024-04-26 | 广东省农业科学院蔬菜研究所 | Black skin collybia albuminosa appearance quality grading system and method thereof |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105654468B (en) * | 2015-12-21 | 2019-03-22 | 中国石油大学(华东) | A kind of tealeaves tender shoots detection method based on deep learning and image edge information |
| CN106560691A (en) * | 2016-10-20 | 2017-04-12 | 中国计量大学 | Identification method for producing area of Wuyi rock tea and with deep learning function |
| CN110070069A (en) * | 2019-04-30 | 2019-07-30 | 重庆大学 | A kind of Classification of Tea method based on convolutional neural networks Automatic Feature Extraction |
-
2019
- 2019-09-14 CN CN201910867669.3A patent/CN110555487B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110555487A (en) | 2019-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110555487B (en) | Fresh tea leaf identification and classification method and system based on convolutional neural network | |
| CN106934418B (en) | Insulator infrared diagnosis method based on convolution recursive network | |
| CN104331710B (en) | On off state identifying system | |
| CN106971152A (en) | A kind of method of Bird's Nest in detection transmission line of electricity based on Aerial Images | |
| CN104766097B (en) | Surface of aluminum plate defect classification method based on BP neural network and SVMs | |
| CN104063686B (en) | Crop leaf diseases image interactive diagnostic system and method | |
| CN110807760B (en) | Tobacco leaf grading method and system | |
| CN109740721B (en) | Wheat ear counting method and device | |
| CN110646425B (en) | Tobacco leaf online auxiliary grading method and system | |
| CN104239899B (en) | A kind of power transmission line spacer recognition methods for unmanned plane inspection | |
| CN106157323A (en) | The insulator division and extracting method that a kind of dynamic division threshold value and block search combine | |
| CN112818827A (en) | Image recognition-based method for judging stage temperature control point in tobacco leaf baking process | |
| CN109886267A (en) | A kind of soft image conspicuousness detection method based on optimal feature selection | |
| CN103177257A (en) | Image identification method and image classification method for coprinus comatus | |
| CN114926511A (en) | High-resolution remote sensing image change detection method based on self-supervision learning | |
| CN107067022B (en) | Method, device and equipment for establishing image classification model | |
| CN110866547B (en) | Automatic classification system and method for traditional Chinese medicine decoction pieces based on multiple features and random forests | |
| Liu et al. | Development of a machine vision algorithm for recognition of peach fruit in a natural scene | |
| CN111046838A (en) | Method and device for identifying wetland remote sensing information | |
| CN113205006B (en) | Multi-temporal remote sensing image rice extraction method based on rice indexes | |
| WO2022222036A1 (en) | Method and apparatus for determining parking space | |
| CN112419278B (en) | Solid wood floor classification method based on deep learning | |
| CN111539931A (en) | Appearance abnormity detection method based on convolutional neural network and boundary limit optimization | |
| CN111968078A (en) | Appearance detection method, device, equipment and storage medium for power transformation equipment | |
| CN116092170A (en) | Patent value analysis system based on big data technology |
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 |