CN115861710B - Method and device for identifying wood tree species based on multi-source feature fusion - Google Patents
Method and device for identifying wood tree species based on multi-source feature fusion Download PDFInfo
- Publication number
- CN115861710B CN115861710B CN202211667095.3A CN202211667095A CN115861710B CN 115861710 B CN115861710 B CN 115861710B CN 202211667095 A CN202211667095 A CN 202211667095A CN 115861710 B CN115861710 B CN 115861710B
- Authority
- CN
- China
- Prior art keywords
- wood
- sample
- section
- characteristic
- dna sequence
- 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
- 239000002023 wood Substances 0.000 title claims abstract description 576
- 230000004927 fusion Effects 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 53
- 239000000126 substance Substances 0.000 claims abstract description 134
- 108091028043 Nucleic acid sequence Proteins 0.000 claims description 111
- 238000013136 deep learning model Methods 0.000 claims description 94
- 239000011159 matrix material Substances 0.000 claims description 68
- 238000012549 training Methods 0.000 claims description 58
- 238000001228 spectrum Methods 0.000 claims description 44
- 238000012360 testing method Methods 0.000 claims description 37
- 238000013527 convolutional neural network Methods 0.000 claims description 27
- 108020004414 DNA Proteins 0.000 claims description 25
- 150000001875 compounds Chemical class 0.000 claims description 24
- 238000000605 extraction Methods 0.000 claims description 15
- 210000003484 anatomy Anatomy 0.000 claims description 14
- 230000007246 mechanism Effects 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 12
- 238000011176 pooling Methods 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 10
- 238000004140 cleaning Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 8
- 238000002790 cross-validation Methods 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 238000001819 mass spectrum Methods 0.000 claims description 6
- 230000003321 amplification Effects 0.000 claims description 5
- 238000011156 evaluation Methods 0.000 claims description 5
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 230000002068 genetic effect Effects 0.000 abstract description 11
- 230000015654 memory Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 239000011148 porous material Substances 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 238000004252 FT/ICR mass spectrometry Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a wood tree species identification method and device based on multi-source feature fusion, and belongs to the technical field of wood identification. The invention simultaneously utilizes the structural characteristics, genetic characteristics and chemical characteristics of the wood to carry out tree species identification, covers various types of tree species information of the wood, solves the problems that the existing wood identification lacks reliable characteristic reference data sets, depends on single characteristics to carry out tree species identification results to be unreliable, is not suitable for complex application scenes and the like, realizes accurate identification of the wood species level, improves the identification precision on one hand, and improves the stability and the applicability of the identification results on the other hand.
Description
Technical Field
The invention relates to the technical field of wood identification, in particular to a wood tree species identification method and device based on multi-source feature fusion.
Background
The wood identification is to compare a sample to be detected with a wood specimen containing complete tree taxonomy and collected information, and give the tree species name of the sample to be detected based on the difference between the wood tree species. The traditional wood identification technology is used for observing the structural characteristics of a sample to be detected on macroscopic and microscopic scales by human eyes, mainly depends on the expertise and the identification experience of an identification person, the identification result is greatly influenced by subjective factors of the person, and in most cases, the traditional wood identification technology only can identify the wood to be in the category or the class, and the identification result of the category level cannot be obtained.
In order to realize accurate identification of the wood species level, new wood identification technologies based on computer technology have been developed in recent years, but although some new technologies are currently used for solving the problem of wood species identification, the existing new wood identification technologies provide species identification results based on single characteristics of anatomy, genetics, chemistry and the like, and the reliability of the identification results is not high due to a certain degree of variability of species of wood. And under different complex application scenes, partial characteristics are difficult to acquire due to the limitation of various conditions, and tree species identification by means of the single characteristics which cannot be acquired is often difficult to realize.
Disclosure of Invention
In order to solve the defects of the prior art, the invention provides the wood tree species identification method and device based on multi-source feature fusion, which realize accurate identification of wood species level, improve identification precision on one hand, and improve stability and applicability of identification results on the other hand.
The technical scheme provided by the invention is as follows:
a method for identifying wood species based on multi-source feature fusion, the method comprising:
establishing a wood classification characteristic reference data set, and dividing the wood classification characteristic reference data set into a training set and a testing set; the wood classification characteristic reference data set comprises a wood section image sample, a wood DNA sequence sample and a wood chemical fingerprint spectrum sample;
Constructing a deep learning model, and training the deep learning model by adopting the training set;
extracting key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample respectively by an autonomous learning method through the deep learning model;
fusing the extracted key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample, verifying the fused result by using the test set, and adjusting and optimizing the deep learning model;
and acquiring a wood section image, a wood DNA sequence and/or a wood chemical fingerprint of the wood sample to be identified, and inputting the wood section image, the wood DNA sequence and/or the wood chemical fingerprint into the deep learning model to obtain the tree species name of the wood sample to be identified.
Further, the building of the wood classification characteristic reference data set and the division into the training set and the testing set includes:
acquiring a timber cross section, a diameter section and a chord section structural image acquired from a timber specimen, and performing data enhancement treatment to obtain a timber section image sample;
obtaining DNA extracted from the wood specimen, performing amplification, sequencing and DNA bar code evaluation, and screening out an effective DNA bar code sequence to obtain a wood DNA sequence sample;
Acquiring mass spectrum data acquired by scanning the surface of the wood sample by adopting a mass spectrometer, and carrying out normalization treatment to obtain the wood chemical fingerprint sample;
performing data cleaning on the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample to remove abnormal values, and establishing the wood classification characteristic reference data set;
and (3) adopting a k-fold cross validation method to divide the wood classification characteristic reference data set into 8: the ratio of 2 is divided into the training set and the test set.
Further, the constructing the deep learning model, and training the deep learning model by using the training set includes:
constructing a deep learning model based on convolutional neural networks, wherein the deep learning model comprises three convolutional neural networks in parallel, and each convolutional neural network comprises an input layer, a convolutional layer, an attention mechanism layer, a pooling layer and a full connection layer;
and respectively inputting the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample of the training set into three convolutional neural networks of the deep learning model for training, extracting features at the convolutional layer, the attention mechanism layer and the pooling layer, and classifying the features through the full connection layer.
Further, the extracting key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample by the deep learning model through an autonomous learning method includes:
extracting the timber anatomy feature information contained in the timber cross section, diameter section and chord section structure images respectively, and carrying out matrixing representation to obtain a key feature matrix of the timber section image sample;
the base sequence of the screened effective DNA bar code sequence is matrixed by adopting a k-mer algorithm to obtain a key feature matrix of the wood DNA sequence sample;
based on the wood chemical fingerprint sample, obtaining the structure and content information of wood characteristic compounds, and carrying out matrixing representation on the molecular weight of each characteristic compound to obtain a key characteristic matrix of the wood chemical fingerprint sample.
Further, the timber cross section, the diametral section and the chord section structure image comprise a macroscopic structure image and a microscopic structure image, and the data enhancement processing comprises image rotation, image scaling, image mirroring and/or image clipping;
when the wood characteristic compound is obtained, peak alignment treatment is carried out on the wood chemical fingerprint sample in a mode of traversing the charge-to-mass ratio data, and the characteristic compound is searched out by comparing the charge-to-mass ratio.
Further, the wood anatomy feature information includes cross section feature information extracted from the wood cross section configuration image, diametral section feature information extracted from the wood diametral section configuration image, and chord section feature information extracted from the wood chord section configuration image;
the cross-section characteristic information comprises a tube hole frequency characteristic, a tube hole diameter characteristic and/or an axial parenchyma frequency characteristic, the diameter section characteristic information comprises a wood ray cell type characteristic and/or a catheter-ray interline hole characteristic, and the chord section characteristic information comprises a wood ray width characteristic, a wood ray height characteristic and/or a wood ray frequency characteristic.
Further, the fusing the extracted key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample, verifying the fused result by using the test set, and adjusting and optimizing the deep learning model, including:
normalizing key feature matrixes of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample;
fusing key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix;
And outputting an identification result based on the fusion matrix, testing and parameter adjustment are carried out on the deep learning model by utilizing the test set, and each parameter of the deep learning model is adjusted and optimized according to the identification result, so that the classification precision of the deep learning model reaches more than 99%.
The method for fusing the key feature matrix of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample comprises the following steps:
superposing the characteristic values with the same number of rows and columns in the key characteristic matrix of the normalized wood section image sample, the normalized wood DNA sequence sample and the normalized wood chemical fingerprint sample to obtain the fusion matrix;
or alternatively;
and performing matrix splicing operation on the key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain the fusion matrix.
Further, the obtaining a wood section image, a wood DNA sequence and/or a wood chemical fingerprint of the wood sample to be identified, inputting the wood section image, the wood DNA sequence and/or the wood chemical fingerprint into the deep learning model to obtain a tree species name of the wood sample to be identified, including:
According to the actual application scene and the acquirable condition of the data of the wood sample to be identified, acquiring one or more of a wood section image, a wood DNA sequence and a wood chemical fingerprint of the wood sample to be identified;
inputting one or more of the obtained wood section images, the wood DNA sequences and the wood chemical fingerprints of the wood samples to be identified into the deep learning model, and outputting the tree species names of the wood samples to be identified.
A wood species identification device based on multi-source feature fusion, the device comprising:
the data set establishing module is used for establishing a wood classification characteristic reference data set and dividing the wood classification characteristic reference data set into a training set and a testing set; the wood classification characteristic reference data set comprises a wood section image sample, a wood DNA sequence sample and a wood chemical fingerprint spectrum sample;
the model construction module is used for constructing a deep learning model and training the deep learning model by adopting the training set;
the feature extraction module is used for respectively extracting key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample by an autonomous learning method through the deep learning model;
The model optimization module is used for fusing the extracted key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample, verifying the fused result by using the test set and adjusting and optimizing the deep learning model;
the tree species identification module is used for acquiring a wood section image, a wood DNA sequence and/or a wood chemical fingerprint of the wood sample to be identified, and inputting the wood section image, the wood DNA sequence and/or the wood chemical fingerprint into the deep learning model to obtain the tree species name of the wood sample to be identified.
Further, the data set establishing module includes:
the wood section image acquisition unit is used for acquiring wood cross section, diameter section and chord section structural images acquired from a wood sample, and carrying out data enhancement treatment to obtain a wood section image sample;
the wood DNA sequence acquisition unit is used for acquiring DNA extracted from the wood specimen, performing amplification, sequencing and DNA bar code evaluation, and screening out an effective DNA bar code sequence to obtain the wood DNA sequence sample;
the wood chemical fingerprint acquisition unit is used for acquiring mass spectrum data acquired by scanning the surface of the wood sample by adopting a mass spectrometer and carrying out normalization treatment to obtain the wood chemical fingerprint sample;
The data cleaning unit is used for cleaning the data of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample, removing abnormal values and establishing the wood classification characteristic reference data set;
the data set dividing unit is used for adopting a k-fold cross validation method to divide the wood classification characteristic reference data set into 8: the ratio of 2 is divided into the training set and the test set.
Further, the model building module includes:
the model construction unit is used for constructing a deep learning model based on convolutional neural networks, the deep learning model comprises three convolutional neural networks in parallel, and each convolutional neural network comprises an input layer, a convolutional layer, an attention mechanism layer, a pooling layer and a full connection layer;
the model training unit is used for respectively inputting the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample of the training set into three convolutional neural networks of the deep learning model for training, extracting features at the convolutional layer, the attention mechanism layer and the pooling layer, and classifying the features through the fully connected layer.
Further, the feature extraction module includes:
The first extraction unit is used for respectively extracting the wood anatomy feature information contained in the wood cross section, diameter section and chord section structural image and carrying out matrixing representation to obtain a key feature matrix of the wood section image sample;
the second extraction unit is used for matrixing the base sequence of the screened effective DNA barcode sequence by adopting a k-mer algorithm to obtain a key feature matrix of the wood DNA sequence sample;
and the third extraction unit is used for acquiring the structure and content information of the wood characteristic compounds based on the wood chemical fingerprint sample, and carrying out matrixing representation on the molecular weight of each characteristic compound to obtain a key characteristic matrix of the wood chemical fingerprint sample.
Further, the timber cross section, the diametral section and the chord section structure image comprise a macroscopic structure image and a microscopic structure image, and the data enhancement processing comprises image rotation, image scaling, image mirroring and/or image clipping;
when the wood characteristic compound is obtained, peak alignment treatment is carried out on the wood chemical fingerprint sample in a mode of traversing the charge-to-mass ratio data, and the characteristic compound is searched out by comparing the charge-to-mass ratio.
Further, the wood anatomy feature information includes cross section feature information extracted from the wood cross section configuration image, diametral section feature information extracted from the wood diametral section configuration image, and chord section feature information extracted from the wood chord section configuration image;
the cross-section characteristic information comprises a tube hole frequency characteristic, a tube hole diameter characteristic and/or an axial parenchyma frequency characteristic, the diameter section characteristic information comprises a wood ray cell type characteristic and/or a catheter-ray interline hole characteristic, and the chord section characteristic information comprises a wood ray width characteristic, a wood ray height characteristic and/or a wood ray frequency characteristic.
Further, the model optimization module includes:
the normalization unit is used for normalizing the key feature matrixes of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample;
the feature fusion unit is used for fusing the key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix;
the model optimization unit is used for outputting an identification result based on the fusion matrix, testing and parameter adjustment are carried out on the deep learning model by utilizing the test set, and each parameter of the deep learning model is adjusted and optimized according to the identification result, so that the classification precision of the deep learning model reaches more than 99%.
Further, the feature fusion unit is configured to:
superposing the characteristic values with the same number of rows and columns in the key characteristic matrix of the normalized wood section image sample, the normalized wood DNA sequence sample and the normalized wood chemical fingerprint sample to obtain the fusion matrix;
or alternatively;
and performing matrix splicing operation on the key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain the fusion matrix.
Further, the tree species identification module includes:
the data acquisition unit is used for acquiring one or more of a wood section image, a wood DNA sequence and a wood chemical fingerprint of the wood sample to be identified according to the actual application scene and the data acquirable condition of the wood sample to be identified;
the tree species identification unit is used for inputting one or more of the obtained wood section images, the wood DNA sequences and the wood chemical fingerprints of the wood samples to be identified into the deep learning model and outputting the tree species names of the wood samples to be identified.
The invention has the following beneficial effects:
according to the method, a wood classification characteristic reference data set of a wood section image sample, a wood DNA sequence sample and a wood chemical fingerprint sample is established by collecting biological information such as structural characteristics, genetic characteristics, chemical characteristics and the like of a wood sample; and constructing a deep learning model, fusing the data information of the anatomical, genetic and chemical classification characteristics of the wood, taking three types of data in the reference data set of the wood classification characteristics as a training set, extracting key characteristics in the three data types by an autonomous learning method, and training to obtain the deep learning model capable of autonomously identifying wood tree types. On the basis, any one or more types of combination of a wood section image, a wood DNA sequence and a wood chemical fingerprint of a wood sample to be identified can be input into a trained deep learning model for tree species identification, and the model automatically gives the tree species name of the sample to be identified.
The invention simultaneously utilizes the structural characteristics, genetic characteristics and chemical characteristics of the wood to carry out tree species identification, covers various types of tree species information of the wood, solves the problems that the existing wood identification lacks reliable characteristic reference data sets, depends on single characteristics to carry out tree species identification results to be unreliable, is not suitable for complex application scenes and the like, realizes accurate identification of the wood species level, improves the identification precision on one hand, and improves the stability and the applicability of the identification results on the other hand.
Drawings
FIG. 1 is a flow chart of a method for identifying wood species based on multi-source feature fusion of the present invention;
fig. 2 is a schematic diagram of a wood species identification device based on multi-source feature fusion according to the present invention.
Detailed Description
In order to make the technical problems, technical solutions and advantages to be solved more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present invention.
The embodiment of the invention provides a wood tree species identification method based on multi-source feature fusion, which comprises the following steps of:
s100: establishing a wood classification characteristic reference data set, and dividing the wood classification characteristic reference data set into a training set and a testing set; the wood classification characteristic reference data set comprises a wood section image sample, a wood DNA sequence sample and a wood chemical fingerprint spectrum sample.
In the step, a wood sample is taken as a library building sample, a wood classification characteristic reference data set comprising a series of wood section image samples, wood DNA sequence samples and wood chemical fingerprint spectrum samples is built, and three types of sample data in the wood classification characteristic reference data set are all from accurately named wood samples and have accurate tree species marks.
S200: and constructing a deep learning model, and training the deep learning model by adopting a training set.
The deep learning model may be a model based on a convolutional neural network and having a layer of attention mechanisms, the deep learning model being trained by a training set.
S300: and respectively extracting key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample by an autonomous learning method through a deep learning model.
The key characteristics of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample are wood structural characteristics, wood genetic characteristics and wood chemical characteristics respectively.
S400: and fusing key features of the extracted wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample, verifying a fused result by using a test set, and adjusting and optimizing the deep learning model.
The key features can be represented in a matrix form, and can be fused in a matrix form when fusion is performed, and verification is performed by using a test set, so that a deep learning model capable of automatically identifying wood tree species is obtained.
S500: and (3) acquiring a wood section image, a wood DNA sequence and/or a wood chemical fingerprint of the wood sample to be identified, and inputting the wood section image, the wood DNA sequence and/or the wood chemical fingerprint into a deep learning model to obtain the tree species name of the wood sample to be identified.
After the deep learning model is trained and optimized, the tree species names can be identified based on the deep learning model, one or more of a wood section image, a wood DNA sequence and a wood chemical fingerprint can be acquired according to conditions for a wood sample to be identified, and the tree species names of the wood sample to be identified can be automatically output by inputting the wood section image, the wood DNA sequence and the wood chemical fingerprint into the deep learning model.
The prior art does not establish a reliable multisource classification characteristic fusion reference data set, cannot integrate comprehensive and complete classification characteristic information of wood, has low reliability of identification results, and is not suitable for diversified complex application scenes.
According to the method, a wood classification characteristic reference data set of a wood section image sample, a wood DNA sequence sample and a wood chemical fingerprint sample is established by collecting biological information such as structural characteristics, genetic characteristics, chemical characteristics and the like of a wood sample; and constructing a deep learning model, fusing the data information of the anatomical, genetic and chemical classification characteristics of the wood, taking three types of data in the reference data set of the wood classification characteristics as a training set, extracting key characteristics in the three data types by an autonomous learning method, and training to obtain the deep learning model capable of autonomously identifying wood tree types. On the basis, any one or more types of combination of a wood section image, a wood DNA sequence and a wood chemical fingerprint of a wood sample to be identified can be input into a trained deep learning model for tree species identification, and the model automatically gives the tree species name of the sample to be identified.
The invention simultaneously utilizes the structural characteristics, genetic characteristics and chemical characteristics of the wood to carry out tree species identification, covers various types of tree species information of the wood, solves the problems that the existing wood identification lacks reliable characteristic reference data sets, depends on single characteristics to carry out tree species identification results to be unreliable, is not suitable for complex application scenes and the like, realizes accurate identification of the wood species level, improves the identification precision on one hand, and improves the stability and the applicability of the identification results on the other hand.
As an improvement of the embodiment of the present invention, the foregoing S100 includes:
s110: and acquiring structural images of a cross section, a diametral section and a chord section of the wood acquired from the wood specimen, and performing data enhancement treatment to obtain a wood section image sample.
For example, an iWood wood image acquisition device can be used for acquiring structural images of transverse, radial and chord three-section surfaces from a wood specimen, or an optical microscope can be used for acquiring structural images of transverse, radial and chord three-section surfaces from the wood specimen, and the acquired structural images of the transverse, radial and chord section surfaces of the wood comprise a macroscopic structural image and a microscopic structural image.
The data enhancement processing comprises image rotation, image scaling, image mirroring and/or image clipping operations on the image, so that more image information is obtained, and sample diversity of the wood classification characteristic reference data set is enriched.
S120: DNA extracted from the wood specimen is obtained, amplified, sequenced and DNA bar code evaluated, and effective DNA bar code sequences (namely DNA base sequences) are screened out to obtain wood DNA sequence samples.
S130: and acquiring mass spectrum data acquired by scanning the surface of the wood sample by adopting a mass spectrometer, and carrying out normalization treatment to obtain a wood chemical fingerprint spectrum sample.
The mass spectrometer may employ a high resolution mass spectrometer such as a real-time direct analysis-fourier transform ion cyclotron resonance mass spectrometer (DART-FTICR-MS). The normalization processing method may be alignment processing or other processing modes.
S140: and (3) performing data cleaning on the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample, removing abnormal values, and establishing a wood classification characteristic reference data set.
S150: adopting a k-fold cross validation method, and taking the reference data set of the wood classification characteristics as 8: the scale of 2 is divided into training and test sets.
The k-fold cross validation firstly divides the reference data set into k mutually exclusive subsets with the same size, k-1 is randomly selected as a training set each time, and the rest 1 is used as a testing set to carry out one round of training. After this round of training is completed, k-1 is selected again as the training set, and the remaining 1 is used as the test set for one round of training. After several rounds of training, the loss function is selected to evaluate the optimal model and parameters.
Further, the S200 may include:
s210: and constructing a deep learning model based on convolutional neural networks, wherein the deep learning model comprises three convolutional neural networks in parallel, and each convolutional neural network comprises an input layer, a convolutional layer, an attention mechanism layer, a pooling layer and a full connection layer.
The three parallel convolutional neural networks are convolutional neural networks based on images, genetic information and chemical fingerprints respectively.
S220: and respectively inputting the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample of the training set into three convolutional neural networks of a deep learning model for training, extracting features in a convolutional layer, an attention mechanism layer and a pooling layer, and classifying the features through the full connection layer.
As one preferable aspect, the S300 includes:
s310: and respectively extracting the timber anatomy feature information contained in the timber cross section, diameter section and chord section structural images, and carrying out matrixing representation to obtain a key feature matrix of the timber section image sample.
The key feature of the wood section image sample is the matrixing representation of the wood anatomy feature information, namely the wood structure feature matrix, which can also be called as the wood image feature. The method for acquiring the characteristic information of the wood anatomy structure comprises the following steps:
1. and extracting cross section characteristic information from the wood cross section structural image, wherein the cross section characteristic information comprises pore frequency characteristics, pore diameter characteristics, axial parenchyma frequency characteristics and the like.
2. And extracting the diametral section characteristic information from the diametral section structural image of the wood, wherein the diametral section characteristic information comprises wood ray cell type characteristics and/or catheter-ray interline hole characteristics and the like.
3. And extracting chord-section characteristic information from the wood chord-section structural image, wherein the chord-section characteristic information comprises wood ray width characteristics, wood ray height characteristics, wood ray frequency characteristics and/or the like.
S320: and (3) matrixing the base sequence of the screened effective DNA barcode sequence by adopting a k-mer algorithm to obtain a key feature matrix of the wood DNA sequence sample.
The key feature matrix of the wood DNA sequence sample is the wood genetic feature.
S330: based on the wood chemical fingerprint sample, obtaining the structure and content information of the wood characteristic compounds, and carrying out matrixing representation on the molecular weight of each characteristic compound to obtain a key characteristic matrix of the wood chemical fingerprint sample.
The key feature matrix of the wood chemical fingerprint spectrum sample is the wood chemical feature, and is the matrixing representation of the wood feature compound. When the wood characteristic compound is obtained, peak alignment treatment can be carried out on the wood chemical fingerprint sample in a mode of traversing the charge-to-mass ratio data, and the characteristic compound is searched out by comparing the charge-to-mass ratio.
As another improvement of the embodiment of the present invention, the foregoing S400 includes:
s410: and carrying out normalization treatment on key feature matrixes of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample.
The normalization process allows the matrices of the three data types to have the same dimensional format and preserve the respective important features.
S420: and fusing the key feature matrix of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix.
The fusion mode can be linear weighted fusion or matrix splicing fusion, and finally a wood classification feature fusion matrix is obtained, wherein:
the linear weighted fusion includes: and superposing the characteristic values with the same number of rows and columns in the key characteristic matrix of the normalized wood section image sample, the normalized wood DNA sequence sample and the normalized wood chemical fingerprint sample to obtain a fusion matrix.
The matrix splicing and fusion comprises the following steps: and performing matrix splicing operation on key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix.
For example, the tf.concat function may be used to splice matrices of three data types after unifying dimensions to obtain a fusion matrix.
S430: and outputting an identification result at a full-connection layer of the deep learning model based on the fusion matrix, testing and parameter adjustment are carried out on the deep learning model by utilizing a test set, and each parameter of the deep learning model is adjusted and optimized according to the identification result, so that the classification precision of the deep learning model reaches more than 99%.
After the training process is performed, the step S500 may include:
s510: according to the actual application scene and the acquisition condition of the data of the wood sample to be identified, one or more of a wood section image, a wood DNA sequence and a wood chemical fingerprint of the wood sample to be identified are acquired.
S520: inputting the obtained one or more wood section images, wood DNA sequences and wood chemical fingerprints of the wood sample to be identified into a deep learning model, and outputting the tree species name of the wood sample to be identified.
The embodiment of the invention also provides a wood tree species identification device based on multi-source feature fusion, as shown in fig. 2, the device comprises:
the data set establishing module 1 is used for establishing a wood classification characteristic reference data set and dividing the wood classification characteristic reference data set into a training set and a testing set; the wood classification characteristic reference data set comprises a wood section image sample, a wood DNA sequence sample and a wood chemical fingerprint spectrum sample.
The model construction module 2 is used for constructing a deep learning model and training the deep learning model by adopting a training set.
And the feature extraction module 3 is used for respectively extracting key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample by an autonomous learning method through a deep learning model.
And the model optimization module 4 is used for fusing key features of the extracted wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample, verifying the fused result by using a test set and adjusting and optimizing the deep learning model.
The tree species identification module 5 is used for acquiring a wood section image, a wood DNA sequence and/or a wood chemical fingerprint of the wood sample to be identified, inputting the wood section image, the wood DNA sequence and/or the wood chemical fingerprint into the deep learning model, and obtaining the tree species name of the wood sample to be identified.
The invention simultaneously utilizes the structural characteristics, genetic characteristics and chemical characteristics of the wood to carry out tree species identification, covers various types of tree species information of the wood, solves the problems that the existing wood identification lacks reliable characteristic reference data sets, depends on single characteristics to carry out tree species identification results to be unreliable, is not suitable for complex application scenes and the like, realizes accurate identification of the wood species level, improves the identification precision on one hand, and improves the stability and the applicability of the identification results on the other hand.
As an improvement of the embodiment of the present invention, the data set creating module includes:
the wood section image acquisition unit is used for acquiring the wood cross section, the diameter section and the chord section structural image acquired from the wood specimen, and carrying out data enhancement processing to obtain a wood section image sample.
Wherein, the wood cross section, the diametral section and the chord section structure images comprise a macroscopic structure image and a microscopic structure image, and the data enhancement processing comprises image rotation, image scaling, image mirroring and/or image clipping;
and the wood DNA sequence acquisition unit is used for acquiring DNA extracted from the wood specimen, performing amplification, sequencing and DNA bar code evaluation, and screening out an effective DNA bar code sequence to obtain a wood DNA sequence sample.
The wood chemical fingerprint acquisition unit is used for acquiring mass spectrum data acquired by scanning the surface of a wood sample by using a mass spectrometer and carrying out normalization treatment to obtain a wood chemical fingerprint sample.
The data cleaning unit is used for cleaning the data of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample, removing abnormal values and establishing a wood classification characteristic reference data set.
The data set dividing unit is used for adopting a k-fold cross validation method to divide the wood classification characteristic reference data set into 8: the scale of 2 is divided into training and test sets.
Further, the model building module includes:
the model construction unit is used for constructing a deep learning model based on convolutional neural networks, wherein the deep learning model comprises three convolutional neural networks in parallel, and each convolutional neural network comprises an input layer, a convolutional layer, an attention mechanism layer, a pooling layer and a full connection layer.
The model training unit is used for respectively inputting the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample of the training set into three convolutional neural networks of the deep learning model for training, extracting features in a convolutional layer, an attention mechanism layer and a pooling layer, and classifying the features through the full connection layer.
Preferably, the feature extraction module includes:
the first extraction unit is used for respectively extracting the timber anatomy feature information contained in the timber cross section, diameter section and chord section structure images, and carrying out matrixing representation to obtain a key feature matrix of the timber section image sample.
Wherein the timber anatomy feature information comprises cross section feature information extracted from a timber cross section construction image, diametral section feature information extracted from a timber diametral section construction image and chord section feature information extracted from a timber chord section construction image.
The cross-section characteristic information comprises a tube hole frequency characteristic, a tube hole diameter characteristic and/or an axial parenchyma frequency characteristic, the diameter section characteristic information comprises a wood ray cell type characteristic and/or a catheter-ray interline hole characteristic, and the chord section characteristic information comprises a wood ray width characteristic, a wood ray height characteristic and/or a wood ray frequency characteristic.
And the second extraction unit is used for matrixing the base sequence of the screened effective DNA barcode sequence by adopting a k-mer algorithm to obtain a key feature matrix of the wood DNA sequence sample.
And the third extraction unit is used for acquiring the structure and content information of the wood characteristic compounds based on the wood chemical fingerprint sample, and carrying out matrixing representation on the molecular weight of each characteristic compound to obtain a key characteristic matrix of the wood chemical fingerprint sample.
The peak alignment treatment can be carried out on the wood chemical fingerprint spectrum sample by traversing the charge-to-mass ratio data, and the characteristic compound is searched out by comparing the charge-to-mass ratio.
As another improvement of the embodiment of the present invention, the model optimization module includes:
the normalization unit is used for normalizing key feature matrixes of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample.
And the feature fusion unit is used for fusing the key feature matrix of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix.
The operation of the key feature matrix fusion can be as follows:
superposing the characteristic values with the same number of rows and columns in the key characteristic matrix of the normalized wood section image sample, the normalized wood DNA sequence sample and the normalized wood chemical fingerprint sample to obtain a fusion matrix;
or alternatively;
and performing matrix splicing operation on key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix.
The model optimizing unit is used for outputting an identification result based on the fusion matrix, testing and parameter adjustment are carried out on the deep learning model by utilizing the test set, and each parameter of the deep learning model is adjusted and optimized according to the identification result, so that the classification precision of the deep learning model reaches more than 99%.
After the training process is completed, the tree species identification module includes:
the data acquisition unit is used for acquiring one or more of a wood section image, a wood DNA sequence and a wood chemical fingerprint of the wood sample to be identified according to the actual application scene and the data acquirable condition of the wood sample to be identified.
The tree species identification unit is used for inputting the acquired one or more wood section images, wood DNA sequences and wood chemical fingerprints of the wood sample to be identified into the deep learning model and outputting the tree species name of the wood sample to be identified.
The device provided in the foregoing embodiments has a one-to-one correspondence between its implementation principle and the technical effects that are produced and the embodiments of the foregoing methods, and for a brief description, reference may be made to the corresponding matters in the embodiments of the foregoing methods where no mention is made in the examples of the device. It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of modules and units described in the apparatus may refer to corresponding procedures in the foregoing method embodiments, and are not described herein again.
The wood species identification method based on multi-source feature fusion provided by the embodiment of the invention can realize business logic through a computer program and is recorded on a storage medium, and the storage medium can be read and executed by a computer to realize the effects of the scheme described in the method embodiment of the specification. Accordingly, embodiments of the present invention also provide a computer readable storage medium for wood species identification comprising a memory for storing processor executable instructions which, when executed by a processor, implement the steps of the wood species identification method based on multi-source feature fusion comprising the previous embodiments.
The storage medium may include physical means for storing information, typically by digitizing the information before storing it in an electronic, magnetic, or optical medium. The storage medium may include: means for storing information using electrical energy such as various memories, e.g., RAM, ROM, etc.; devices for storing information using magnetic energy such as hard disk, floppy disk, magnetic tape, magnetic core memory, bubble memory, and USB flash disk; devices for optically storing information, such as CDs or DVDs. Of course, there are other ways of readable storage medium, such as quantum memory, graphene memory, etc.
The above description of the storage medium according to the method embodiment may further include other implementations, and the implementation principle and the generated technical effects of the embodiment are the same as those of the foregoing method embodiment, and specific reference may be made to the description of the related method embodiment, which is not repeated herein.
The embodiment of the invention also provides equipment for wood species identification, which can be a single computer or can comprise actual operation devices and the like using one or more of the methods or one or more embodiment devices of the specification. The wood species identification device may include at least one processor and a memory storing computer executable instructions that when executed by the processor implement the steps of any one or more of the above-described multi-source feature fusion-based wood species identification methods.
The above description of the apparatus according to the method embodiment may further include other implementations, and the implementation principle and the generated technical effects of the embodiment are the same as those of the foregoing method embodiment, and specific reference may be made to the description of the related method embodiment, which is not repeated herein.
Finally, it should be noted that: the above examples are only specific embodiments of the present invention, and are not intended to limit the scope of the present invention, but it should be understood by those skilled in the art that the present invention is not limited thereto, and that the present invention is described in detail with reference to the foregoing examples: any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or perform equivalent substitution of some of the technical features, while remaining within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the corresponding technical solutions. Are intended to be encompassed within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (2)
1. A method for identifying wood species based on multi-source feature fusion, the method comprising:
establishing a wood classification characteristic reference data set, and dividing the wood classification characteristic reference data set into a training set and a testing set; the wood classification characteristic reference data set comprises a wood section image sample, a wood DNA sequence sample and a wood chemical fingerprint spectrum sample;
constructing a deep learning model, and training the deep learning model by adopting the training set;
extracting key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample respectively by an autonomous learning method through the deep learning model;
fusing the extracted key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample, verifying the fused result by using the test set, and adjusting and optimizing the deep learning model;
acquiring a wood section image, a wood DNA sequence and a wood chemical fingerprint of a wood sample to be identified, and inputting the wood section image, the wood DNA sequence and the wood chemical fingerprint into the deep learning model to obtain a tree species name of the wood sample to be identified;
The building of the wood classification characteristic reference data set and the division into a training set and a testing set comprises the following steps:
acquiring a timber cross section, a diameter section and a chord section structural image acquired from a timber specimen, and performing data enhancement treatment to obtain a timber section image sample;
obtaining DNA extracted from the wood specimen, performing amplification, sequencing and DNA bar code evaluation, and screening out an effective DNA bar code sequence to obtain a wood DNA sequence sample;
acquiring mass spectrum data acquired by scanning the surface of the wood sample by adopting a mass spectrometer, and carrying out normalization treatment to obtain the wood chemical fingerprint sample;
performing data cleaning on the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample to remove abnormal values, and establishing the wood classification characteristic reference data set;
and (3) adopting a k-fold cross validation method to divide the wood classification characteristic reference data set into 8:2 into the training set and the testing set;
the method for extracting key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample by the deep learning model through an autonomous learning method comprises the following steps:
Extracting the timber anatomy feature information contained in the timber cross section, diameter section and chord section structure images respectively, and carrying out matrixing representation to obtain a key feature matrix of the timber section image sample;
the base sequence of the screened effective DNA bar code sequence is matrixed by adopting a k-mer algorithm to obtain a key feature matrix of the wood DNA sequence sample;
acquiring structure and content information of wood characteristic compounds based on the wood chemical fingerprint sample, and matrixing the molecular weight of each characteristic compound to obtain a key characteristic matrix of the wood chemical fingerprint sample;
the method for fusing the extracted key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample, verifying the fused result by using the test set, and adjusting and optimizing the deep learning model comprises the following steps:
normalizing key feature matrixes of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample;
fusing key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix;
Based on the fusion matrix output identification result, the test set is utilized to test and tune the deep learning model, and each parameter of the deep learning model is adjusted and optimized according to the identification result, so that the classification precision of the deep learning model reaches more than 99%;
the constructing the deep learning model, and training the deep learning model by adopting the training set comprises the following steps:
constructing a deep learning model based on convolutional neural networks, wherein the deep learning model comprises three convolutional neural networks in parallel, and each convolutional neural network comprises an input layer, a convolutional layer, an attention mechanism layer, a pooling layer and a full connection layer;
respectively inputting the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample of the training set into three convolutional neural networks of the deep learning model for training, extracting features in the convolutional layer, the attention mechanism layer and the pooling layer, and classifying the features through the fully connected layer;
the timber cross section, the diametral section and the chord section structure images comprise a macroscopic structure image and a microscopic structure image, and the data enhancement processing comprises image rotation, image scaling, image mirroring and/or image clipping;
When the wood characteristic compound is obtained, carrying out peak alignment treatment on the wood chemical fingerprint sample in a traversal charge-mass ratio data mode, and searching out the characteristic compound by comparing the charge-mass ratio;
the wood anatomy feature information includes cross section feature information extracted from the wood cross section configuration image, diametral section feature information extracted from the wood diametral section configuration image, and chord section feature information extracted from the wood chord section configuration image;
the cross section characteristic information comprises a tube hole frequency characteristic, a tube hole diameter characteristic and/or an axial parenchyma frequency characteristic, the diameter section characteristic information comprises a wood ray cell type characteristic and/or a catheter-ray interline hole characteristic, and the chord section characteristic information comprises a wood ray width characteristic, a wood ray height characteristic and/or a wood ray frequency characteristic;
the fusing of the key feature matrix of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample comprises the following steps:
superposing the characteristic values with the same number of rows and columns in the key characteristic matrix of the normalized wood section image sample, the normalized wood DNA sequence sample and the normalized wood chemical fingerprint sample to obtain the fusion matrix;
Or alternatively;
performing matrix splicing operation on the key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix;
the method for obtaining the wood section image, the wood DNA sequence and the wood chemical fingerprint of the wood sample to be identified comprises the steps of:
acquiring a wood section image, a wood DNA sequence and a wood chemical fingerprint of the wood sample to be identified according to the actual application scene and the data acquirable condition of the wood sample to be identified;
inputting the obtained wood section image, wood DNA sequence and wood chemical fingerprint of the wood sample to be identified into the deep learning model, and outputting the tree species name of the wood sample to be identified.
2. A wood species identification device based on multi-source feature fusion, the device comprising:
the data set establishing module is used for establishing a wood classification characteristic reference data set and dividing the wood classification characteristic reference data set into a training set and a testing set; the wood classification characteristic reference data set comprises a wood section image sample, a wood DNA sequence sample and a wood chemical fingerprint spectrum sample;
The model construction module is used for constructing a deep learning model and training the deep learning model by adopting the training set;
the feature extraction module is used for respectively extracting key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample by an autonomous learning method through the deep learning model;
the model optimization module is used for fusing the extracted key features of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint sample, verifying the fused result by using the test set and adjusting and optimizing the deep learning model;
the wood species identification module is used for acquiring a wood section image, a wood DNA sequence and a wood chemical fingerprint of a wood sample to be identified, and inputting the wood section image, the wood DNA sequence and the wood chemical fingerprint into the deep learning model to obtain a wood species name of the wood sample to be identified;
the data set establishing module includes:
the wood section image acquisition unit is used for acquiring wood cross section, diameter section and chord section structural images acquired from a wood sample, and carrying out data enhancement treatment to obtain a wood section image sample;
The wood DNA sequence acquisition unit is used for acquiring DNA extracted from the wood specimen, performing amplification, sequencing and DNA bar code evaluation, and screening out an effective DNA bar code sequence to obtain the wood DNA sequence sample;
the wood chemical fingerprint acquisition unit is used for acquiring mass spectrum data acquired by scanning the surface of the wood sample by adopting a mass spectrometer and carrying out normalization treatment to obtain the wood chemical fingerprint sample;
the data cleaning unit is used for cleaning the data of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample, removing abnormal values and establishing the wood classification characteristic reference data set;
the data set dividing unit is used for adopting a k-fold cross validation method to divide the wood classification characteristic reference data set into 8:2 into the training set and the testing set;
the feature extraction module includes:
the first extraction unit is used for respectively extracting the wood anatomy feature information contained in the wood cross section, diameter section and chord section structural image and carrying out matrixing representation to obtain a key feature matrix of the wood section image sample;
The second extraction unit is used for matrixing the base sequence of the screened effective DNA barcode sequence by adopting a k-mer algorithm to obtain a key feature matrix of the wood DNA sequence sample;
the third extraction unit is used for obtaining the structure and content information of the wood characteristic compounds based on the wood chemical fingerprint sample, and carrying out matrixing representation on the molecular weight of each characteristic compound to obtain a key characteristic matrix of the wood chemical fingerprint sample;
the model optimization module comprises:
the normalization unit is used for normalizing the key feature matrixes of the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample;
the feature fusion unit is used for fusing the key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix;
the model optimization unit is used for outputting an identification result based on the fusion matrix, testing and adjusting parameters of the deep learning model by utilizing the test set, and adjusting and optimizing each parameter of the deep learning model according to the identification result so that the classification precision of the deep learning model reaches more than 99%;
The model construction module comprises:
the model construction unit is used for constructing a deep learning model based on convolutional neural networks, the deep learning model comprises three convolutional neural networks in parallel, and each convolutional neural network comprises an input layer, a convolutional layer, an attention mechanism layer, a pooling layer and a full connection layer;
the model training unit is used for respectively inputting the wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample of the training set into three convolutional neural networks of the deep learning model for training, extracting features in the convolutional layer, the attention mechanism layer and the pooling layer, and classifying the features through the fully connected layer;
the timber cross section, the diametral section and the chord section structure images comprise a macroscopic structure image and a microscopic structure image, and the data enhancement processing comprises image rotation, image scaling, image mirroring and/or image clipping;
when the wood characteristic compound is obtained, carrying out peak alignment treatment on the wood chemical fingerprint sample in a traversal charge-mass ratio data mode, and searching out the characteristic compound by comparing the charge-mass ratio;
the wood anatomy feature information includes cross section feature information extracted from the wood cross section configuration image, diametral section feature information extracted from the wood diametral section configuration image, and chord section feature information extracted from the wood chord section configuration image;
The cross section characteristic information comprises a tube hole frequency characteristic, a tube hole diameter characteristic and/or an axial parenchyma frequency characteristic, the diameter section characteristic information comprises a wood ray cell type characteristic and/or a catheter-ray interline hole characteristic, and the chord section characteristic information comprises a wood ray width characteristic, a wood ray height characteristic and/or a wood ray frequency characteristic;
the feature fusion unit is used for:
superposing the characteristic values with the same number of rows and columns in the key characteristic matrix of the normalized wood section image sample, the normalized wood DNA sequence sample and the normalized wood chemical fingerprint sample to obtain the fusion matrix;
or alternatively;
performing matrix splicing operation on the key feature matrixes of the normalized wood section image sample, the wood DNA sequence sample and the wood chemical fingerprint spectrum sample to obtain a fusion matrix;
the tree species identification module comprises:
the data acquisition unit is used for acquiring a wood section image, a wood DNA sequence and a wood chemical fingerprint of the wood sample to be identified according to the actual application scene and the data acquirable condition of the wood sample to be identified;
the tree species identification unit is used for inputting the obtained wood section image, the wood DNA sequence and the wood chemical fingerprint of the wood sample to be identified into the deep learning model and outputting the tree species name of the wood sample to be identified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211667095.3A CN115861710B (en) | 2022-12-22 | 2022-12-22 | Method and device for identifying wood tree species based on multi-source feature fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211667095.3A CN115861710B (en) | 2022-12-22 | 2022-12-22 | Method and device for identifying wood tree species based on multi-source feature fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115861710A CN115861710A (en) | 2023-03-28 |
| CN115861710B true CN115861710B (en) | 2023-10-27 |
Family
ID=85654444
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211667095.3A Active CN115861710B (en) | 2022-12-22 | 2022-12-22 | Method and device for identifying wood tree species based on multi-source feature fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115861710B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2003000040A2 (en) * | 2001-06-25 | 2003-01-03 | Pulp And Paper Research Institute Of Canada | A nucleic acid-based method for tree phenotype prediction: dna markers for fibre and wood quality properties |
| CN114186625A (en) * | 2021-12-09 | 2022-03-15 | 中国林业科学研究院木材工业研究所 | Wood identification method and system based on image feature fusion |
-
2022
- 2022-12-22 CN CN202211667095.3A patent/CN115861710B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2003000040A2 (en) * | 2001-06-25 | 2003-01-03 | Pulp And Paper Research Institute Of Canada | A nucleic acid-based method for tree phenotype prediction: dna markers for fibre and wood quality properties |
| CN114186625A (en) * | 2021-12-09 | 2022-03-15 | 中国林业科学研究院木材工业研究所 | Wood identification method and system based on image feature fusion |
Non-Patent Citations (2)
| Title |
|---|
| "基于深度强化学习的木材缺陷图像识别及分割模型研究";张旭中等;《电子测量技术》(第17期);第86-92页 * |
| "木材树种计算机视觉识别技术发展与应用";何拓等;《林业工程学》;第 18-27页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115861710A (en) | 2023-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109766872B (en) | Image recognition method and device | |
| CN107729721B (en) | Metabolite identification and disorder pathway analysis method | |
| CN110880019B (en) | Method for adaptively training target domain classification model through unsupervised domain | |
| CN112700325A (en) | Method for predicting online credit return customers based on Stacking ensemble learning | |
| CN112508901B (en) | Underwater structure disease identification method, system, device and storage medium | |
| CN111210869A (en) | Protein cryoelectron microscope structure analysis model training method and analysis method | |
| WO2018218788A1 (en) | Third-generation sequencing sequence alignment method based on global seed scoring optimization | |
| CN115021679B (en) | Photovoltaic equipment fault detection method based on multi-dimensional outlier detection | |
| CN111506785B (en) | Social text-based network public opinion topic identification method and system | |
| CN113392894A (en) | Cluster analysis method and system for multi-group mathematical data | |
| CN113259388A (en) | Network flow abnormity detection method, electronic equipment and readable storage medium | |
| CN116843955A (en) | Microorganism classification and identification method and system based on computer vision | |
| CN116110542A (en) | Data analysis method based on trusted multi-view | |
| CN114547365A (en) | Image retrieval method and device | |
| CN115861710B (en) | Method and device for identifying wood tree species based on multi-source feature fusion | |
| CN115910217B (en) | Base determination method, device, computer equipment and storage medium | |
| CN111863135B (en) | False positive structure variation filtering method, storage medium and computing device | |
| CN114283884A (en) | Original spectrogram processing method and device, computer equipment and storage medium | |
| CN112362292A (en) | Method for anomaly detection of wind tunnel test data | |
| CN113922823B (en) | Social media information propagation graph data compression method based on constraint sparse representation | |
| CN113887674B (en) | Abnormal behavior detection method and system based on big data | |
| CN114186625A (en) | Wood identification method and system based on image feature fusion | |
| EP1691311A1 (en) | Method, system and software for carrying out biological interpretations of microarray experiments | |
| CN117437976B (en) | Disease risk screening method and system based on gene detection | |
| Lu et al. | Accelerating the discovery of biodiversity by detecting “new” species based on machine learning method |
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 |