Wood tree species identification method and device based on structural feature digitization

Abstract

The invention discloses a wood tree species identification method and device based on structural feature digitization, and belongs to the technical field of wood identification. The method comprises the following steps: firstly, constructing a wood microstructure feature image data set and training a wood microstructure feature extraction model; then constructing a wood construction feature digital matrix data set and training a wood tree species identification model; finally, inputting a to-be-identified wood tissue slice image into the trained wood structure feature extraction model, and extracting a digital matrix of the to-be-identified wood tissue slice image; and inputting the digital matrix of the to-be-identified wood tissue slice image into the trained wood tree species identification model to obtain the tree species to which the to-be-identified wood tissue slice image belongs. According to the method, the wood structure features in the image are converted into the digital matrix, and quantitative data of the wood structure features are accurately recorded, so that accurate identification of wood tree species is realized.

Description

Technical Field

[0001] The invention relates to the technical field of wood identification, and in particular to a wood species identification method and device based on digitized structural features. Background Art

[0002] Wood identification is to compare the sample to be tested with a wood specimen that contains complete tree taxonomy and collection information, and give the species name of the sample to be tested based on the differences between wood species. The traditional method of wood species identification is to manually observe the microscopic structural characteristics of the wood, and then compare it with the wood specimen to identify the wood species. However, due to human subjectivity, traditional wood identification methods are generally unable to accurately record a large amount of information such as the location, size, tissue ratio, etc. of the wood structural characteristics. Therefore, they can generally only identify the wood to the "genus" or "category", and cannot achieve accurate identification of wood at the "species" level. Summary of the invention

[0003] In order to solve the defects of the prior art, the present invention provides a method and device for identifying wood species based on digitized structural features, which can convert the wood structural features in an image into a digital matrix, accurately record the quantitative data of the wood structural features, and thus realize accurate identification of wood species.

[0004] The present invention provides the following technical solutions:

[0005] A method for identifying wood species based on digitized structural features, the method comprising:

[0006] Construct a wood microstructure feature image dataset;

[0007] The wood microstructure feature image data set includes a plurality of first-class training samples, each of which includes a correctly named wood tissue slice image and corresponding first-class label data, and the first-class label data is a manually annotated digital matrix representing the wood structure features of the correctly named wood tissue slice image;

[0008] Constructing a wood structural feature extraction model, and using the wood microscopic structural feature image data set to train the wood structural feature extraction model;

[0009] Construct a digital matrix dataset of wood structural characteristics;

[0010] The digital matrix data set of wood structural characteristics includes a plurality of second-class training samples, each of which includes a digital matrix of a correctly named wood tissue slice image and corresponding second-class label data, wherein the second-class label data is the tree species to which the correctly named wood tissue slice image belongs, which is manually annotated;

[0011] Constructing a wood species identification model, and using the wood structural feature digitized matrix data set to train the wood species identification model;

[0012] Obtaining a wood tissue slice image to be identified, and inputting the trained wood structure feature extraction model to extract a digital matrix of the wood tissue slice image to be identified;

[0013] The digital matrix of the wood tissue slice image to be identified is input into the trained wood species identification model to obtain the tree species to which the wood tissue slice image to be identified belongs.

[0014] Further, the digitized matrix is ​​labeled by the following steps:

[0015] Taking the upper left corner of the correctly named wood tissue slice image as the coordinate origin, the contour points of each structural feature are marked with coordinates to obtain the coordinate matrix of each structural feature;

[0016] The coordinate matrices of all structural features are combined to obtain the digitized matrix.

[0017] Furthermore, the structural features include pores, axial parenchyma and wood rays, and the number of contour points marked on each structural feature is not less than 10.

[0018] Furthermore, the wood structure feature extraction model includes an input layer, a convolutional layer, a pooling layer and a fully connected layer, and is embedded with a channel attention mechanism module.

[0019] Further, the correctly named wood tissue slice image and the wood tissue slice image to be identified contain at least one complete wood growth ring.

[0020] A device for identifying wood species based on digitized structural features, the device comprising:

[0021] The first data set construction module is used to construct a wood microstructure feature image data set;

[0022] The wood microstructure feature image data set includes a plurality of first-class training samples, each of which includes a correctly named wood tissue slice image and corresponding first-class label data, and the first-class label data is a manually annotated digital matrix representing the wood structure features of the correctly named wood tissue slice image;

[0023] A first model training module is used to construct a wood structure feature extraction model and train the wood structure feature extraction model using the wood microstructure feature image dataset;

[0024] The second data set construction module is used to construct a digital matrix data set of wood structural characteristics;

[0025] The digital matrix data set of wood structural characteristics includes a plurality of second-class training samples, each of which includes a digital matrix of a correctly named wood tissue slice image and corresponding second-class label data, wherein the second-class label data is the tree species to which the correctly named wood tissue slice image belongs, which is manually annotated;

[0026] A second model training module is used to construct a wood species identification model and train the wood species identification model using the wood structural feature digitized matrix data set;

[0027] A digital matrix extraction module is used to obtain the wood tissue slice image to be identified, and input the trained wood structure feature extraction model to extract the digital matrix of the wood tissue slice image to be identified;

[0028] The wood species identification module is used to input the digital matrix of the wood tissue slice image to be identified into the trained wood species identification model to obtain the tree species to which the wood tissue slice image to be identified belongs.

[0029] Further, the digitized matrix is ​​labeled by the following steps:

[0030] Taking the upper left corner of the correctly named wood tissue slice image as the coordinate origin, the contour points of each structural feature are marked with coordinates to obtain the coordinate matrix of each structural feature;

[0031] The coordinate matrices of all structural features are combined to obtain the digitized matrix.

[0032] Furthermore, the structural features include pores, axial parenchyma and wood rays, and the number of contour points marked on each structural feature is not less than 10.

[0033] Furthermore, the wood structure feature extraction model includes an input layer, a convolutional layer, a pooling layer and a fully connected layer, and is embedded with a channel attention mechanism module.

[0034] Further, the correctly named wood tissue slice image and the wood tissue slice image to be identified contain at least one complete wood growth ring.

[0035] The present invention has the following beneficial effects:

[0036] The present invention digitizes the structural features of wood through images to form a digital matrix, records the relative position, size, shape and other information of the structural features of wood, and on this basis, accurately identifies the wood species. The present invention accurately selects the structural features in the wood image by artificial marking, and then trains the deep learning model so that the model can automatically detect the structural features in the wood image. Then, by means of coordinates, the wood structural features in the image are converted into a digital matrix, and the quantitative data of the wood structural features are accurately recorded, thereby realizing accurate identification of wood species. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flow chart of the wood species identification method based on structural feature digitization of the present invention;

[0038] Figure 2 This is an example of a wood tissue slice image;

[0039] Figure 3 A schematic diagram for labeling wood tissue slice images;

[0040] Figure 4 It is a schematic diagram of the wood species identification device based on the digitization of structural features of the present invention. DETAILED DESCRIPTION

[0041] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0042] The embodiment of the present invention provides a method for identifying wood species based on digitized structural features, such as Figure 1 As shown, the method includes:

[0043] S1: Construct a wood microstructure feature image dataset.

[0044] Among them, the wood microscopic structure feature image dataset includes multiple first-category training samples, each of which includes a correctly named wood tissue slice image and corresponding first-category label data, and the first-category label data is a manually annotated digital matrix representing the wood structure features of the correctly named wood tissue slice image.

[0045] When constructing a wood microstructure feature image dataset, we first collect images from correctly named wood tissue slices through an optical microscope. The collected wood tissue slice images should contain at least one complete wood ring, with an image magnification of 5 times and an image resolution of no less than 4000 pixels * 3000 pixels. At least 3 different wood tissue slice samples should be collected for each tree species, and at least 5 images of different fields of view should be collected for each wood tissue slice sample. An example of the collected wood tissue slice image is as follows: Figure 2 shown.

[0046] Then you need to annotate the digitized matrix for the wood tissue slice image as a label. The process of annotating the digitized matrix is ​​as follows:

[0047] 1. Take the upper left corner of the correctly named wood tissue slice image as the coordinate origin (0, 0), mark the coordinates of the contour points of each structural feature, and obtain the coordinate matrix of each structural feature.

[0048] The aforementioned structural features include wood structural features such as pores, axial parenchyma and wood rays. Each structural feature has no less than 10 contour points marked, and each structural feature forms a coordinate matrix. An example of a coordinate mark is Figure 3 shown.

[0049] 2. Combine the coordinate matrices of all structural features to obtain a digitized matrix.

[0050] The digital matrix records the relative position, size, shape and other information of each structural feature.

[0051] S2: Construct a wood structural feature extraction model and use the wood microstructure feature image dataset to train the wood structural feature extraction model. Obtain a wood structural feature extraction model that can extract a digital matrix.

[0052] The wood structure feature extraction model includes an input layer, a convolution layer, a pooling layer and a fully connected layer, and is embedded with a channel attention mechanism module. The aforementioned wood microstructure feature image dataset is input into the model for training, and the above-mentioned wood structure features can be automatically detected through the attention module.

[0053] S3: Construct a digital matrix dataset of wood structural characteristics.

[0054] Among them, the digital matrix data set of wood structural characteristics includes multiple second-category training samples, each of which includes a digital matrix of correctly named wood tissue slice images and corresponding second-category label data, and the second-category label data is the tree species to which the correctly named wood tissue slice images belong that are manually annotated.

[0055] S4: Construct a wood species identification model, and use the wood structural feature digital matrix data set to train the wood species identification model, so as to obtain a wood species identification model that can identify wood species.

[0056] S5: Obtain the wood tissue slice image to be identified, and input it into the trained wood structure feature extraction model to extract the digital matrix of the wood tissue slice image to be identified.

[0057] Similar to the correctly named wood tissue slice image, the wood tissue slice image to be identified is also required to contain at least one complete wood growth ring.

[0058] S6: Inputting the digitized matrix of the wood tissue slice image to be identified into the trained wood species identification model to obtain the tree species to which the wood tissue slice image to be identified belongs.

[0059] The present invention digitizes the structural features of wood through images to form a digital matrix, records the relative position, size, shape and other information of the structural features of wood, and on this basis, accurately identifies the wood species. The present invention accurately selects the structural features in the wood image by artificial marking, and then trains the deep learning model so that the model can automatically detect the structural features in the wood image. Then, by means of coordinates, the wood structural features in the image are converted into a digital matrix, and the quantitative data of the wood structural features are accurately recorded, thereby realizing accurate identification of wood species.

[0060] The embodiment of the present invention also provides a device for identifying wood species based on digitized structural features, such as Figure 4 As shown, the device comprises:

[0061] The first data set construction module 1 is used to construct a wood microstructure feature image data set.

[0062] Among them, the wood microscopic structure feature image dataset includes multiple first-category training samples, each of which includes a correctly named wood tissue slice image and corresponding first-category label data, and the first-category label data is a manually annotated digital matrix representing the wood structure features of the correctly named wood tissue slice image.

[0063] The first model training module 2 is used to construct a wood structure feature extraction model and use a wood microstructure feature image dataset to train the wood structure feature extraction model.

[0064] The second data set construction module 3 is used to construct a digital matrix data set of wood structural characteristics.

[0065] Among them, the digital matrix data set of wood structural characteristics includes multiple second-category training samples, each of which includes a digital matrix of correctly named wood tissue slice images and corresponding second-category label data, and the second-category label data is the tree species to which the correctly named wood tissue slice images belong that are manually annotated.

[0066] The second model training module 4 is used to construct a wood species identification model and train the wood species identification model using a wood structural feature digitized matrix data set.

[0067] The digital matrix extraction module 5 is used to obtain the wood tissue slice image to be identified and input the trained wood structure feature extraction model to extract the digital matrix of the wood tissue slice image to be identified.

[0068] The wood species identification module 6 is used to input the digitized matrix of the wood tissue slice image to be identified into the trained wood species identification model to obtain the tree species to which the wood tissue slice image to be identified belongs.

[0069] The present invention digitizes the structural features of wood through images to form a digital matrix, records the relative position, size, shape and other information of the structural features of wood, and on this basis, accurately identifies the wood species. The present invention accurately selects the structural features in the wood image by artificial marking, and then trains the deep learning model so that the model can automatically detect the structural features in the wood image. Then, by means of coordinates, the wood structural features in the image are converted into a digital matrix, and the quantitative data of the wood structural features are accurately recorded, thereby realizing accurate identification of wood species.

[0070] In the present invention, the digitized matrix can be annotated by the following steps:

[0071] The upper left corner of the correctly named wood tissue slice image is taken as the coordinate origin, and the contour points of each structural feature are marked with coordinates to obtain the coordinate matrix of each structural feature.

[0072] The coordinate matrices of all construction features are combined to obtain a digitized matrix.

[0073] The structural features include pores, axial parenchyma and wood rays, and the number of contour points marked on each structural feature is not less than 10.

[0074] The wood structure feature extraction model includes input layer, convolution layer, pooling layer and fully connected layer, and is embedded with a channel attention mechanism module.

[0075] The aforementioned correctly named wood tissue slice image and the wood tissue slice image to be identified contain at least one complete wood growth ring.

[0076] The device provided in the above embodiment, its implementation principle and the technical effect produced correspond to the embodiment of the above method one by one. For the sake of brief description, the part not mentioned in the embodiment of the device can refer to the corresponding content in the embodiment of the above method. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the modules and units described in the device can refer to the corresponding process in the embodiment of the above method, and will not be repeated here.

[0077] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the aforementioned embodiments, those of ordinary skill in the art should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed by the present invention, or can easily conceive of changes, or make equivalent replacements for some of the technical features therein. Such modifications, changes or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention.

Claims

1. A method for identifying wood species based on digitized structural features, characterized in that: The method comprises: Construct a wood microstructure feature image dataset; The wood microstructure feature image data set includes a plurality of first-class training samples, each of which includes a correctly named wood tissue slice image and corresponding first-class label data, and the first-class label data is a manually annotated digital matrix representing the wood structure features of the correctly named wood tissue slice image; Constructing a wood structural feature extraction model, and using the wood microscopic structural feature image data set to train the wood structural feature extraction model; Construct a digital matrix dataset of wood structural characteristics; The digital matrix data set of wood structural characteristics includes a plurality of second-class training samples, each of which includes a digital matrix of a correctly named wood tissue slice image and corresponding second-class label data, wherein the second-class label data is the tree species to which the correctly named wood tissue slice image belongs, which is manually annotated; Constructing a wood species identification model, and using the wood structural feature digitized matrix data set to train the wood species identification model; Obtaining a wood tissue slice image to be identified, and inputting the trained wood structure feature extraction model to extract a digital matrix of the wood tissue slice image to be identified; The digital matrix of the wood tissue slice image to be identified is input into the trained wood species identification model to obtain the tree species to which the wood tissue slice image to be identified belongs.

2. The method for identifying wood species based on structural feature digitization according to claim 1, characterized in that: The digitized matrix is ​​labeled by the following steps: Taking the upper left corner of the correctly named wood tissue slice image as the coordinate origin, the contour points of each structural feature are marked with coordinates to obtain the coordinate matrix of each structural feature; The coordinate matrices of all structural features are combined to obtain the digitized matrix.

3. The method for identifying wood species based on structural feature digitization according to claim 2, characterized in that: The structural features include pores, axial parenchyma and wood rays, and the number of contour points marked on each structural feature is not less than 10.

4. The method for identifying wood species based on structural feature digitization according to claim 3 is characterized in that: The wood structure feature extraction model includes an input layer, a convolution layer, a pooling layer and a fully connected layer, and is embedded with a channel attention mechanism module.

5. The method for identifying wood species based on structural feature digitization according to claim 4, characterized in that: The correctly named wood tissue slice image and the wood tissue slice image to be identified contain at least one complete wood growth ring.

6. A device for identifying wood species based on digital structural features, characterized in that: The device comprises: The first data set construction module is used to construct a wood microstructure feature image data set; The wood microstructure feature image data set includes a plurality of first-class training samples, each of which includes a correctly named wood tissue slice image and corresponding first-class label data, and the first-class label data is a manually annotated digital matrix representing the wood structure features of the correctly named wood tissue slice image; A first model training module is used to construct a wood structure feature extraction model and train the wood structure feature extraction model using the wood microstructure feature image dataset; The second data set construction module is used to construct a digital matrix data set of wood structural characteristics; The digital matrix data set of wood structural characteristics includes a plurality of second-class training samples, each of which includes a digital matrix of a correctly named wood tissue slice image and corresponding second-class label data, wherein the second-class label data is the tree species to which the correctly named wood tissue slice image belongs, which is manually annotated; A second model training module is used to construct a wood species identification model and train the wood species identification model using the wood structural feature digitized matrix data set; A digital matrix extraction module is used to obtain the wood tissue slice image to be identified, and input the trained wood structure feature extraction model to extract the digital matrix of the wood tissue slice image to be identified; The wood species identification module is used to input the digital matrix of the wood tissue slice image to be identified into the trained wood species identification model to obtain the tree species to which the wood tissue slice image to be identified belongs.

7. The device for identifying wood species based on structural feature digitization according to claim 6 is characterized in that: The digitized matrix is ​​labeled by the following steps: Taking the upper left corner of the correctly named wood tissue slice image as the coordinate origin, the contour points of each structural feature are marked with coordinates to obtain the coordinate matrix of each structural feature; The coordinate matrices of all structural features are combined to obtain the digitized matrix.

8. The device for identifying wood species based on structural feature digitization according to claim 7, characterized in that: The structural features include pores, axial parenchyma and wood rays, and the number of contour points marked on each structural feature is not less than 10.

9. The device for identifying wood species based on structural feature digitization according to claim 8, characterized in that: The wood structure feature extraction model includes an input layer, a convolution layer, a pooling layer and a fully connected layer, and is embedded with a channel attention mechanism module.

10. The device for identifying wood species based on structural feature digitization according to claim 9, characterized in that: The correctly named wood tissue slice image and the wood tissue slice image to be identified contain at least one complete wood growth ring.

Images