A fuzzy inference classification and recognition method for open set of wood species based on near-infrared spectroscopy analysis

Abstract

A fuzzy inference classification and recognition method for open-set wood species based on near-infrared spectroscopy analysis, belonging to the field of wood classification. In order to improve the recognition accuracy of wood and reduce the occurrence of misjudgment. In the present invention, near-infrared spectra of wood species samples are collected, and the obtained near-infrared spectral data of wood species samples are subjected to PCA dimensionality reduction processing to obtain the near-infrared spectral data of wood species samples after dimensionality reduction processing, which are divided into wood species training samples and wood species test samples; an open-set fuzzy inference classifier is used to train the wood species training samples to generate fuzzy rules; an open-set fuzzy inference classification and recognition judgment is performed on the wood species test samples; for unknown wood species, first, an open-set fuzzy inference classification and recognition judgment is performed on the obtained near-infrared spectral data of the unknown wood species samples after dimensionality reduction processing. The present invention realizes "rejecting" unknown tree species samples and improves the accuracy of tree species recognition.

Description

Technical Field

[0001] The present invention belongs to the field of wood classification, and particularly relates to a fuzzy inference classification and recognition method for an open set of wood tree species based on near-infrared spectroscopy analysis. Background Art

[0002] Solid wood is a basic wood industrial product and is the main raw material in the furniture and building materials industries. Different types of wood have different characteristic parameters, so their uses, physical properties, and prices vary greatly. Resources in nature are very rich, and there are also a wide variety of tree species. According to the statistics of the International Union for Conservation of Nature (IUCN), there are a total of 60,065 tree species globally. It is unrealistic and unnecessary to study a wood tree species classification and recognition system to identify all these tree species. In fact, in many cases, it is necessary to study specific wood tree species in specific regions and categories, which will have theoretical research significance and practical application backgrounds. For example, the precious and rare tree species, rosewood, includes a total of 5 genera and 33 species. To prevent the occurrence of situations where inferior products are passed off as rosewood tree species in the market, it is necessary to develop a specific classification and recognition system for these 33 rosewood species.

[0003] Currently, the mainstream detection method in the research of wood tree species recognition is non-destructive testing methods, such as image processing methods, spectral analysis methods, etc. Near-infrared spectroscopy analysis technology has the advantages of low cost, high efficiency, fast speed, non-destructive, convenient detection, good test reproducibility, etc., and has been widely used in the qualitative and quantitative research of product quality in various fields. However, the recognition accuracy of the classification and recognition system for wood is not high, and there are cases of misjudgment. Summary of the Invention

[0004] The problem to be solved by the present invention is to improve the recognition accuracy of wood and reduce the occurrence of misjudgment, and a fuzzy inference classification and recognition method for an open set of wood tree species based on near-infrared spectroscopy analysis is proposed.

[0005] To achieve the above object, the present invention is realized through the following technical solutions:

[0006] A fuzzy inference classification and recognition method for an open set of wood tree species based on near-infrared spectroscopy analysis, comprising the following steps:

[0007] S1. Collect near-infrared spectra of wood tree species samples to obtain near-infrared spectral data of wood tree species samples;

[0008] S2. Perform PCA dimensionality reduction processing on the near-infrared spectral data of the wood tree species samples collected in step S1 to obtain the near-infrared spectral data of the wood tree species samples after dimensionality reduction processing;

[0009] S3. Divide the near-infrared spectral data of the wood tree species samples after dimensionality reduction processing in step S2 into wood tree species training samples and wood tree species test samples;

[0010] S4. Use an open-set fuzzy inference classifier to train the wood species training samples and generate fuzzy rules;

[0011] S5. Perform open-set fuzzy inference classification and recognition judgment on the wood species test samples;

[0012] S6. For unknown wood species, first perform steps S1 - S2, and then perform the open-set fuzzy inference classification and recognition judgment in step S5 on the near-infrared spectral data of the unknown wood species samples obtained after dimensionality reduction processing.

[0013] Furthermore, each wood species in step S1 includes 50 samples, and the near-infrared spectrum acquisition method includes the following steps:

[0014] S1.1. Place the wood species sample to be measured on the bracket of the spectrometer with the surface to be measured facing down. The fiber optic probe is 5 mm away from the surface of the wood species sample to be measured, and the diameter of the circular field of view is 6.35 mm;

[0015] S1.2. The near-infrared spectrum is acquired by the diffuse reflection method. The spectral region acquisition range is 950 - 1650 nm, and the spectral wavelength resolution is 5.4 nm;

[0016] S1.3. Before collecting the spectrum, perform spectral whiteboard calibration and dark calibration. The whiteboard calibration uses a standard polytetrafluoroethylene whiteboard as the background, and the dark calibration is performed in a light-shielding manner. At the same time, turn on the electronic dark noise correction and stray light correction buttons;

[0017] S1.4. Set the parameters in the software SpectraSuite supporting the spectral instrument: the integration time is 600 ms, repeat scanning 900 times and take the average, and the smoothness is 5;

[0018] S1.5. When collecting samples, randomly select 4 points on the cross-section of each sample for spectral collection and calculate their average value. After measuring 5 samples, perform a standard whiteboard calibration once.

[0019] Furthermore, after the PCA dimensionality reduction processing in step S2, the first 4 principal components are retained for the near-infrared spectral data of the wood species samples.

[0020] Furthermore, the number of wood species training samples in step S3 is 40, and the number of wood species test samples is 10.

[0021] Furthermore, the specific implementation method of step S4 includes the following steps:

[0022] S4.1. The fuzzy rule formula for establishing an open-set fuzzy inference classifier is:

[0023] Rule R j : If xp1 is A j1 and…x pn is A jn ; Then Class D j with CF j , j = 1, 2, …N;

[0024] where N is the total number of rules, R j is the rule, Class D j is Class D j , CF j is the confidence coefficient of rule R j ;

[0025] The training samples are x p = (x p1 , x p2 , …x pn ), p = 1, 2, …m, m is the number of training samples, n is the feature dimension of the samples;

[0026] In the field of near-infrared spectroscopy classification of wood species, the feature vector of the near-infrared spectral data of the wood species samples after dimensionality reduction is represented as (PC1, PC2, PC3, PC4), so n = 4. At this time, x p = (x p1 , x p2 , …x p4 ) = (PC1, PC2, PC3, PC4);

[0027] Each tree species includes 50 samples, 40 of which are used as training samples, A j1 , …A jn are the fuzzy sets on the corresponding sample feature components, C i is a known tree species category corresponding to this rule, D j = C i , i = 1, 2, …C; C is the total number of known tree species categories in the training set, then m = 40C;

[0028] S4.2, Set the known tree species category h = C i , i = 1, 2, …C, calculate the sum of the products of the membership function values of all training samples of this category corresponding to this rule R j . The specific calculation formula is:

[0029]

[0030] β Class h (j) is the sum of the products of the membership function values of the known tree species category h, μ j (x p ) is the sample xp Membership value;

[0031] S4.3. Search for Class corresponding to β Class h (j) maximum value, the specific calculation formula is:

[0032]

[0033] is β Class h (j) maximum value;

[0034] If there is a unique exists, then the class corresponding to this rule is Calculate the corresponding confidence coefficient CF j , generate a fuzzy rule, the calculation formula is:

[0035]

[0036]

[0037] is the average value of the sum of the product of the membership function values of the known tree species category h, excluding the tree species

[0038] S4.4. Repeat steps S4.1 - 4.3 to generate all N fuzzy rules.

[0039] Furthermore, the specific implementation method of step S5 includes the following steps:

[0040] S5.1. For the test sample x, it is known that x = (x1,..., x4) = (PC1,..., PC4);

[0041] S5.2. For the rule R j , set the fuzzy set of rule R j to be A ji , i = 1, 2, 3, 4; for the probability that the test sample x = (PC1,..., PC4) belongs to the fuzzy set A ji , the calculation formula is:

[0042]

[0043] where I, L, K, D represent the number of fuzzy sets of PC1,..., PC4; these 4 probabilities are independent of each other and satisfy the condition

[0044] S5.3. Calculate for the rule R j the probability that the sample to be recognized belongs to the rule R jprobability is as follows:

[0045]

[0046] S5.4. Define the probability that a test sample belongs to a certain tree species category as

[0047]

[0048] S5.5. Construct the Generalized Basic Probability Distribution Transformation Algorithm GBPA for test samples. The GBPA of a test sample belonging to a known tree species category is defined as m{C i}:

[0049]

[0050] The GBPA of a test sample belonging to an unknown tree species category is defined as:

[0051]

[0052] S5.7. Conduct classification judgment. According to give a judgment conclusion. If obtains the maximum value, determine x as an unknown tree species category; otherwise, if m{C i} obtains the maximum value, determine it as the known tree species category C i .

[0053] Advantages of the present invention:

[0054] A fuzzy inference classification and recognition method for wood tree species in an open set based on near-infrared spectroscopy analysis according to the present invention proposes a near-infrared spectroscopy classification and recognition system for wood tree species in an open set situation. It can not only accurately identify samples of known tree species categories, but also effectively identify and "reject" samples of unknown tree species categories. It first uses the Principal Component Analysis (PCA) algorithm to extract PC1-PC4 of the near-infrared spectrum of the wood cross-section. Secondly, according to the fuzzy inference algorithm in the open set, use the training set samples of known tree species categories to train the fuzzy inference classifier to generate an effective if-then rule set. Each rule in this rule set corresponds to a known tree species category and also has a rule confidence coefficient to represent the credibility of the rule. Finally, when an unknown sample x enters this fuzzy inference system, first calculate the probability of the 4 fuzzy sets of the PC1-PC4 features of this sample belonging to a certain rule R j , and calculate the probability of this sample belonging to this rule on this basis Then, for a certain known tree species C i , find all the rules of this tree species and calculate the The maximum value among them is the GBPA of this tree species category, that is Finally, calculate the GBPA of this sample belonging to the unknown tree species category, that is The category corresponding to the maximum value of these GBPA probabilities is the final classification result.

[0055] A fuzzy inference classification and recognition method for open set of wood tree species based on near-infrared spectroscopy according to the present invention realizes "rejecting recognition" of unknown tree species samples and improves the accuracy of tree species recognition. It shows that the specific classification recognition rate is closely related to the number of known / unknown tree species, the number of samples of the two major tree species, and the design of the membership function of the fuzzy inference classifier. Brief Description of the Drawings

[0056] Figure 1 It is a flowchart of a fuzzy inference classification and recognition method for open set of wood tree species based on near-infrared spectroscopy according to the present invention. Detailed Embodiments

[0057] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the specific embodiments described are only a part of the embodiments of the present invention, rather than all of the specific embodiments. The components of the specific embodiments of the present invention usually described and shown in the drawings here can be arranged and designed in various different configurations, and the present invention can also have other embodiments.

[0058] Therefore, the detailed description of the specific embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but only represents the selected specific embodiments of the present invention. All other specific embodiments obtained by those skilled in the art based on the specific embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0059] In order to further understand the content, features and effects of the present invention, the following specific embodiments are exemplified and described in detail in conjunction with the accompanying drawings as follows: Specific Embodiment 1:

[0061] A fuzzy inference classification and recognition method for open set of wood tree species based on near-infrared spectroscopy includes the following steps:

[0062] S1. Collect near-infrared spectra of wood tree species samples to obtain near-infrared spectral data of wood tree species samples;

[0063] Further, each wood species in step S1 includes 50 samples, and the near-infrared spectrum acquisition method includes the following steps:

[0064] S1.1. Place the wood species sample to be measured on the bracket of the spectrometer, with the surface to be measured facing downwards. The fiber optic probe is 5 mm away from the surface of the wood species sample to be measured, and the diameter of the circular field of view is 6.35 mm;

[0065] S1.2. The near-infrared spectrum is acquired in a diffuse reflection mode. The spectral region acquisition range is 950 - 1650 nm, and the spectral wavelength resolution is 5.4 nm;

[0066] S1.3. Before acquiring the spectrum, perform spectral whiteboard calibration and dark calibration. The whiteboard calibration uses a standard polytetrafluoroethylene whiteboard as the background, and the dark calibration is carried out in a light-shielding manner. At the same time, turn on the electronic dark noise correction and stray light correction buttons;

[0067] S1.4. Set the parameters in the software SpectraSuite supporting the spectral instrument: the integration time is 600 ms, repeat the scan 900 times and take the average, and the smoothness is 5;

[0068] S1.5. When acquiring the sample, randomly select 4 points on the cross-section of each sample for spectral acquisition and calculate their average value. After measuring every 5 samples, perform a standard whiteboard calibration once;

[0069] S2. Perform PCA dimensionality reduction processing on the near-infrared spectrum data of the wood species samples acquired in step S1 to obtain the near-infrared spectrum data of the wood species samples after dimensionality reduction processing;

[0070] Further, after the PCA dimensionality reduction processing in step S2, the dimension of the near-infrared spectrum data of the wood species samples retains the first 4 principal components;

[0071] S3. Divide the near-infrared spectrum data of the wood species samples after dimensionality reduction processing in step S2 into wood species training samples and wood species test samples;

[0072] Further, the number of wood species training samples in step S3 is 40, and the number of wood species test samples is 10;

[0073] S4. Use an open-set fuzzy inference classifier to train the wood species training samples to generate fuzzy rules;

[0074] Further, the specific implementation method of step S4 includes the following steps:

[0075] S4.1. The fuzzy rule formula for establishing an open-set fuzzy inference classifier is:

[0076] Rule R j : If x p1is A j1 and…x pn is A jn ; Then Class D j with CF j , j = 1, 2, … N; (1)

[0077] where N is the total number of rules, R j is a rule, Class D j is Class D j , CF j is the confidence coefficient of rule R j ;

[0078] The training sample is x p = (x p1 , x p2 , … x pn ), p = 1, 2, … m, m is the number of training samples, and n is the feature dimension of the samples;

[0079] In the field of near-infrared spectroscopy classification of wood species, the feature vector of the near-infrared spectroscopy data of the wood species samples after dimensionality reduction is represented as (PC1, PC2, PC3, PC4). Therefore, n = 4. At this time, x p = (x p1 , x p2 , … x p4 ) = (PC1, PC2, PC3, PC4);

[0080] Each tree species includes 50 samples, and 40 of them are used as training samples, A j1 , … A jn are fuzzy sets on the corresponding sample feature components, C i is a known tree species category corresponding to this rule, D j = C i , i = 1, 2, … C; C is the total number of known tree species categories in the training set, then m = 40C;

[0081] S4.2. Set the known tree species category h = C i , i = 1, 2, … C, and calculate the sum of the products of the membership function values of all training samples of this category corresponding to this rule R j . The specific calculation formula is:

[0082]

[0083] β Class h (j) is the sum of the products of the membership function values of the known tree species category h, μ j (x p ) is the sample x pMembership value;

[0084] S4.3. Search for Class Corresponding to β Class h (j) maximum value, the specific calculation formula is:

[0085]

[0086] Is β Class h (j) maximum value;

[0087] If there is a unique Exists, then the class corresponding to this rule is Calculate the corresponding confidence coefficient CF j , generate a fuzzy rule, the calculation formula is:

[0088]

[0089]

[0090] Is the average value of the sum of the product of the membership function values of the known tree species categories h, excluding the tree species

[0091] S4.4. Repeat steps S4.1 - 4.3 to generate all N fuzzy rules;

[0092] S5. Perform open - set fuzzy inference classification and recognition judgment on the test samples of wood tree species;

[0093] Furthermore, the specific implementation method of step S5 includes the following steps:

[0094] S5.1. For the test sample x, it can be known that x = (x1,..., x4) = (PC1,..., PC4);

[0095] S5.2. For the rule R j , set the fuzzy set of the rule R j as A ji , i = 1, 2, 3, 4; For the probability that the test sample x = (PC1,..., PC4) belongs to the fuzzy set A ji , the calculation formula is:

[0096]

[0097] Where I, L, K, D represent the number of fuzzy sets of PC1,..., PC4; These 4 probabilities Are independent of each other and satisfy the condition

[0098] B5.3. Calculate the probability that the sample to be recognized belongs to rule R j The probability that the sample to be recognized belongs to rule R j is as follows:

[0099]

[0100] S5.4. Define the probability that the test sample belongs to a certain tree species category as

[0101]

[0102] S5.5. Construct the Generalized Basic Probability Assignment transformation algorithm GBPA for the test sample. The GBPA that the test sample belongs to a known tree species category is defined as m{C i}:

[0103]

[0104] The GBPA that the test sample belongs to an unknown tree species category is defined as:

[0105]

[0106] S5.7. Make a classification judgment. According to give a judgment conclusion. If obtains the maximum value, determine x as the unknown tree species category; otherwise, if m{C i} obtains the maximum value, determine it as the known tree species category C i ;

[0107] S6. For unknown wood tree species, first execute steps S1 - S2, and then perform the open - set fuzzy inference classification and recognition judgment in step S5 on the near - infrared spectral data of the unknown wood tree species samples obtained after dimensionality reduction processing.

[0108] The following is an example to illustrate the technical effect of the present invention:

[0109] For example, for 20 existing wood tree species, take 10 tree species as known categories and the remaining 10 tree species as unknown categories. The number of fuzzy rules included in the effective rule set can be obtained as 301, while the complete rule set contains 37856 rules. That is to say, in the rules represented by formula (1), N is the total number of rules and N = 301. For a sample to be tested, after performing the 4 steps in this section, the GBPA shown in Table 1 is obtained, and thus the sample is determined to be an unknown tree species category.

[0110] Table 1. Calculation results of GBPA for the sample to be tested

[0111]

[0112] Some specific experimental results are given below. Due to space limitations, only the classification results under two representative datasets are presented here. The first dataset includes 10 known tree species and 10 unknown tree species (each tree species contains 50 samples), and is divided into two groups. In Group 1.1, the samples of known tree species are divided into a training set and a test set according to the "hold-out method" at a ratio of 4:1; that is, the training set has 400 samples and the test set has 100 samples. The 500 samples of unknown tree species are all included in the test set. Therefore, there are a total of 600 test samples, which are classified into known and unknown tree species categories at a ratio of 1:5. In Group 1.2, the samples of known tree species are processed in the same way as in Group 1.1, but for the unknown tree species category, only 10 test samples are selected for each tree species. In this way, there are a total of 200 test samples, which are classified into known and unknown tree species categories at a ratio of 1:1.

[0113] The second dataset includes 2 known tree species and 18 unknown tree species (each tree species contains 50 samples), which is a highly imbalanced dataset and is also divided into two groups. In Group 2.1, the samples of known tree species are still divided into a training set and a test set at a ratio of 4:1; that is, the training set has 80 samples and the test set has 20 samples. The 900 samples of unknown tree species are all included in the test set. Therefore, there are a total of 920 test samples, which are classified into known and unknown tree species categories at a ratio of 1:45. In Group 2.2, the samples of known tree species are processed in the same way as in Group 2.1, but for the unknown tree species category, only 10 test samples are selected for each tree species. In this way, there are a total of 200 test samples, which are classified into known and unknown tree species categories at a ratio of 1:9.

[0114] The classification results in two groups are shown in Tables 2 and 3. To compare the classification accuracy, three classification metrics, namely F-Score, Kappa coefficient, and Overall Recognition Accuracy (ORA), are adopted. In addition, visible light images (RGB) of the cross-sections of wood samples are collected, and texture operators such as improved-basic gray level aura matrix (I-BGLAM, Zamri et al. 2016), local binary pattern (LBP), and fractal texture analysis (FTA, Xu et al. 2009) are used to extract texture features as classification features. The classifier used is Support Vector Machine (SVM). For one-class classifiers, one-class SVM (OC-SVM; Chang and Lin, 2011), Support Vector Data Description (SVDD) and Weight-SVDD (Rahmanimanesh et al. 2004), Bayesian-SVDD (Sotiris et al. 2010), genetic algorithm based SVDD (GA-SVDD; Shon and Moon, 2007), and particle swarm optimized SVDD (PSO-SVDD; Duan et al. 2016) are selected. During classification, a two-level structure is adopted, that is, first, a one-class classifier is used to classify the samples into two major categories: known tree species and unknown tree species, and then the known tree species category is further subdivided into a specific tree species; the classification feature used is the texture feature of the RGB image.

[0115] Comparison of Classification Accuracy of the First Group of Datasets in Table 2

[0116]

[0117]

[0118]

[0119] Comparison of Classification Accuracy of the Second Group of Datasets in Table 3

[0120]

[0121]

[0122]

[0123] In Tables 2 and 3, the F-Score cannot be calculated in a few cases because the denominator in the fraction becomes zero during the calculation process, resulting in the inability to continue the calculation. The present invention adopts a novel GBPA conversion calculation method to convert the near-infrared spectral fuzzy inference classification results of wood species under the closed set framework into GBPA under the open set framework. When making classification decisions, according to the corresponding decision conclusions are given.

[0124] It should be noted that relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

[0125] Although the present application has been described above with reference to specific embodiments, various improvements can be made to it and components can be replaced with equivalents without departing from the scope of the present application. In particular, as long as there is no structural conflict, the various features in the specific embodiments disclosed in the present application can be combined with each other in any way. The exhaustive description of these combinations is not given in this specification only for the sake of saving space and resources. Therefore, the present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A fuzzy inference classification and recognition method for open set of wood species based on near-infrared spectroscopy analysis, characterized in that: It includes the following steps: S1. Collect the near-infrared spectra of the wood species samples to obtain the near-infrared spectral data of the wood species samples; S2. Perform PCA dimensionality reduction on the near-infrared spectral data of the wood species samples collected in step S1 to obtain the near-infrared spectral data of the wood species samples after dimensionality reduction; S3. Divide the near-infrared spectral data of the wood species samples after dimensionality reduction in step S2 into wood species training samples and wood species test samples; S4. Use an open-set fuzzy inference classifier to train the wood species training samples to generate fuzzy rules; S5. Perform open-set fuzzy inference classification and recognition judgment on the wood species test samples; The specific implementation method of step S5 includes the following steps: S5.

1. For the test sample x, it is known that x=(x1,…,x4)=(PC1,…,PC4), and (PC1,…,PC4) represents the feature vector of the near-infrared spectral data of the wood species samples after dimensionality reduction; S5.

2. For rule R j , set the fuzzy set of rule R j as A ji , where i = 1, 2, 3, 4; for the probability that the test sample x = (PC1, …, PC4) belongs to the fuzzy set A ji , the calculation formula is as follows: where I, L, K, D represent the number of fuzzy sets of PC1, …, PC4; these four probabilities are mutually independent and satisfy the condition μ j () is the membership value of the sample (). S5.

3. Calculate the probability that j the sample to be recognized belongs to rule R j is: as follows Among them, CF j is the confidence coefficient; S5.

4. Define the probability that a test sample belongs to a certain tree species as Among them, D j is the category, C i is a known tree species category corresponding to this rule, and max is the maximum value function; S5.

5. Construct the Generalized Basic Probability Assignment (GBPA) transformation algorithm for test samples. The GBPA where the test sample belongs to a known tree species is defined as m{C i}: The GBPA that the test sample belongs to an unknown tree species is defined as: S5.

7. Conduct classification judgment. According to give a judgment conclusion. If obtains the maximum value, determine x as the unknown tree species category; otherwise, if m{C i} obtains the maximum value, determine it as the known tree species category C i , where C is the total number of known tree species categories in the training set; S6. For unknown wood species, first perform steps S1 - S2, and perform the open-set fuzzy inference classification and recognition judgment in step S5 on the near-infrared spectral data of the unknown wood species samples after dimensionality reduction obtained; 2. The fuzzy inference classification and recognition method for open set of wood species based on near-infrared spectroscopy analysis according to claim 1, wherein: In step S1, each wood species includes 50 samples, and the near-infrared spectrum collection method includes the following steps: S1.

1. Place the wood species sample to be measured on the bracket of the spectrometer, with the surface to be measured facing down, the fiber optic probe 5 mm away from the surface of the wood species sample to be measured, and the diameter of the circular field of view being 6.35 mm; S1.

2. The near-infrared spectrum is collected in a diffuse reflection mode, the spectral region collection range is 950 - 1650 nm, and the spectral wavelength resolution is 5.4 nm; S1.

3. Before collecting the spectrum, perform spectral whiteboard calibration and dark calibration. The whiteboard calibration uses a standard polytetrafluoroethylene whiteboard as the background, and the dark calibration is carried out in a light-shielding manner. At the same time, turn on the electronic dark noise correction and stray light correction buttons; S1.

4. Set the parameters in the software SpectraSuite supporting the spectral instrument: the integration time is 600 ms, repeat scanning 900 times and take the average, and the smoothness is 5; S1.

5. When collecting samples, randomly select 4 points on the cross-section of each sample for spectral collection and calculate their average value. After measuring 5 samples, perform a standard whiteboard calibration once.

3. A fuzzy inference classification and recognition method for an open set of wood tree species based on near-infrared spectroscopy analysis according to claim 2, characterized in that: In step S3, the number of wood species training samples is 40, and the number of wood species test samples is 10.

4. A fuzzy inference classification and recognition method for an open set of wood tree species based on near-infrared spectroscopy analysis according to claim 3, characterized in that: The specific implementation method of step S4 includes the following steps: S4.

1. The fuzzy rule formula for establishing an open-set fuzzy inference classifier is: Rule R j : If x p1 is A j1 and…x pn is A jn ; Then Class D j with CF j , j = 1, 2, … N; where N is the total number of rules, R j is a rule, and Class D j is class D j , and CF j is the confidence factor of rule R j . The training sample is x p =(x p1 , x p2 , … x pn ), where p = 1, 2, … m, m is the number of training samples, and n is the feature dimension of the samples; In the field of near-infrared spectroscopy classification of wood species, the feature vector of the near-infrared spectral data of the wood species samples after dimensionality reduction is represented as (PC1, PC2, PC3, PC4). Therefore, n = 4, and at this time x p =(x p1 , x p2 , … x p4 ) = (PC1, PC2, PC3, PC4); Each tree species includes 50 samples, 40 of which are used as training samples, A j1 ,…A jn is a fuzzy set on the corresponding sample feature component, C i is a known tree species category corresponding to this rule, D j = C i , i = 1, 2, … C; C is the total number of known tree species categories in the training set, then m = 40C; S4.

2. Set the known tree species category h = C i , where i = 1, 2,..., C, and calculate the sum of the products of the membership function values corresponding to all training samples of this category for this rule R j . The specific calculation formula is as follows: Among them, β Class h (j) is the sum of the product of the membership function values of the known tree species category h, μ j (x p ) is the membership value of the sample x p ; S4.

3. Search for Class corresponding to β Class h (j) maximum value, and the specific calculation formula is: Among them, is the maximum value of β Class h (j); If there is a unique Class exists, then the category corresponding to this rule is calculate the corresponding confidence coefficient CF j , and generate a fuzzy rule, the calculation formula is: Among them, is the average value of the sum of the product of the membership function values of the known tree species category h, excluding the tree species Class S4.

4. Repeat steps S4.1 - 4.3 to generate all N fuzzy rules.

Images