CN109064553A - Solid wood board knot form inversion method based on near-infrared spectrum analysis - Google Patents
Solid wood board knot form inversion method based on near-infrared spectrum analysis Download PDFInfo
- Publication number
- CN109064553A CN109064553A CN201811265532.2A CN201811265532A CN109064553A CN 109064553 A CN109064553 A CN 109064553A CN 201811265532 A CN201811265532 A CN 201811265532A CN 109064553 A CN109064553 A CN 109064553A
- Authority
- CN
- China
- Prior art keywords
- defect
- spectrum
- sample
- knot
- neural network
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 239000002023 wood Substances 0.000 title claims abstract description 48
- 239000007787 solid Substances 0.000 title claims abstract description 40
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 30
- 238000004458 analytical method Methods 0.000 title claims abstract description 19
- 230000007547 defect Effects 0.000 claims abstract description 141
- 238000001228 spectrum Methods 0.000 claims abstract description 57
- 230000003595 spectral effect Effects 0.000 claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000001514 detection method Methods 0.000 claims abstract description 9
- 238000013528 artificial neural network Methods 0.000 claims description 36
- 230000008569 process Effects 0.000 claims description 21
- 230000009467 reduction Effects 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 11
- 230000005856 abnormality Effects 0.000 claims description 10
- 238000004611 spectroscopical analysis Methods 0.000 claims description 9
- 238000009499 grossing Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 claims description 2
- 238000002203 pretreatment Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 12
- 238000010183 spectrum analysis Methods 0.000 abstract description 7
- 239000000284 extract Substances 0.000 abstract description 5
- 238000009877 rendering Methods 0.000 abstract 1
- 238000002474 experimental method Methods 0.000 description 5
- 239000000463 material Substances 0.000 description 5
- 230000000877 morphologic effect Effects 0.000 description 4
- 241000218652 Larix Species 0.000 description 3
- 235000005590 Larix decidua Nutrition 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000002604 ultrasonography Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 244000166124 Eucalyptus globulus Species 0.000 description 1
- 241000544657 Larix gmelinii Species 0.000 description 1
- 241001138451 Nothofagus betuloides Species 0.000 description 1
- 241001236212 Pinus pinaster Species 0.000 description 1
- 235000005105 Pinus pinaster Nutrition 0.000 description 1
- 241001480055 Quercus mongolica Species 0.000 description 1
- 238000002835 absorbance Methods 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013531 bayesian neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000009643 growth defect Effects 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000002715 modification method Methods 0.000 description 1
- 238000013425 morphometry Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000009659 non-destructive testing Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3563—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Theoretical Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Biochemistry (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Pathology (AREA)
- Analytical Chemistry (AREA)
- Immunology (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
基于近红外光谱分析的实木板材节子形态反演方法，属于实木缺陷检测识别技术领域。本发明的目的是利用近红外光谱分析技术准确地刻画出节子在木材内部的形态，进而对实木板材力学性能和缺陷类别进行准确预测与识别。首先利用图像提取表面节子的缺陷区域，计算出相应中心位置；然后，提取节子缺陷边缘的多点位置，采集这些边缘点的光谱信息并进行基线漂移与去噪处理；运用主成分与马氏距离结合剔除异常光谱，利用K‑S划分校正样本集，并运用ISOMAP对缺陷光谱信息特征进行有效提取；最后，应用小波网络建立节子边缘光谱与该点倾角间的关系，并利用Solidworks软件实现节子椎体模型的三维呈现。本发明用于实木板材节子形态的建模。
The invention discloses a knot shape inversion method based on near-infrared spectrum analysis, and belongs to the technical field of solid wood defect detection and recognition. The purpose of the present invention is to use near-infrared spectral analysis technology to accurately describe the shape of knots inside the wood, and then accurately predict and identify the mechanical properties and defect categories of solid wood boards. First, use the image to extract the defect area of the surface knot, and calculate the corresponding center position; then, extract the multi-point position of the edge of the knot defect, collect the spectral information of these edge points, and perform baseline drift and denoising processing; K-S is used to divide the correction sample set, and ISOMAP is used to effectively extract the characteristics of defect spectral information; finally, the wavelet network is used to establish the relationship between the node edge spectrum and the inclination angle of the point, and the solidworks software is used to Realize the three-dimensional rendering of the knot-shaped vertebral body model. The invention is used for modeling the shape of knots of solid wood boards.
Description
Technical field
The present invention relates to a kind of solid wood board knot form inversion methods, belong to solid wood defects detection identification technology field.
Background technique
Knot destroys the uniformity and integrality of timber, not only influences product appearance as timber growth defect, but also
Knot affects the mechanical property of solid wood board, how accurately to depict knot in the form of wood internal, and then calculate reality
Wood plank mechanical property is the problem in science with application value.
In recent years, since non-destructive testing technology has the characteristics that non-demolition, flexible, reliable, many scholars have applied machine
The methods of vision, ultrasound examination, X-ray scanning and laser imaging have carried out coherent detection research to solid wood defect.
Gonzalo A.Ruz[1]Automatic Visual Inspection (AVI) system has been built, correct point of 11 kinds of defect classification image sets is realized
Class；Zhang Yizhuo[2,3]Sort research is carried out for wood surface defect and texture type, firstly, application image is divided
Technology extracts defect, completes defect characteristic fusion and identification using PCA and SOM；Then, solid wood figure is extracted using dual-tree complex wavelet
Picture frequency spectrum information completes the Fast Classification of Feature Dimension Reduction and texture, defect classification with PSO and CS.The ultrasound examination of defect
It is to complete the qualitative estimation of knot information and wood performance, Jose the time transmitted in solid wood board according to acoustic signals[4]Structure
It has built the ultrasonic wave detecting system of defect and maritime pine has been tested, defect area ratio is at 20% to 30%, accuracy rate
It is 75% to 83%；Zhang Houjiang[5]Vibration test system is constructed, discovery transverse mode of vibration can effectively measure plank
Test material intrinsic frequency, Elastic modulus prediction value are slightly larger than the elasticity modulus that static mode measures.Olsson[6]It proposes to apply laser
Imaging technique measures wood surface fiber angle, and then predicts defects in timber state mechanical property related to estimation timber.X-ray is swept
Retouch the Density Distribution and defect shape information that can grasp solid wood board comprehensively with machine vision technique[7], but the device is complicated, at
This is higher；Mohamad[8]The defect form, defect concentration and the net wooden density that solid wood board is measured using the technology, pass through meter
The KDR parameter for calculating defect, has estimated MOE the and MOR mechanics parameter of timber.
Currently, the defects detection based on machine vision solves the detection of solid wood blemish surface and identification, ultrasound examination can
To judge the presence of defect, although X-ray can comprehensively grasp solid wood information, its testing cost is higher.
Near-infrared spectral analysis technology is judged by detection timber hydric group, i.e. C-H, N-H, O-H content of material
Timber related physical and chemical property.Liang Hao[9]The correlation of near infrared spectrum with solid wooden floor board defect type is analyzed, and is led to
Crossing Bayesian neural network realizes the identification of defect；Yang Zhong[10]It is built respectively using near infrared spectrum combination SIMCA method
The defect recognition model of eucalyptus is found；Zhang Yizhuo etc.[11,12]Using near-infrared spectrum technique, be utilized respectively GN sample preferably with
CPLS modeling, Isomap-PLS method predict the related mechanical property of Mongolian oak.Although near-infrared spectral analysis technology can be with
Solid wood board mechanical property and defect classification are predicted and identified, but how using technology completion defect plate mechanics
Performance evaluation is not yet carried out.
Summary of the invention
The object of the present invention is to provide a kind of solid wood board knot form inversion method based on near-infrared spectrum analysis, with
Knot is accurately depicted using near-infrared spectral analysis technology in the form of wood internal, and then to solid wood board mechanical property
Accurate Prediction and identification are carried out with defect classification.
The technical solution adopted by the present invention to solve the above technical problem is:
A kind of solid wood board knot form inversion method based on near-infrared spectrum analysis, the realization process of the method
Are as follows:
Training process:
The defect borderline region that solid wood board upper and lower surfaces knot is extracted first with image processing techniques is found out up and down
The center of corresponding boundary point and knot blemish surface on face knot defect boundary obtains one table of solid wood board
On knot defect boundary point A, the point place corresponding central point O of blemish surface and another surface corresponding to this point on face
The boundary point inclination angle OAA ' that is constituted of defect boundary point A ',
Using the spectral information at near-infrared spectrometers acquisition defect boundary point inclination angle, the spectral information is carried out pre-
Effective training set is obtained after processing；
It is effectively extracted with spectral information of the ISOMAP to the defects of training set boundary point inclination angle, with lacking for extraction
The spectral effective feature for falling into boundary point inclination angle is input, using defect boundary point tilt angles as output, constructs Wavelet Neural Network
Network is to obtain the Nonlinear Mapping relationship between spectral effective feature and boundary point inclination angle OAA；
Detection process:
The defect borderline region that a certain face knot of solid wood board is extracted first with image processing techniques, finds out the surface
Multiple boundary points on knot defect boundary；
The spectral information of the multiple boundary point is acquired using near-infrared spectrometers, and carries out baseline drift and denoising
Processing, the validity feature for extracting each spectrum input to trained wavelet neural network, obtain each defect boundary point and incline
Angle angle is presented using these tilt angles by the three-dimensional of Solidworks software realization knot vertebral model.
In the training process to the spectral information carry out pretreatment include: first carry out baseline drift and denoising, then
With principal component in conjunction with mahalanobis distance rejecting abnormalities spectrum, then using K-S divide calibration samples collection.
It is effectively extracted with spectral information of the ISOMAP to the defects of training set boundary point inclination angle, process are as follows:
When using Isomap algorithm to plate defect edge spectral signature dimensionality reduction, dimensionality reduction number range d is set as 1~15,
The range of neighbour k is 2~20, and the various combination for choosing d and k tests Data Dimensionality Reduction effect, works as k=19, when d=12,
It is 12.53 with the smallest SECV value, defect angle prediction related coefficient R is 0.83, and prediction relation analysis error RPD is
1.80。
12 dimension data of validity feature of each spectrum after Isomap is optimized is exported as input as 1 dimension data, use
Wavelet-neural network model predicts plate defect angle, sets 14 for the hidden layer node of wavelet neural network, wavelet neural
The wavelet basis function of network uses Morlet function, the learning rate of setting, anticipation error, study number be respectively 0.01,
0.001,1000, it is pre- for boundary point inclination angle to be trained to obtain trained wavelet-neural network model with 103 samples
It surveys.
First carry out the process of baseline drift and denoising are as follows:
It carries out first derivative to the near infrared spectrum of defect sample first to handle to obtain derivative spectrum, first derivative spectrum is pre-
Processing method is as follows:
In formula, xtFor the discrete spectrum at wavelength t, g is window width；
Then, spectrum high-frequency noise is removed using S-G smoothing processing, improves signal-to-noise ratio, calculate it is smoothed at wavelength t after
Mean Value Formulas is as follows:
In formula, xt,smoothIt is the numerical value at smooth rear wavelength t, hiFor S-G smoothing factor, H is normalization factor, and w is flat
Sliding window size；
With the process of principal component rejecting abnormalities spectrum in conjunction with mahalanobis distance are as follows:
Defect Edge point sample spectrum is considered as exception with the averaged spectrum sample that similarity is low in principal component space, is led to
It crosses and different weight coefficient e values is selected to carry out size adjustment to threshold range, and using model prediction result come rejecting abnormalities
It is as follows to calculate threshold range existing for unusual sample for sample:
In formula, AtFor threshold range,The geneva matrix for being near infrared spectrum data to averaged spectrum in principal component space
Average value；σ is the standard deviation of A；E is the weight coefficient for adjusting threshold range；
The spectroscopic data of Defect Edge point sample i is Ai, work as Ai-AtIt is worth smaller, both shows that similarity is higher；Otherwise also
So.Therefore, e value is bigger, and similarity increases therewith；Conversely, similarity is lower；
The process of calibration samples collection is divided using K-S are as follows:
(1) in defect spectrum samples n, calibration set number of samples m is set；
(2) the Euclidean distance l between spectrum is calculatedij, select lijMaximum sample n1And n2Into calibration set；
(3) remaining spectrum is calculated separately at a distance from calibration set spectrum, and is minimized and is constituted set l, l=min (l1v,
l2v,,…,lqv), wherein 1~q indicates the number for being selected into calibration set sample, and v indicates the number of remaining sample to be selected；
(4) it chooses the corresponding sample of maximum value in l and enters calibration set, which is that distance has the farthest sample of calibration samples
This；
(5) it is repeated in and carries out step (3), (4), until calibration set number of samples reaches preset number m.
The wavelet neural network replaces the activation primitive of BP network hidden layer, wavelet neural network using wavelet basis function
Input parameter be characteristic wavelength point that defect sample point preferably goes out, the prediction output of wavelet neural network is that defect point being inclined
Angle；It is x in input signal sequenceiWhen (i=1,2 ..., k), the hidden layer of wavelet neural network exports calculation formula are as follows:
In formula, h (j) is the output of j-th of node of hidden layer of wavelet neural network, wijFor the input of wavelet neural network
The connection weight of layer and hidden layer, hjFor wavelet basis function, bjFor hjShift factor, ajFor hjContraction-expansion factor；
Shown in the calculation formula of wavelet neural network output layer such as formula (5):
In formula, wikFor the connection weight of hidden layer to output layer, h (i) is the output of i-th of hidden layer node, and l is implicit
Node layer number, p are output layer number of nodes.
The beneficial effects of the present invention are: defect morphological analysis of the primary study of the present invention based on near-infrared spectral analysis technology
Method, the knot three-dimensional oblique cone model proposed using Pablo[13], by acquiring the near infrared light spectrum information of Defect Edge, structure
The numerical value spectrum for building defect oblique cone incidence angle Yu defect composition transfer utilizes the multiple spot angle information quantitative forecast defect at edge
Form inside solid wood board, so for the quantitative forecast containing defective solid wood board mechanical property provide theories integration and
With reference to.
A kind of solid wood knot inversion method based on near-infrared spectrum analysis proposed by the present invention, it is anti-to give knot defect
The basic step drilled.Method calculates respective center position first with the defect area of image zooming-out face knot；Then,
The multiple spot position for extracting knot Defect Edge, acquires the spectral information of these marginal points and carries out baseline drift and denoising；
With principal component in conjunction with mahalanobis distance rejecting abnormalities spectrum, using K-S divide calibration samples collection, and with ISOMAP to defect
Spectral information characteristics are effectively extracted；Finally, the relationship between knot edge spectrum and the inclination angle is established using wavelet network,
And it is presented using the three-dimensional of Solidworks software realization knot vertebral model.Experiment is using the conduct pair of larch solid wood board
As 160 groups of spectroscopic datas of 40 knots being chosen and acquiring, by measuring the relative tertiary location of upper and lower surfaces knot, meter
Calculate the true value at each point inclination angle.The experimental results showed that using S-G it is smooth+first derivative carries out Pretreated spectra, this method obtains
The spectral profile arrived is apparent, and absorption peak becomes apparent from；Using Isomap dimension reduction method, Nonlinear Dimension Reduction number d=12, neighbour are chosen
Number k=19, SECV are minimum, and method eliminates the redundant data of spectral information；The knot inclination angle established using wavelet neural network
Nonlinear model, prediction related coefficient 0.88, prediction standard difference are 7.65, and relation analysis error is 2.14.
Detailed description of the invention
Fig. 1 is defect oblique cone morphometry schematic diagram；Fig. 2 is experiment flow figure；Fig. 3 is wavelet neural network structure
Figure；Fig. 4 and Fig. 5 is that single order leads the smooth larch defect sample spectrogram of+S-G, in figure: Fig. 4 is the defect sample before pretreatment
This spectrogram, Fig. 5 are pretreated defect sample spectrogram；Fig. 6 is PCA combination mahalanobis distance rejecting abnormalities defect sample
Figure；Fig. 7 is the SECV of the Isomap-PLS model under different d, k relationships；Fig. 8 is that the wavelet network based on ISOMAP predicts knot
Fruit figure, Fig. 9 are the wavelet network prediction result figure based on full spectrum；Figure 10 is defect sample edge sample point diagram；Figure 11 is scarce
Morphological Simulation figure is fallen into, in figure: (a) drawing oblique cone aspect graph for SolidWorks, (b) indicate defect plate schematic diagram.
Specific embodiment
As shown in Figure 1 to 11, the solid wood board knot form described in present embodiment based on near-infrared spectrum analysis
The realization process of inversion method is as follows:
1 materials and methods
1.1 experimental materials and experiment flow
For the external morphology of defect, the three-dimensional knot model proposed using Pablo, i.e., knot ideal at oblique circle
Cone[13], defect form is as shown in Figure 1, the basic model needs 3 parameters to determine, respectively O, A and θ.Wherein, O is defect
In the ellipse center location that solid wood board surface is presented；A is the marginal point of solid wood board surface defect, and OA distance can pass through
Image processing techniques is calculated；θ is the inclination angle of A point on Defect Edge.Physically, can by defect on solid wood board,
The relative position of lower surface calculates the external morphology of defect, and then calculates the θ value on marginal point, and the value is as data sample
True value is associated with near infrared spectrum data foundation, completes the prediction at Defect Edge point inclination angle.Theoretically, with Defect Edge information
The increase of acquisition, defect form are more approached with actual conditions.
Experiment selects larch (Larix gmelinii) that plate is made afterwards through processing, therefrom selects containing defect, nothing
Obvious 40, color difference sample is simultaneously numbered.With NIRQuest512 type light-near infrared optical fiber spectrograph and
The near infrared light spectrum information of 4 points of SpectraSuite software collection Defect Edge obtains 160 groups of data.In order to improve modeling
Precision, using exceptional sample rejecting, sample set division, the baseline drift of spectral information and denoising, spectral signature wavelength
The pre-treatment steps such as extraction；When Defect Edge inclination angle is predicted, using wavelet neural network establish spectroscopic data and degree of tilt it
Between relationship；It is presented finally, completing the three-dimensional of data using SolidWorks, completes defect morphological Simulation.Detailed process is as schemed
2。
The near infrared spectrum of 1.2 defect sample pre-processes
1.2.1 spectral information baseline drift, denoising
In order to eliminate baseline drift existing for the near infrared spectrum at plate defect edge and spectra overlapping, and then can be obvious
The variation tendency of spectrum is found out on ground, carries out first derivative to the near infrared spectrum of defect sample first and handles to obtain derivative spectrum,
First derivative spectrum preprocess method is as follows:
In formula, xtFor the discrete spectrum at wavelength t, g is window width.
Then, spectrum high-frequency noise is removed using S-G smoothing processing, improves signal-to-noise ratio, calculate it is smoothed at wavelength t after
Mean Value Formulas is as follows:
In formula, hiFor S-G smoothing factor, H is normalization factor.
1.2.2 sample set divides
For defect spectrum samples unevenly distributed, intermediate sample is easy to cause prediction result to deviate true value, presents
" equalization " phenomenon out divides here, completing calibration set using K-S method.Defect spectrum samples collection division methods based on K-S
It is as follows:
(1) in defect spectrum samples n, calibration set number of samples m is set；
(2) the Euclidean distance l between spectrum is calculatedij, select lijMaximum sample n1And n2Into calibration set；
(3) remaining spectrum is calculated separately at a distance from calibration set spectrum, and is minimized and is constituted set l.L=min (l1v,
l2v,,…,lqv), wherein 1~q indicates the number for being selected into calibration set sample, and v indicates the number of remaining sample to be selected；
(4) it chooses the corresponding sample of maximum value in l and enters calibration set, which is that distance has the farthest sample of calibration samples
This；
(5) it is repeated in and carries out step (3), (4), until calibration set number of samples reaches preset number m.
1.2.3 exceptional sample is rejected
The predictive ability of model can be improved in the rejecting of abnormal defect sample, herein using principal component in conjunction with mahalanobis distance
Method complete exceptional sample rejecting[14].Defect Edge point sample spectrum is similar in principal component space to averaged spectrum
It spends low sample and is considered as exception, size adjustment is carried out to threshold range by selecting different weight coefficient e values, and use mould
Type prediction result carrys out rejecting abnormalities sample.It is as follows that this method calculates threshold range existing for unusual sample:
In formula, AtFor threshold range,The geneva square for being near infrared spectrum data to averaged spectrum in principal component space
Battle array average value；σ is the standard deviation of A；E is the weight coefficient for adjusting threshold range.
The spectroscopic data of Defect Edge point sample i is Ai, work as Ai-AtIt is worth smaller, both shows that similarity is higher；Otherwise also
So.Therefore, e value is bigger, and similarity increases therewith；Conversely, similarity is lower.
1.2.4 characteristic wavelength extracts
Spectroscopic data pretreatment after, in order to improve the accuracy of model, need to collected near infrared light spectrum information into
Row characteristic wavelength extracts, and deletes the redundancy of information.Early-stage study shows that Isomap algorithm has good Nonlinear Dimension Reduction energy
Power[12], extracted herein using the characteristic wavelength that Isomap carries out Defect Edge sample point.Isomap algorithm is broadly divided into three steps
It is rapid: firstly, setting dimensionality reduction number d and neighbour number k, obtains approximate geodesic distance by the shortest path in the data neighborhood figure of building
From；Then, d dimension data is found out using MDS；Finally, effectively output low-dimensional insertion, using d dimension data and angle true value as model
Input, obtains characteristic wavelength by seeking minimal error.
The 1.3 plate defect angle prediction models based on wavelet network
Wavelet network, instead of the activation primitive of BP network hidden layer, is mentioned using wavelet basis function so that wavelet network has
The advantages of taking local message completes the weight of BP network and the optimization of threshold value, overcomes BP network vulnerable to local extremum shadow
It rings[15].The network structure of wavelet neural network is as shown in Figure 3.
In Fig. 3, X1,X2,…,XkThe input parameter of wavelet neural network, i.e., defect sample point preferably go out characteristic wavelength
Point, Y are the prediction output of wavelet neural network, and numerical value is the inclination angle of defect point.It is x in input signal sequencei(i=1,
2 ..., k) when, the hidden layer of wavelet neural network exports calculation formula are as follows:
In formula, h (j) is the output of j-th of node of hidden layer of wavelet neural network, wijFor the input of wavelet neural network
The connection weight of layer and hidden layer, hjFor wavelet basis function, bjFor hjShift factor, ajFor hjContraction-expansion factor.
Shown in the calculation formula of wavelet neural network output layer such as formula (5):
In formula, wikFor the connection weight of hidden layer to output layer, h (i) is the output of i-th of hidden layer node, and l is implicit
Node layer number, p are output layer number of nodes.
Wavelet neural network uses the weight and wavelet basis function parameter of gradient modification method corrective networks, is calculated by the amendment
The amendment of method approaches wavelet neural network prediction output gradually with desired output, and error between the two is smaller and smaller.
2 experimental results
The near infrared spectrum of 2.1 defect sample pre-processes
It is pre-processed using near infrared spectrum of the method that first derivative, S-G are smoothly combined to plate defect sample,
As a result see Fig. 4 and Fig. 5.S-G is smooth+and the major absorbance peak of first derivative treated spectrogram becomes apparent from, while having filtered out more
More high-frequency noise bring interference, spectral profile are more clear, smoothly.
To collected 160 plate defect samples, the ratio between calibration set number and forecast set number are set using K-S method
For 3:1, obtaining calibration set sample is 120, and forecast set sample is 40；Weight coefficient is taken using principal component combination mahalanobis distance
E is 1.25, rejects 19 abnormal defect sample altogether, 17 exceptional samples are wherein eliminated in calibration set, eliminate 2 in forecast set
A exceptional sample, rejecting abnormalities sample results are as shown in Figure 6.
When using Isomap algorithm to plate defect edge spectral signature dimensionality reduction, dimensionality reduction number range d is set as 1~15,
The range of neighbour k is 2~20, and the various combination for choosing d and k tests Data Dimensionality Reduction effect, and model selection is traditional
As a result PLS is shown in Fig. 7.As shown in Figure 7, work as k=19, when d=12, there is the smallest SECV value, SECV 12.53, at this point, lacking
Falling into angle prediction related coefficient R is 0.83, and prediction relation analysis error RPD is 1.80.
2.2 wavelet neural networks predict plate defect angle
12 dimension datas after Isomap is optimized are exported as input as 1 dimension data, pre- using wavelet-neural network model
Plate defect angle to be surveyed, sets 14 for hidden layer node, wavelet basis function uses Morlet function, the learning rate of setting,
Anticipation error, study number are respectively 0.01,0.001,1000, are trained with 103 samples, after network convergence, to 38
A sample is predicted that prediction result is as shown in Figure 8.At this point, Prediction Parameters R, SEP, RPD are respectively 0.89,7.65,2.14.
Fig. 9 is the wavelet network prediction result under all-wave length effect, and Prediction Parameters R, SEP, RPD are respectively 0.41,19.33,1.09.
Parameter comparison shows more accurate based on wavelet network prediction result of the ISOMAP feature after preferred.
In order to verify the superiority of wavelet network, PLS, BP network and wavelet network prediction result is selected to carry out respectively
Compare, the relevant parameter of 3 models is as shown in table 1.The prediction phase relation of wavelet-neural network model prediction plate defect angle
Number R, RPD are above the prediction result of PLS and BP model, and SEP is minimum.Experiment shows that near infrared spectrum data is inclined with defect
There is certain non-linear spectrum between angle, wavelet neural network prediction result is better than PLS linear model；Since BP network needs to save
Point setting and optimization, therefore, neural network forecast is low compared with wavelet network precision of prediction.
Prediction result of the different modeling methods of table 1 to defect angle
Table.1 The prediction comparison of defect angle based on different
models
2.3 plate defect three-dimensional configuration invertings
Defect sample picture is as shown in Figure 10, and 4 spectroscopic datas of sheet material measurement Defect Edge are obtained by model prediction
Corresponding defect inclination angle, specific number are as shown in table 2.By the spatial information of 4 defect points and inclination data input
The form of defect is finally inversed by SolidWorks2016 software, as shown in figure 11.
2 defect morphological parameters of table
Table.2 The parameters of defect shape
3 conclusions of the invention
The present invention applies near-infrared spectrum analysis and mode identification technology in Wood Defects Testing, is plate defect shape
State detection provides a kind of new feasible method, provides reference frame to sheet mechanical property influence for quantitative analysis defect.
The experimental results showed that smooth, single order is led the method combined and pre-processed near infrared spectrum data using S-G, can obtain
To relatively sharp, smooth spectrum；Effective extraction that characteristic wavelength is realized using Isomap Method of Nonlinear Dimensionality Reduction, is improved
The modeling accuracy at defect inclination angle；Effective modeling at Defect Edge inclination angle is realized using wavelet network non-linear modeling method, benefit
Go out the corresponding oblique cone of plate defect with SolidWorks2016 Software on Drawing, realizes the simulation to defect form.It builds
In mold process, due to that can have certain measurement error in Defect Edge positioning, the error of training set will affect defect inclination angle
Prediction models the robustness at the precise positioning of primary study Defect Edge and defect inclination angle in the research of follow-up phase.
The bibliography that the present invention quotes is as follows:
[1]Ruz,G.A.,Estévez,P.A.and Ramirez,P.A.International Journal of
Systems Science,2009,40(2):163
[2]Zhang Yizhuo,Xu Chao,Li Chao,et.al.Journal of forestry research,
2015,26(3):745.
[3]Zhang Yizhuo,Liu Sijia,Cao Jun,et.al.Wood Science and Technology,
2016,50(3):297.
[4] Machado, J.S., Sardinha, R.A., Cruz, H.P.Wood Science and Technology,
2004,38:277.
[5] Zhang Houjiang, Shen Shijie, Cui Yingying, et.al (Zhang Houjiang, Shen Shijie, Cui Ying
Grain husk, etc.) .Journal of Beijing Forestry University (Beijing Forestry University's journal), 2005,27 (6):
91.[6]Olsson,A.,Oscarsson,J.,Serrano,E.et al.European Journal of Wood
Products.2013,71(3):319.
[7]Jungkwon O,Kugbo S,Kwangmo K,et al.Journal of Wood Science,2009,55
(4):264.[8]Hittawe M M,Sidibe D,Meriaudeau F.Proceedings of SPIE-The
International Society for Optical Engineering,2015,9534.
[9] Liang Hao, Cao Jun, Lin Xue, et al (Liang Hao, Cao Jun, Lin Xue, etc.) .Spectroscopy
And Spectral Analysis (spectroscopy and spectrum analysis), 2017,37 (7): 2041.
[10] Yang Zhong, Zhang Mao Mao, Chen Ling, et al.Bioresources, 2015,10 (2):
3314.
[11] Zhang Yizhuo, Su Yaowen, Li Chao, et al (Zhang Yizhuo, Su Yaowen, Li Chao, etc.)
.Journal of Forestry Engineering (Forestry Engineering journal), 2016,1 (6): 17.
[12] Zhang Yizhuo, Su Yaowen, Li Chao, et al (Zhang Yizhuo, Su Yaowen, Li Chao, etc.)
.Journal of Beijing Forestry University (Beijing Forestry University's journal), 2016,38 (8): 99.
[13]Pablo Guindos,Manuel Guaita.Wood Science and Technology,2013,47
(3):585.
[14] Zhang Lingshua, Wang Weidong, Gu Yunhon, et al (Zhang Lingshuai, Wang Weidong, Gu Yunhong,
Deng) .Spectroscopy and Spectral Analysis (spectroscopy and spectrum analysis), 2011,31 (5): 1254.
[15] Song Rushun (Song Rushun) .Control and Decision (control and decision), 2000,15 (6):
765.
Claims (6)
1. a kind of solid wood board knot form inversion method based on near-infrared spectrum analysis, which is characterized in that the method
Realization process are as follows:
Training process:
The defect borderline region that solid wood board upper and lower surfaces knot is extracted first with image processing techniques, finds out upper and lower surface
The center of corresponding boundary point and knot blemish surface on knot defect boundary obtains on one surface of solid wood board
Knot defect boundary point A, the corresponding central point O of blemish surface where the point and another surface corresponding to this point on lack
The boundary point inclination angle OAA ' that boundary point A ' is constituted is fallen into,
Using the spectral information at near-infrared spectrometers acquisition defect boundary point inclination angle, the spectral information is pre-processed
After obtain effective training set；
It is effectively extracted with spectral information of the ISOMAP to the defects of training set boundary point inclination angle, with the defect side of extraction
The spectral effective feature at boundary point inclination angle be input, using defect boundary point tilt angles as output, construct wavelet neural network from
And obtain the Nonlinear Mapping relationship between spectral effective feature and boundary point inclination angle OAA；
Detection process:
The defect borderline region that a certain face knot of solid wood board is extracted first with image processing techniques, finds out the face knot
Multiple boundary points on defect boundary；
The spectral information of the multiple boundary point is acquired using near-infrared spectrometers, and is carried out at baseline drift and denoising
Reason, the validity feature for extracting each spectrum input to trained wavelet neural network, obtain each defect boundary point inclination angle
Angle is presented using these tilt angles by the three-dimensional of Solidworks software realization knot vertebral model.
2. the solid wood board knot form inversion method according to claim 1 based on near-infrared spectrum analysis, feature
It is, carrying out pretreatment to the spectral information in the training process includes: first to carry out baseline drift and denoising, then use
Then principal component rejecting abnormalities spectrum in conjunction with mahalanobis distance divides calibration samples collection using K-S.
3. the solid wood board knot form inversion method according to claim 2 based on near-infrared spectrum analysis, feature
It is, is effectively extracted with spectral information of the ISOMAP to the defects of training set boundary point inclination angle, process are as follows:
When using Isomap algorithm to plate defect edge spectral signature dimensionality reduction, dimensionality reduction number range d is set as 1~15, neighbour k
Range be 2~20, the various combination for choosing d and k tests Data Dimensionality Reduction effect, works as k=19, when d=12, has most
Small SECV value is 12.53, and defect angle prediction related coefficient R is 0.83, and prediction relation analysis error RPD is 1.80.
4. the solid wood board knot form inversion method according to claim 3 based on near-infrared spectrum analysis, feature
It is,
12 dimension data of validity feature of each spectrum after Isomap is optimized is exported as input as 1 dimension data, using small echo
Neural Network model predictive plate defect angle sets 14 for the hidden layer node of wavelet neural network, wavelet neural network
Wavelet basis function use Morlet function, the learning rate of setting, anticipation error, study number be respectively 0.01,0.001,
1000, it is trained to obtain trained wavelet-neural network model for the prediction of boundary point inclination angle with 103 samples.
5. the solid wood board knot form inversion method according to claim 2,3 or 4 based on near-infrared spectrum analysis,
It is characterized in that,
First carry out the process of baseline drift and denoising are as follows:
First derivative is carried out to the near infrared spectrum of defect sample first to handle to obtain derivative spectrum, first derivative spectrum pretreatment
Method is as follows:
In formula, xtFor the discrete spectrum at wavelength t, g is window width；
Then, spectrum high-frequency noise is removed using S-G smoothing processing, improves signal-to-noise ratio, calculate it is smoothed at wavelength t after be averaged
It is as follows to be worth formula:
In formula, xt,smoothIt is the numerical value at smooth rear wavelength t, hiFor S-G smoothing factor, H is normalization factor, and w is smoothing windows
Mouth size；
With the process of principal component rejecting abnormalities spectrum in conjunction with mahalanobis distance are as follows:
Defect Edge point sample spectrum is considered as exception with the averaged spectrum sample that similarity is low in principal component space, passes through choosing
Size adjustment carried out to threshold range with different weight coefficient e values, and using model prediction result come rejecting abnormalities sample,
It is as follows to calculate threshold range existing for unusual sample:
In formula, AtFor threshold range,The geneva matrix for being near infrared spectrum data to averaged spectrum in principal component space is average
Value；σ is the standard deviation of A；E is the weight coefficient for adjusting threshold range；
The spectroscopic data of Defect Edge point sample i is Ai, work as Ai-AtIt is worth smaller, both shows that similarity is higher；Vice versa.Cause
This, e value is bigger, and similarity increases therewith；Conversely, similarity is lower；
The process of calibration samples collection is divided using K-S are as follows:
(1) in defect spectrum samples n, calibration set number of samples m is set；
(2) the Euclidean distance l between spectrum is calculatedij, select lijMaximum sample n1And n2Into calibration set；
(3) remaining spectrum is calculated separately at a distance from calibration set spectrum, and is minimized and is constituted set l, l=min (l1v,
l2v,,…,lqv), wherein 1~q indicates the number for being selected into calibration set sample, and v indicates the number of remaining sample to be selected；
(4) it chooses the corresponding sample of maximum value in l and enters calibration set, which is that distance has the farthest sample of calibration samples；
(5) it is repeated in and carries out step (3), (4), until calibration set number of samples reaches preset number m.
6. the solid wood board knot form inversion method according to claim 5 based on near-infrared spectrum analysis, feature
It is,
The wavelet neural network replaces the activation primitive of BP network hidden layer using wavelet basis function,
The input parameter of wavelet neural network is the characteristic wavelength point that defect sample point preferably goes out, and the prediction of wavelet neural network is defeated
It is out the inclination angle of defect point；It is x in input signal sequenceiWhen (i=1,2 ..., k), the hidden layer of wavelet neural network exports meter
Calculate formula are as follows:
In formula, h (j) is the output of j-th of node of hidden layer of wavelet neural network, wijFor wavelet neural network input layer and
The connection weight of hidden layer, hjFor wavelet basis function, bjFor hjShift factor, ajFor hjContraction-expansion factor；
Shown in the calculation formula of wavelet neural network output layer such as formula (5):
In formula, wikFor the connection weight of hidden layer to output layer, h (i) is the output of i-th of hidden layer node, and l is hidden layer section
Points, p are output layer number of nodes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811265532.2A CN109064553B (en) | 2018-10-26 | 2018-10-26 | Solid wood board section morphology inversion method based on near infrared spectrum analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811265532.2A CN109064553B (en) | 2018-10-26 | 2018-10-26 | Solid wood board section morphology inversion method based on near infrared spectrum analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109064553A true CN109064553A (en) | 2018-12-21 |
| CN109064553B CN109064553B (en) | 2023-07-07 |
Family
ID=64767565
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811265532.2A Expired - Fee Related CN109064553B (en) | 2018-10-26 | 2018-10-26 | Solid wood board section morphology inversion method based on near infrared spectrum analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109064553B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109870421A (en) * | 2019-03-27 | 2019-06-11 | 东北林业大学 | An incremental method for classification and identification of wood species based on visible light/near-infrared spectral analysis |
| CN110009089A (en) * | 2019-03-15 | 2019-07-12 | 重庆科技学院 | An intelligent design modeling and decision parameter optimization method for autism hugging machine based on PLS-PSO neural network |
| CN110489838A (en) * | 2019-08-08 | 2019-11-22 | 东南大学 | A kind of unseasoned lumber thermal parameter inversion method based on Bayesian inference |
| CN113049500A (en) * | 2021-03-19 | 2021-06-29 | 杭州海康威视数字技术股份有限公司 | Water quality detection model training and water quality detection method, electronic equipment and storage medium |
| CN113496490A (en) * | 2021-09-06 | 2021-10-12 | 南通弈驰新型建材科技有限公司 | Wood board surface defect detection method and system based on computer vision |
| CN114577671A (en) * | 2022-03-17 | 2022-06-03 | 东北林业大学 | A near-infrared wood density detection method based on parameter correction and transfer learning |
| CN116843998A (en) * | 2023-08-29 | 2023-10-03 | 四川省分析测试服务中心 | Spectrum sample weighting method and system |
| CN117664909A (en) * | 2024-01-30 | 2024-03-08 | 上海优珏新型装饰材料有限公司 | Thermal cycle plate nondestructive testing method based on near infrared spectrum |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020025061A1 (en) * | 2000-08-23 | 2002-02-28 | Leonard Metcalfe | High speed and reliable determination of lumber quality using grain influenced distortion effects |
| US6597761B1 (en) * | 2001-02-23 | 2003-07-22 | Invision Technologies, Inc. | Log evaluation using cylindrical projections |
| CN101952712A (en) * | 2007-08-28 | 2011-01-19 | Posco公司 | Device and method for optically detecting surface defect of round wire rod |
| CN104616275A (en) * | 2013-11-04 | 2015-05-13 | 北京兆维电子(集团)有限责任公司 | Defect detecting method and defect detecting device |
| CN105608695A (en) * | 2015-12-22 | 2016-05-25 | 北京光电技术研究所 | Detection method and system of building surface defects |
| US20180059318A1 (en) * | 2011-03-09 | 2018-03-01 | Flex Lighting Ii, Llc | Film-based light fixture with light reflecting layer and fastener |
-
2018
- 2018-10-26 CN CN201811265532.2A patent/CN109064553B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020025061A1 (en) * | 2000-08-23 | 2002-02-28 | Leonard Metcalfe | High speed and reliable determination of lumber quality using grain influenced distortion effects |
| US6597761B1 (en) * | 2001-02-23 | 2003-07-22 | Invision Technologies, Inc. | Log evaluation using cylindrical projections |
| CN101952712A (en) * | 2007-08-28 | 2011-01-19 | Posco公司 | Device and method for optically detecting surface defect of round wire rod |
| US20180059318A1 (en) * | 2011-03-09 | 2018-03-01 | Flex Lighting Ii, Llc | Film-based light fixture with light reflecting layer and fastener |
| CN104616275A (en) * | 2013-11-04 | 2015-05-13 | 北京兆维电子(集团)有限责任公司 | Defect detecting method and defect detecting device |
| CN105608695A (en) * | 2015-12-22 | 2016-05-25 | 北京光电技术研究所 | Detection method and system of building surface defects |
Non-Patent Citations (2)
| Title |
|---|
| PABLO GUINDOS ET AL.: "A three-dimensional wood material model to simulate the behavior of wood with any type of knot at the macro-scale", 《WOOD SCIENCE AND TECHNOLOGY》 * |
| 张瑞滨: "板材节子视频检测与计算机再现理论研究", 《中国优秀博硕士学位论文全文数据库 (硕士) 信息科技辑》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009089A (en) * | 2019-03-15 | 2019-07-12 | 重庆科技学院 | An intelligent design modeling and decision parameter optimization method for autism hugging machine based on PLS-PSO neural network |
| CN109870421A (en) * | 2019-03-27 | 2019-06-11 | 东北林业大学 | An incremental method for classification and identification of wood species based on visible light/near-infrared spectral analysis |
| CN109870421B (en) * | 2019-03-27 | 2021-04-06 | 东北林业大学 | Incremental wood tree species classification and identification method based on visible light/near infrared spectrum analysis |
| CN110489838A (en) * | 2019-08-08 | 2019-11-22 | 东南大学 | A kind of unseasoned lumber thermal parameter inversion method based on Bayesian inference |
| CN113049500A (en) * | 2021-03-19 | 2021-06-29 | 杭州海康威视数字技术股份有限公司 | Water quality detection model training and water quality detection method, electronic equipment and storage medium |
| CN113049500B (en) * | 2021-03-19 | 2022-12-06 | 杭州海康威视数字技术股份有限公司 | Water quality detection model training and water quality detection method, electronic equipment and storage medium |
| CN113496490B (en) * | 2021-09-06 | 2021-11-09 | 南通弈驰新型建材科技有限公司 | Wood board surface defect detection method and system based on computer vision |
| CN113496490A (en) * | 2021-09-06 | 2021-10-12 | 南通弈驰新型建材科技有限公司 | Wood board surface defect detection method and system based on computer vision |
| CN114577671A (en) * | 2022-03-17 | 2022-06-03 | 东北林业大学 | A near-infrared wood density detection method based on parameter correction and transfer learning |
| CN116843998A (en) * | 2023-08-29 | 2023-10-03 | 四川省分析测试服务中心 | Spectrum sample weighting method and system |
| CN116843998B (en) * | 2023-08-29 | 2023-11-14 | 四川省分析测试服务中心 | Spectrum sample weighting method and system |
| CN117664909A (en) * | 2024-01-30 | 2024-03-08 | 上海优珏新型装饰材料有限公司 | Thermal cycle plate nondestructive testing method based on near infrared spectrum |
| CN117664909B (en) * | 2024-01-30 | 2024-04-16 | 上海优珏新型装饰材料有限公司 | Thermal cycle plate nondestructive testing method based on near infrared spectrum |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109064553B (en) | 2023-07-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109064553A (en) | Solid wood board knot form inversion method based on near-infrared spectrum analysis | |
| CN101915744B (en) | Near infrared spectrum nondestructive testing method and device for material component content | |
| Jie et al. | Review on the recent progress of non-destructive detection technology for internal quality of watermelon | |
| CN111968080A (en) | Hyperspectrum and deep learning-based method for detecting internal and external quality of Feicheng peaches | |
| CN109253985B (en) | Method for identifying wood grade for koto panel by near infrared spectrum based on neural network | |
| CN109324013A (en) | A method for constructing near-infrared fast analysis of crude oil properties using Gaussian process regression model | |
| CN106248621B (en) | A kind of evaluation method and system | |
| CN105424653B (en) | The fruit pulp tissue optical property detecting system and method popped one's head in integrated optical fiber | |
| He et al. | Fast discrimination of apple varieties using Vis/NIR spectroscopy | |
| CN111474137A (en) | Citrus sugar degree nondestructive testing method based on convolutional neural network | |
| CN113267466B (en) | Non-destructive detection method for sugar content and acidity of fruit based on spectral wavelength optimization | |
| CN108693139A (en) | The near infrared prediction model method for building up of electronics tobacco tar physical and chemical index and application | |
| CN111537469A (en) | A rapid non-destructive testing method for apple quality based on near-infrared technology | |
| CN109142251B (en) | LIBS Quantitative Analysis Method of Random Forest Aided Artificial Neural Network | |
| CN110186871A (en) | A kind of method of discrimination in the fresh tea leaves place of production | |
| CN108827909B (en) | A rapid soil classification method based on visible-near-infrared spectroscopy and multi-object fusion | |
| CN107290299B (en) | Method for detecting sugar degree and acidity of peaches in real time in nondestructive mode | |
| CN105738311A (en) | Apple sweetness non-damage quick detection method based on near-infrared spectrum technology | |
| CN109961179A (en) | A kind of aquatic product quality detection method and portable Raman device | |
| CN114913406A (en) | A method for estimating chlorophyll content based on inversion of optical property parameters | |
| Chen et al. | Prediction of apple internal qualities using hyperspectral imaging techniques | |
| CN110567385A (en) | Construction thickness detection method of building reflective heat insulation coating based on hyperspectral technology | |
| Jin et al. | An innovative fusion feature method of spectrum and visual image for diagnosing ‘Akizuki’pear cork spot disorder | |
| CN110186870A (en) | A kind of method that extreme learning machine spectral model differentiates the beautiful dew fresh tea leaves place of production of bestowing favour | |
| CN113418889B (en) | Real-time detection method for water content and total colony count of dried vegetables based on deep learning |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20230707 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |