A wood breakage rate image measurement method based on dyeing technology and semantic segmentation

Abstract

The application discloses a kind of wood breakage rate image measurement methods based on dyeing technology and semantic segmentation, it includes the following steps: S1, dyeing adhesive is bonded into two pieces of wood and is carried out shear failure along grain shear test specimen, and S2, different types of wood breakage rate image are collected, original image is manually annotated, and wood breakage rate dataset is made;S3, the wood breakage rate image dataset is trained and predicted by semantic segmentation, and wood breakage rate prediction image is automatically generated;S4, wood breakage rate automatic measurement is carried out to wood breakage rate prediction image, the product of wood breakage rate of the two adhesive surfaces of sample and respective training weight, namely, the wood breakage rate of sample.The application can automatically identify wood damage area, quickly realize wood breakage rate measurement, and can improve measurement accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wood science, furniture structure, building structure and the like, in particular to a wood breakage rate image measurement method based on dyeing technology and semantic segmentation. BACKGROUND

[0002] Wood breakage rate is an important index for characterizing the gluing performance of glued wood beams and wood-based integrated materials. At present, wood breakage rate measurement is mainly evaluated by visual inspection. This method is low in efficiency, slow in speed, and has large differences in measurement results. In order to improve the measurement accuracy, a scientific wood breakage rate image measurement method is in urgent need.

[0003] There are methods for quickly measuring wood breakage rate in the field. Among them, patent (method for quickly detecting wood breakage rate) application number: 201410353883.4 discloses a method for quickly detecting wood breakage rate, which can be used to detect the shear surface area of the test piece damaged after detecting the gluing strength of wood or plywood. Compared with the present application, this method mainly converts the image into a gray-scale image, then performs binary segmentation thresholding (maximum inter-class variance method), and then calculates the wood breakage rate in the gluing strength test process through image type conversion, image enhancement, image segmentation and morphological processing operations. This method does not classify each pixel in the wood breakage rate image, does not correspond each pixel with its represented category, and only measures one gluing surface, resulting in large measurement error. Secondly, when the color of the adhesive is similar to the color of the wood, this method cannot be measured. Finally, with the vigorous development of composite materials, different types of wood are bonded by adhesives to be used as furniture and building materials. The color of such composite materials also has great difference. The more obvious the color difference between the two, the more conducive to computer measurement. The smaller the color difference, the less conducive to computer measurement. Therefore, there is a great difference in simply using the average value as the final evaluation index.

[0004] Therefore, in the field of wood processing, it is necessary to find a more scientific and reasonable wood breakage rate image measurement method. SUMMARY

[0005] In view of the above shortcomings in the prior art, the wood breakage rate image measurement method based on dyeing technology and semantic segmentation provided by the present application solves the problem of difficulty in existing wood breakage rate measurement.

[0006] In order to achieve the above-mentioned application purposes, the technical scheme adopted by the present application is as follows:

[0007] The present application provides a wood breakage rate image measurement method based on dyeing technology and semantic segmentation, which comprises the following steps:

[0008] S1, dyeing the adhesive (2b) to bond two pieces of wood (2a) into a parallel grain shear test sample, and performing shear failure;

[0009] S2, collect different types of wood breakage rate images, manually label the original images, and make a wood breakage rate dataset;

[0010] S3, train and predict the wood breakage rate image dataset by semantic segmentation, and automatically generate wood breakage rate prediction images;

[0011] S4, automatically measure the wood breakage rate of the wood breakage rate prediction image, and the sum of the wood breakage rate of the two glued surfaces of the sample and the respective training weight is the wood breakage rate of the sample.

[0012] Further, the S1 order shear sample is a wood and wood, bamboo and bamboo, bamboo and wood furniture and building plate adhesive strength test order shear sample processed according to the requirements of industry standard, and there is a fiber tear phenomenon of adhesive material;

[0013] Further, the type of dyed adhesive in S1 can be one of isocyanate adhesive, phenolic resin adhesive, urea-formaldehyde resin adhesive, and the color agent can be one of organic, inorganic or water-oil compatible, and the color of the dyed adhesive can be one of black, brown, blue, red;

[0014] Further, the wood breakage rate image in S2 is collected by electronic digital device and image acquisition device for shooting image or video. The wood breakage rate image can be one or more of wet state, dry state, color, and gray. The wood breakage rate image can be one or more of shallow or deep layer. The angle between the collected wood breakage rate image order shear sample and the lens of the image acquisition device is 0°-180°.

[0015] Further, the enhanced feature extraction network in semantic segmentation in S3 directly performs two times up-sampling when up-sampling, and finally obtains a feature layer with the same height and width as the input picture.

[0016] Compared with the prior art, the advantages of the present technology are:

[0017] Semantic segmentation was chosen as the method for measuring wood breakage rate. Experiments showed that it significantly improved upon methods such as visual inspection, Photoshop, and Matlab in reducing subjective factors. The grain-cut specimens, bonded with pigment-stained adhesive, exhibited a large color difference and clear boundary between the adhesive and the damaged wood area, facilitating rapid wood breakage rate measurement. The enhanced feature extraction network in semantic segmentation directly performs double upsampling during upsampling, resulting in feature layers with the same height and width as the input image, improving the network's versatility. The weighted average obtained by assigning entropy values ​​to different full-values ​​of the grain-cut specimens reduces the significant error introduced by the average value. The wood breakage rate image acquisition equipment ensures consistent image quality. In summary, the beneficial effects of this invention are:

[0018] 1. High efficiency: This invention directly dyes the adhesive, which reduces measurement time and lowers measurement costs compared to the traditional method of dyeing both the adhesive and the substrate.

[0019] 2. Scientific validity: Semantic segmentation reduces the impact of human factors on image acquisition when measuring wood breakage rate; the wood breakage rate image acquisition device reduces the impact of environmental factors such as angle and lighting on image acquisition; thus greatly improving the scientific validity of wood breakage rate measurement.

[0020] 3. Rationality: Semantic segmentation directly performs a 2x upsampling during upsampling, resulting in a feature layer with the same height and width as the input image, facilitating network construction and improving versatility. Entropy weighting assigns different weights to the two glued surfaces; the sum of the products of the wood breakage rates of the two glued surfaces and each weight value represents the final wood breakage rate of a shear test specimen along the grain. This avoids the limitations of traditional methods that use the wood breakage rate of one glued surface or the average of the wood breakage rates of both glued surfaces as the final evaluation metric.

[0021] It effectively extracts low-dimensional and high-dimensional image features to accurately identify wood damage areas and calculate wood breakage rate. It is applicable to the measurement of wood breakage rate of different types of glued plywood, such as materials, adhesives, and glued areas. It saves labor costs and labor, reduces errors caused by human measurement, and improves work efficiency and measurement accuracy. Attached Figure Description

[0022] Figure 1 This invention provides a flowchart of a wood breakage rate image measurement method based on staining technology and semantic segmentation, as an embodiment of the present invention.

[0023] Figure 2 The parallel shear specimen provided in the embodiment of the present invention.

[0024] Figure 3is a wood failure rate image of a parallel shear specimen collected in Example 1 of the method according to the present application, a manually annotated image and a prediction result thereof.

[0025] Figure 4 is a wood failure rate image of a parallel shear specimen collected in Example 2 of the method according to the present application, a manually annotated image and a prediction result thereof.

[0026] Figure 5 is a wood failure rate image of a parallel shear specimen collected in Example 3 of the method according to the present application, a manually annotated image and a prediction result thereof.

[0027] Figure 6 is a wood failure rate image of a parallel shear specimen collected in Comparative Example 1 of the method according to the present application, a manually annotated image and a prediction result thereof. DETAILED DESCRIPTION

[0028] The specific embodiments of the present application will be described below in detail with reference to the accompanying drawings of the embodiments of the present application, but it should be understood that the scope of protection of the present application is not limited by the specific embodiments.

[0029] Unless otherwise explicitly indicated, throughout the specification and claims, the term "comprise" or its variants such as "comprises" or "comprising" will be understood to encompass the stated element or elements, but not to exclude the presence of other elements or additional elements.

[0030] As shown in Figure 1 , a wood failure rate image measurement method based on dyeing technology and semantic segmentation includes the following steps:

[0031] S1, a dyeing adhesive is used to bond two pieces of wood into a parallel shear specimen, and shear failure is performed;

[0032] S2, different types of wood failure rate images are collected, the original images are manually annotated, and a wood failure rate dataset is prepared;

[0033] S3, the wood failure rate image dataset is trained and predicted by semantic segmentation, and wood failure rate prediction images are automatically generated;

[0034] S4, wood failure rate prediction images are subjected to wood failure rate automatic measurement, and the sum of the wood failure rates of the two glued surfaces of the specimen multiplied by the respective training weights is the wood failure rate of the specimen.

[0035] Example 1

[0036] The implementation method of this embodiment is as described above, and the specific steps will not be described in detail, and only the case implementation conditions and case data will be used to show the effect.

[0037] The carbonized bamboo and larch dyed by FC are pressed into the parallel shear samples by the polymer waterborne isocyanate, and the samples are subjected to shear failure under wet conditions.

[0038] Example 2

[0039] The wood failure rate is measured according to the method of Example 1, except that the parallel shear sample is subjected to shear failure under dry conditions.

[0040] Example 3

[0041] The wood failure rate is measured according to the method of Example 1, except that the base material of the parallel shear sample is larch and larch.

[0042] Comparative Example 1

[0043] The wood failure rate is measured according to the method of Example 3, except that the polymer waterborne isocyanate is not subjected to dyeing treatment.

[0044] Comparative Example 2

[0045] After measuring the wood failure rate of the two glued surfaces of the parallel shear sample by the caliper measurement method, the average value is the final wood failure rate of the sample.

[0046] Table 1 Comparison of wood failure rate test results by different methods

[0047] Number A face B face Sum Comparative Example 2 Absolute error Relative error Example 1 40.80％ 42.29％ 41.54％ 41％ 0.54％ 1.3％ Example 2 86.68％ 90.17％ 88.42％ 87％ 1.42％ 1.6％ Example 3 76.60％ 78.28％ 77.77％ 78％ 0.23％ 0％ Comparative Example 1 99.10％ 98.9％ 99％ 75％ 24％ 32％

[0048] As can be seen from the results in Table 1, the present application can effectively measure the wood failure rate and reduce the measurement error (absolute error, relative error).

[0049] In summary, the method of the present application can fully utilize the adhesive color technology and semantic segmentation to measure the wood failure rate, while the traditional image measurement method is greatly affected by the experimental conditions and has poor adaptability. On the other hand, the absolute error of the wood failure rate measured by the traditional image and the true value is large. The wood failure rate image measurement method based on dyeing technology and semantic segmentation proposed in the present application uses a semantic segmentation framework to adapt to complex problems in multiple scenarios, and the measurement value is closer to the true value, and the measurement error is small. The automatic, industrial and intelligent measurement of the wood failure rate is realized.

Claims

1. A wood breakage rate image measurement method based on dyeing technology and semantic segmentation, characterized by, The method comprises the following steps: S1, a dyed adhesive (2b) is used to bond two pieces of wood (2a) into a parallel grain shear specimen, and shear failure is performed; S2, different types of wood failure images are collected, the original images are manually labeled, and a wood failure dataset is prepared; S3, the wood failure image dataset is trained and predicted through semantic segmentation, and a wood failure prediction image is automatically generated; S4, the wood failure prediction image is automatically measured for wood failure, and the sum of the wood failure of the two glued surfaces of the specimen and the product of the respective training weights is the wood failure of the specimen.

2. The wood breakage rate image measurement method based on the dyeing technique and the semantic segmentation according to claim 1, characterized in that, The parallel grain shear specimen in S1 is a wood-to-wood, or bamboo-to-bamboo, or bamboo-to-wood furniture and building panel adhesive strength test parallel grain shear specimen processed according to industry standard requirements, and there is a fiber tear phenomenon of the adhesive; the dyed adhesive in S1 is an isocyanate adhesive, a phenol formaldehyde resin adhesive, or a urea formaldehyde resin adhesive, and the color agent is one of organic, inorganic, or water-oil compatible; the dyed adhesive is black, brown, blue, red, or other colors that are clearly distinguishable from wood.

3. The method according to claim 1, wherein the method is based on dyeing technique and semantic segmentation. The wood failure image in S2 is collected by an electronic digital device and an image acquisition device for shooting images or videos; wherein the wood failure image is one or more of wet, dry, color, and gray; the wood failure image has one or more of shallow or deep wood failure levels; the collected wood failure image parallel grain shear specimen and the image acquisition device lens angle are 0°-180°.

4. The method according to claim 1, wherein the method is based on dyeing technique and semantic segmentation. In S3, the enhanced feature extraction network in semantic segmentation directly performs two times up-sampling when up-sampling, and finally obtains a feature layer with the same height and width as the input image.

Images