A rapid detection system for wooden furniture materials

Abstract

The present application relates to the technical field of wood detection, and specifically discloses a rapid detection system for wooden furniture materials, which uses a portable spectral scanner to collect spectral scanning images of wooden furniture materials to be tested from multiple positions, and introduces spectral image processing technology based on deep learning, extracts spectral features from spectral scanning images at multiple positions, performs correlation analysis between spectral features at each position, and performs complementary information enhanced aggregation processing to construct a global spectral feature description of the wooden furniture materials to be tested, and then realizes rapid and accurate detection of wooden furniture materials based on matching analysis between spectral features of the wooden furniture materials to be tested and spectral features of known wood types. In this way, the limitations of traditional wood identification methods and the shortcomings of existing spectral detection methods can be effectively overcome, and the accuracy and efficiency of wooden furniture material detection can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of wood detection, and more specifically, to a rapid detection system for wooden furniture materials. Background Art

[0002] As an important part of home decoration, wooden furniture is widely favored by consumers for its natural beauty, durability and environmental protection. With the continuous expansion of the market and the improvement of consumers' quality requirements, the identification and detection of wooden furniture materials are of great significance to ensure product quality, prevent counterfeit and inferior products from entering the market, and satisfy consumers' right to know. Traditional wood identification methods rely on microscopic observation, chemical analysis or manual judgment based on physical characteristics such as density and texture. However, manual observation methods are highly dependent on the experience and subjective judgment of the inspectors, and are easily affected by personal skill levels, fatigue and environmental factors, resulting in instability and uncertainty of test results. Although physical property testing and chemical analysis methods can provide more objective data support, they usually require sample destruction, are time-consuming and costly, and are not suitable for rapid detection or large-scale screening scenarios.

[0003] With the development of science and technology, spectral analysis, as a non-destructive and rapid detection method, has shown great potential in the field of material composition analysis and type identification. Spectral analysis technology can obtain information about the internal composition and structure of a material by measuring the absorption, reflection or transmission characteristics of light, thereby realizing non-destructive and rapid detection of the material. In the field of wooden furniture material detection, spectral analysis technology has initially shown its application prospects, but most of the existing spectral detection methods are based on single-point measurement and simple statistical analysis of spectral images, which is difficult to fully and accurately reflect the complexity and diversity of the internal composition and structure of wooden furniture materials, especially in the presence of uneven distribution or surface treatment, which may lead to misjudgment.

[0004] Therefore, an optimized rapid detection system for wooden furniture materials is expected. Summary of the invention

[0005] In order to solve the above-mentioned technical problems, the present application is proposed. An embodiment of the present application provides a rapid detection system for wooden furniture materials, which uses a portable spectral scanner to collect spectral scanning images of wooden furniture materials to be tested from multiple positions, and introduces spectral image processing technology based on deep learning, and performs spectral feature extraction, correlation analysis between spectral features of each position, and complementary information enhanced aggregation processing on the spectral scanning images of multiple positions, so as to construct a global spectral feature description of the wooden furniture materials to be tested, and then based on the matching analysis between the spectral features of the wooden furniture materials to be tested and the spectral features of known wood types, rapid and accurate detection of wooden furniture materials can be achieved. In this way, the limitations of traditional wood identification methods and the shortcomings of existing spectral detection methods can be effectively overcome, and the accuracy and efficiency of wooden furniture material detection can be improved.

[0006] Accordingly, according to one aspect of the present application, a rapid detection system for wooden furniture materials is provided, comprising:

[0007] A multi-position spectral scanning image acquisition module, used to acquire spectral scanning images of multiple positions of the wooden furniture material to be tested collected by a portable spectral scanner;

[0008] A multi-position spectral feature encoding module, used for calculating the spectral feature encoding vector of the wooden furniture material to be tested based on the spectral scanning images of the multiple positions;

[0009] A reference feature extraction module is used to extract a set of spectral feature encoding vectors of known wood types from a wood material spectral feature database;

[0010] The query matching module is used to query and match the spectral feature coding vector of the wooden furniture material to be tested with the set of spectral feature coding vectors of the known wood types to obtain a matching analysis result, wherein the matching analysis result is the closest wood type.

[0011] Compared with the prior art, the rapid detection system for wooden furniture materials provided by the present application uses a portable spectral scanner to collect spectral scanning images of wooden furniture materials to be tested from multiple positions, and introduces spectral image processing technology based on deep learning to extract spectral features of spectral scanning images at multiple positions, perform correlation analysis between spectral features at each position, and perform complementary information enhanced aggregation processing to construct a global spectral feature description of the wooden furniture materials to be tested, and then achieve rapid and accurate detection of wooden furniture materials based on matching analysis between spectral features of wooden furniture materials to be tested and spectral features of known wood types. In this way, the limitations of traditional wood identification methods and the shortcomings of existing spectral detection methods can be effectively overcome, and the accuracy and efficiency of wooden furniture material detection can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0013] Figure 1 4 is a block diagram of a rapid detection system for wooden furniture materials according to an embodiment of the present application.

[0014] Figure 2 Schematic diagram of data flow of a rapid detection system for wooden furniture materials according to an embodiment of the present application.

[0015] Figure 3 It is a block diagram of a multi-position spectral feature encoding module in a rapid detection system for wooden furniture materials according to an embodiment of the present application.

[0016] Figure 4 It is a block diagram of a feature dynamic compensation aggregation unit in a rapid detection system for wooden furniture materials according to an embodiment of the present application.

[0017] Figure 5 4 is a block diagram of a significance scoring subunit in a rapid detection system for wooden furniture materials according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0019] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0020] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0021] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0022] It is worth noting that in this application, all actions to obtain data are carried out in compliance with the relevant data protection laws and policies of the country where the data is located, and with the authorization given by the owner of the corresponding device.

[0023] As described in the above background technology, with the growth of market demand for wooden furniture and the improvement of quality standards, accurate identification and detection of wooden furniture materials are particularly important for ensuring product quality, preventing defective products from entering the market, and protecting consumers' right to know. Traditional wood identification methods usually rely on microscopic inspection, chemical testing, or subjective evaluation based on physical properties such as density and texture. This manual observation method is highly dependent on the experience and personal judgment of the inspector and is easily affected by personal skills, fatigue and working environment, thus affecting the consistency and reliability of the results. Although physical property tests and chemical analysis can provide more objective data support, they often require the destruction of samples, which is time-consuming and costly, and is not suitable for rapid detection or large-scale screening.

[0024] Based on this, spectral analysis, as a non-invasive and rapid detection technology, has shown great potential in component analysis and species identification. This technology obtains information about the internal composition and structure of a substance by measuring its absorption, reflection or transmission characteristics of light, thus achieving a non-destructive and efficient detection process. In the detection of wooden furniture materials, spectral analysis technology has shown initial application prospects.

[0025] Traditional spectral analysis often uses single-point measurement to collect data. Usually a specific location is selected, where a spectrometer is used to emit light within a specific wavelength range to illuminate the wood surface and record the information of the reflected light. Due to the heterogeneity and complexity of the wood itself, different parts may have significantly different physical and chemical properties. Single-point measurement can only capture information from a certain local area and it is difficult to fully represent the overall characteristics of the entire sample. In addition, many wooden furniture undergoes surface treatments such as painting and polishing, which change the optical properties of the material surface, so that the results of single-point measurement may not truly reflect the characteristics of untreated wood. In addition, although single-point measurement is easy to implement, it lacks sensitivity to changes in the internal structure of the material, especially when facing multi-layered or thick boards. This limitation is more obvious.

[0026] To make up for the shortcomings of single-point measurement, traditional methods also rely on simple spectral image statistical analysis to process the acquired data. This type of analysis focuses on averaging multiple single-point measurement results in an attempt to reduce the impact of random errors and obtain more stable results. However, this approach can easily mask subtle differences that are critical for distinguishing different wood types. For example, by looking for absorption peaks or reflection valleys on the spectral curve and comparing them using their position, intensity, etc. as characteristic parameters. Although this method is intuitive and easy to understand, it may be interfered by noise and cause misjudgment in practical applications. Based on the spectral feature template of the known wood type, the most similar sample is found through visual observation or simple algorithms (such as Euclidean distance). This method is inefficient and has limited accuracy, especially when faced with complex and changeable actual samples, it is prone to matching failures.

[0027] In response to the above technical problems, this application proposes an optimized rapid detection system for wooden furniture materials, which uses a portable spectral scanner to collect spectral scanning images of wooden furniture materials to be tested from multiple positions, and introduces spectral image processing technology based on deep learning to extract spectral features of spectral scanning images at multiple positions, analyze the correlation between spectral features at each position, and perform complementary information enhanced aggregation processing to construct a global spectral feature description of the wooden furniture materials to be tested, and then achieve rapid and accurate detection of wooden furniture materials based on the matching analysis between the spectral features of the wooden furniture materials to be tested and the spectral features of known wood types. In this way, the limitations of traditional wood identification methods and the shortcomings of existing spectral detection methods can be effectively overcome, and the accuracy and efficiency of wooden furniture material detection can be improved.

[0028] Figure 1 4 is a block diagram of a rapid detection system for wooden furniture materials according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the rapid detection system for wooden furniture materials according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the wooden furniture material rapid detection system 100 includes: a multi-position spectral scanning image acquisition module 110, which is used to acquire spectral scanning images of multiple positions of the wooden furniture material to be tested collected by a portable spectral scanner; a multi-position spectral feature encoding module 120, which is used to calculate the spectral feature encoding vector of the wooden furniture material to be tested based on the spectral scanning images of the multiple positions; a reference feature extraction module 130, which is used to extract a set of spectral feature encoding vectors of known wood types from a wooden material spectral feature database; and a query matching module 140, which is used to query and match the spectral feature encoding vector of the wooden furniture material to be tested with the set of spectral feature encoding vectors of the known wood types to obtain a matching analysis result, wherein the matching analysis result is the closest wood type.

[0029] In the above-mentioned rapid detection system for wooden furniture materials, the multi-position spectral scanning image acquisition module 110 is used to acquire spectral scanning images of multiple positions of the wooden furniture material to be tested collected by the portable spectral scanner. It should be understood that the portable spectral scanner emits light of a specific wavelength and illuminates the surface of the wooden furniture material. Since the material has different absorption and reflection characteristics for light of different wavelengths, the instrument captures the reflected light and generates a spectral scanning image according to the change in the characteristics of the light, thereby being able to reflect the chemical composition and structural information of the material. In particular, since the wood material may have certain differences in different parts of the furniture, it is difficult to fully represent the overall material characteristics by only acquiring spectral information from a single position. Therefore, the present application acquires spectral scanning images of the wooden furniture material to be tested from multiple positions, thereby fully acquiring spectral information of each part of the furniture material, so as to reduce detection errors caused by local differences and improve detection accuracy.

[0030] Since wood is a natural product, its internal structure and chemical composition may vary due to factors such as the growth environment and the distribution of annual rings. Even on the same piece of wood, different parts may exhibit different physical and chemical characteristics. For example, the part near the center of the trunk is usually harder and denser than the edge; the processed surface may also have coatings or other treatments, which further affect the spectral characteristics. Therefore, it is difficult to fully represent the material properties of the entire sample by obtaining spectral information from only a single location, which may lead to large deviations in the test results. To solve this problem, a multi-location sampling strategy can be adopted to ensure that as many representative areas as possible are covered to reduce errors caused by local differences and improve overall detection accuracy.

[0031] In order to collect spectral scanning images of multiple positions of the wooden furniture material to be tested, first of all, the purpose and scope of the test must be clear, which will affect the sampling location, quantity and specific parameter settings of the spectral analysis. In terms of equipment selection, near-infrared (NIR) spectroscopy technology is a good choice for wood materials, which can provide information about the internal structure and composition of wood without the need for destructive treatment of samples. After selecting a suitable portable spectral scanner, it is also necessary to ensure that the instrument has been calibrated to ensure the accuracy of the measurement results. In addition, all auxiliary tools, such as cleaning cloths, fixing fixtures, etc., should be prepared to ensure that foreign contaminants are not introduced or materials are moved during the scanning process.

[0032] Secondly, it is necessary to select appropriate sampling points. Considering the overall design and functional use of the furniture, surfaces that are easily exposed and accessible should be given priority as sampling points. At the same time, it is also necessary to avoid obvious defects such as cracks and holes, because the spectral characteristics of these places often cannot reflect the state of normal wood. Considering the heterogeneity of the wood itself, it is recommended to sample at different levels (such as the surface, middle, and deep layers) in order to better understand the vertical distribution characteristics of the material. For large or complex-shaped furniture, it can also be divided into several sub-areas according to its geometric form, and several representative points can be selected in each sub-area.

[0033] Next comes the sample preparation stage. Wooden furniture usually needs to be cleaned to remove dust, dirt or other substances that may interfere with the spectral signal. This step is crucial to obtain a clear and accurate spectral image. At the same time, the size and shape of the furniture should be considered and the scanning path should be planned reasonably to ensure that a representative area can be covered. When it comes to large or complex-shaped furniture, a simple diagram can be drawn in advance to mark the planned scanning points so that data collection can be carried out systematically.

[0034] Then comes the actual acquisition process, where the operator places the portable spectral scanner at the predetermined position and then starts the scanning procedure. Modern spectrometers are usually equipped with an automatic focusing function, which can obtain clear spectral images at different distances. Each scan records the intensity of reflected or transmitted light in a specific wavelength range, which is converted into a digital signal and stored. Due to the inhomogeneity of wood and the local property changes that may be caused by surface treatment, it is recommended to perform multiple repetitive scans at each selected position to improve the reliability and representativeness of the data. In addition, repeating the above steps under different lighting conditions can help verify the consistency of the results. During the acquisition process, attention should also be paid to the influence of environmental conditions. Factors such as temperature and humidity may change the physical properties of wood and thus affect the spectral characteristics. Therefore, try to maintain a constant working environment, or at least record the environmental parameters at the time for reference in subsequent data analysis.

[0035] Finally, as the data from each position gradually accumulates, a multi-position spectral scanning image collection can be constructed. This collection not only contains detailed spectral information from a single position, but also reflects the similarities and differences in the spatial distribution of the entire sample. The data obtained in this way provides a solid foundation for the subsequent multi-position spectral feature encoding module, enabling a more comprehensive and accurate reflection of the complexity and diversity of the internal composition and structure of wooden furniture materials, thereby providing strong support for accurate identification and classification.

[0036] In the above-mentioned rapid detection system for wooden furniture materials, the multi-position spectral feature encoding module 120 is used to calculate the spectral feature encoding vector of the wooden furniture material to be tested based on the spectral scanning images of the multiple positions. It should be understood that due to the huge and complex amount of original spectral scanning image data, it is difficult for traditional spectral image processing technology to directly extract effective spectral features therefrom. To this end, in the technical solution of the present application, a spectral image processing technology based on deep learning is introduced, which performs feature extraction and correlation analysis on the spectral scanning images collected at multiple positions to construct a global spectral feature description of the wooden furniture material to be tested, and generate the spectral feature encoding vector of the wooden furniture material to be tested. Among them, Figure 3 FIG. 1 is a block diagram of a multi-position spectral feature encoding module in a rapid detection system for wooden furniture materials according to an embodiment of the present application. Figure 3 As shown, the multi-position spectral feature encoding module 120 includes: a spectral feature extraction unit 121, used to extract the spectral features of the spectral scanning image at each position respectively to obtain multiple position spectral feature encoding vectors; a feature dynamic compensation aggregation unit 122, used to perform attention-driven feature dynamic compensation aggregation on the multiple position spectral feature encoding vectors to obtain the spectral feature encoding vector of the wooden furniture material to be tested.

[0037] Specifically, in a specific example of the present application, the spectral feature extraction unit 121 is used to: input the spectral scanning images of the various positions into a spectral feature extractor based on the FPT model to obtain the spectral feature encoding vectors of the multiple positions. It should be understood that, considering that there are usually multi-scale and multi-level spectral information in the spectral scanning image, such as the texture distribution of the material surface, the spectral reflectivity and absorptivity of different bands, etc. Therefore, in order to comprehensively and accurately extract the spectral information of the spectral scanning images of various positions, the present application adopts the FPT (Feature Pyramid Transformer) model as the spectral feature extractor. Those skilled in the art should know that the FPT model is a Transformer-based feature pyramid network structure, which combines the advantages of the feature pyramid structure and the Transformer architecture. It can extract features from spectral scanning images at different scales, capture spectral feature information at different levels in the image, and use the self-attention mechanism of the Transformer architecture to model the correlation and dependency between spectral features at different levels on a global scale, thereby achieving fully active feature interaction across space and scale, thereby effectively extracting a more comprehensive and accurate spectral feature representation, comprehensively reflecting the internal composition and structure of wooden furniture materials, and generating multiple position spectral feature encoding vectors.

[0038] Specifically, the feature dynamic compensation aggregation unit 122 is used to perform attention-driven feature dynamic compensation aggregation on the multiple position spectral feature coding vectors to obtain the spectral feature coding vector of the wooden furniture material to be tested. It should be understood that the multiple position spectral feature coding vectors respectively represent the spectral information of different positions of the furniture material, but due to the complexity and unevenness of the wood material, there may be differences and complementarities between the multiple position spectral feature coding vectors. Therefore, in order to construct a global and accurate spectral feature description, the present application proposes an attention-driven feature dynamic compensation aggregation method, which mines the representative core information in the multiple position spectral feature coding vectors to construct a feature compensation-aggregation mechanism based on the core information to dynamically adjust and optimize the complementary relationship between the spectral features of each position, thereby realizing the significance enhancement aggregation between features. Among them, Figure 4 FIG. 1 is a block diagram of a characteristic dynamic compensation aggregation unit in a rapid detection system for wooden furniture materials according to an embodiment of the present application. Figure 4 As shown, the feature dynamic compensation aggregation unit 122 includes: a hub feature extraction subunit 1221, which is used to input the multiple position spectral feature coding vectors into the hub feature extraction network to obtain a multi-position spectral hub feature coding vector; a significance scoring subunit 1222, which is used to perform significance scoring on each position spectral feature coding vector based on the complementary information of each position spectral feature coding vector in the multiple position spectral feature coding vectors relative to the multi-position spectral hub feature coding vector to obtain the attention weights of the complementary information of the spectral features of multiple position nodes; a feature compensation aggregation subunit 1223, which is used to perform dynamic compensation aggregation on the multiple position spectral feature coding vectors based on the attention weights of the complementary information of the spectral features of the multiple position nodes to obtain the spectral feature coding vector of the wooden furniture material to be tested.

[0039] More specifically, the pivot feature extraction subunit 1221 is expressed as follows:

[0040]

[0041] in, represents the hub feature extraction network, represents a set of spectral feature encoding vectors of the plurality of positions, , , and respectively represent the first, second, and third positions in the set of spectral feature encoding vectors. and The spectral feature encoding vector of each position, is the number of spectral feature encoding vectors at the position, and Respectively represent the weight parameter matrix and bias term of the hub feature extraction network, represents the hub feature relevance score transformation vector of the hub feature extraction network, Indicates the The corresponding hub feature relevance scoring factor, represents matrix multiplication, represents the normalized exponential function, Indicates the The corresponding normalized hub feature relevance score factor, Represents the multi-position spectral hub feature encoding vector.

[0042] That is, a hub feature extraction network is constructed through a specific neural network structure, and the hub feature extraction network can perform hub feature correlation analysis on the spectral feature encoding vectors at each position to learn the intrinsic dependency and correlation pattern of the initial spectral feature distribution of the wooden furniture material to be tested, so that the most representative spectral feature representation in the spectral feature encoding vectors at multiple positions, i.e., the multi-position spectral hub feature encoding vector, can be identified and extracted. In this way, a more refined and meaningful data representation can be provided for subsequent analysis.

[0043] Figure 5 FIG. 4 is a block diagram of a significance scoring subunit in a rapid detection system for wooden furniture materials according to an embodiment of the present application. Figure 5 As shown, the significance scoring subunit 1222 includes: a complementary information extraction secondary subunit 12221, which is used to calculate the complementary information of each position spectral feature coding vector in the multiple position spectral feature coding vectors relative to the multi-position spectral hub feature coding vector to obtain multiple position node-spectral hub feature complementary information embedded coding vectors; a significance identification secondary subunit 12222, which is used to input each position node-spectral hub feature complementary information embedded coding vector in the multiple position node-spectral hub feature complementary information embedded coding vectors into the complementary information significance identification module based on the attention mechanism to obtain the attention weights of the multiple position node spectral feature complementary information.

[0044] In a specific example of the present application, the complementary information extraction secondary subunit 12221 is used to: perform point convolution encoding based on the Sigmoid activation function on the position spectrum feature coding vector and the multi-position spectrum hub feature coding vector to obtain a standardized position spectrum feature coding vector and a standardized multi-position spectrum hub feature coding vector; calculate the differential features between the standardized position spectrum feature coding vector and the standardized multi-position spectrum hub feature coding vector to obtain a position node-spectrum hub feature differential feature coding vector; based on the position node-spectrum hub feature differential feature coding vector, perform complementary feature enhancement modulation aggregation encoding on the position spectrum feature coding vector and the multi-position spectrum hub feature coding vector to obtain the position node-spectrum hub feature complementary information embedding coding vector. That is, the spectral feature representations of each position are integrated and associated with the multi-position spectrum hub feature coding vector as the core to achieve a comprehensive analysis and accurate description of the overall spectral features of the wooden furniture material to be tested, which can be expressed as follows:

[0045]

[0046] in, represents the Sigmoid activation function, Indicates the The corresponding standardized position spectral feature encoding vector, express The corresponding standardized multi-position spectral hub feature encoding vector, and represents different weight matrices, represents point convolutional coding, represents the position node-spectral hub feature differential feature encoding vector, represents the difference operation, Indicates the The corresponding position node-spectral hub feature complementary information is embedded into the encoding vector.

[0047] That is, by calculating the complementary information of the spectral feature coding vectors at each position relative to the multi-position spectral hub feature coding vectors, the characteristic differences and unique contributions of the spectral feature coding vectors at each position relative to the global core information are captured, and multiple position node-spectral hub feature complementary information embedding coding vectors are generated, so as to restore the local characteristic components of the spectral feature coding vectors at each position that are not fully considered and absorbed in the hub feature extraction process. In this way, not only the complementarity and differences between the spectral feature coding vectors at each position can be revealed, but also the unique information of each position can be retained and reflected, thereby providing support for a more comprehensive understanding of material properties.

[0048] In a specific example of the present application, the significant identification secondary subunit 12222 is expressed by the formula:

[0049]

[0050]

[0051] in, express The corresponding significant identification factor is and They represent the weight parameter matrix and bias term of the complementary information saliency identification module, represents the hub feature relevance score transformation vector of the complementary information saliency identification module, represents the exponential function operation with e as the base, express The corresponding position node spectral feature complementary information attention weight.

[0052] That is, by introducing the attention mechanism, the importance of embedding the encoding vector of the complementary information of the feature of each node-spectral hub at each position is evaluated. In this way, the weight of the embedding encoding vector of the complementary information of the feature of each node-spectral hub at each position in the final global spectral feature description can be dynamically adjusted to ensure that more representative spectral features are strengthened in the final global spectral feature description, while noise or unimportant feature information is weakened or ignored. In this way, the global spectral feature description can be flexibly optimized according to the contribution of different positions, highlighting key information while filtering out irrelevant data, thereby improving the accuracy and reliability of the overall analysis.

[0053] More specifically, in a specific example of the present application, the feature compensation aggregation subunit 1223 is used to: based on the attention weights of the multiple position node spectral feature complementary information, perform attention modulation on the multiple position node-spectral hub feature complementary information embedded coding vectors to obtain multiple significant position node-spectral hub feature complementary information embedded coding vectors,

[0054]

[0055] in, , , and Respectively , , and The corresponding position node-spectral hub feature complementary information is embedded in the encoding vector, , , and Respectively , , and The corresponding position node spectral feature complementary information attention weight, Represents a set of embedding coding vectors of complementary information of the plurality of significant position nodes-spectral hub features.

[0056] That is, based on the calculated attention weights of the complementary information of the spectral features of multiple position nodes, the attention degree of the embedded coding vectors of the complementary information of the multiple position nodes-spectral hub features is modulated, and by enhancing or suppressing the features, it is ensured that the modulated complementary information embedded coding vectors of the multiple significant position nodes-spectral hub features can more accurately reflect the key role and contribution of the spectral features of each position in the overall spectral feature description, while also retaining the spatial distribution characteristics of the original features. In this way, the uniqueness and importance of each position can be better captured, thereby improving the accuracy and reliability of the overall analysis.

[0057] In a specific example of the present application, the feature compensation aggregation subunit 1223 is further used to: fuse the multi-position spectral hub feature encoding vector and the plurality of significant position node-spectral hub feature complementary information embedding encoding vectors to obtain the spectral feature encoding vector of the wooden furniture material to be tested, which is expressed as:

[0058]

[0059] in, represents a cascade function, A spectral feature encoding vector representing the wooden furniture material to be tested.

[0060] That is, the complementary information of the multiple significant position nodes-spectral hub features is embedded in the coding vector and cascaded with the multi-position spectral hub feature coding vector to comprehensively consider the global core information of the spectral features of the wooden furniture material to be tested and the local detail information and unique contribution of the spectral features of each position node, so as to improve the accuracy and comprehensiveness of the feature expression, ensure that the global spectral feature description of the wooden furniture material to be tested can more accurately reflect the internal composition and structural characteristics of the wooden furniture material to be tested, and thus provide more detailed and reliable information support. Through this fusion strategy, the unique details of each position can be fully captured while retaining the global features, so that the overall analysis has both the accuracy of the macro perspective and the fineness of the micro level.

[0061] In the above-mentioned rapid detection system for wooden furniture materials, the reference feature extraction module 130 is used to extract a set of spectral feature coding vectors of known wood types from the wooden material spectral feature database. That is, by extracting the spectral features of known wood types as reference standards, the spectral features of the wooden furniture material to be tested are compared and analyzed with the spectral features of the known wood types, so as to quickly and accurately determine the wood type to which the material to be tested belongs. It should be understood that the spectral feature coding vector of the known wood type is a spectral feature generated by performing spectral scanning and analysis of the material of the known wood type at multiple positions using a portable spectral scanner, and performing image encoding using the same spectral image processing technology as the wooden furniture material to be tested, and is stored in the database in the form of a spectral feature coding vector for subsequent rapid retrieval and comparison.

[0062] In the above-mentioned rapid detection system for wooden furniture materials, the query matching module 140 is used to query and match the spectral feature coding vector of the wooden furniture material to be tested with the set of spectral feature coding vectors of the known wood types to obtain a matching analysis result, and the matching analysis result is the closest wood type. In a specific example of the present application, the query matching module 140 is used to: respectively calculate the cosine similarity between the spectral feature coding vector of the wooden furniture material to be tested and each spectral feature coding vector in the set of spectral feature coding vectors of the known wood types, so as to determine the wood type corresponding to the spectral feature coding vector corresponding to the maximum cosine similarity as the matching analysis result. It should be understood that cosine similarity is a commonly used similarity measurement method, which measures the similarity between two vectors by calculating the cosine value of the angle between the two vectors. The closer the cosine value is to 1, the more similar the two are. In this application, cosine similarity is used to evaluate the similarity between the spectral feature encoding vector of the wooden furniture material to be tested and the spectral feature encoding vector of the known wood type, and the most likely wood type is determined as the matching analysis result according to the ranking of the similarity, thereby providing the user with the most likely wood type suggestion, which not only improves the accuracy and efficiency of wood type identification, but also provides strong technical support for the identification and quality control of wooden furniture materials. In addition, in order to improve the accuracy and reliability of matching, a similarity threshold can also be set. Only when the maximum cosine similarity exceeds the threshold, the wood type represented by it is determined as a matching result. Otherwise, it can be regarded as a matching failure. At this time, the user can be prompted to re-scan the spectrum or use other methods for further identification and analysis.

[0063] Specifically, when the matching process fails to successfully identify the type of wooden furniture material to be tested, the system should provide a clear and instructive feedback mechanism to help users understand and deal with this situation. First, the system needs to quickly analyze the reasons for the matching failure and classify the possible problems. This may be due to poor sample quality, insufficient spectral data, anomalies in the feature encoding process, or the material to be tested itself does not belong to any known wood type in the database. Based on the preliminary diagnosis results, the system will generate a clear and easy-to-understand message to explain to the user the specific reason for the matching failure and recommend the next step.

[0064] For example, if the problem is due to poor spectral image quality (such as excessive noise, insufficient resolution, etc.), the message may state: "The detected spectral image quality is low, which may have caused the match failure. Please ensure that the sample surface is clean and undamaged, and rescan the spectrum in a well-lit environment." In order to help users successfully complete the rescan, the system should provide detailed operating instructions, including how to clean the sample surface to avoid using detergents that may interfere with the spectral signal; remind users to keep the sample dry to prevent moisture from affecting the spectral characteristics; give recommendations on the best working environment conditions, such as a stable temperature and humidity range, appropriate lighting intensity, etc.; emphasize the importance of maintaining consistent environmental parameters to ensure data comparability; introduce in detail the correct use of portable spectral scanners, from power-on preheating to final data transmission; specially mark steps that are prone to errors, such as instrument calibration, position fixing, etc.; guide users to choose appropriate scanning paths and sampling points to ensure coverage of representative areas; encourage data collection from multiple angles and levels to improve the matching success rate.

[0065] In particular, in the technical solution of the present application, the spectral feature coding vector of the wooden furniture material to be tested is obtained by dynamically compensating and aggregating the spectral features at each position. However, considering that the spectral features at each position have noise and each spectral feature coding vector in the set of spectral feature coding vectors of the known wood types is a pure spectral feature, therefore, in the process of calculating the cosine similarity between the spectral feature coding vector of the wooden furniture material to be tested and each spectral feature coding vector in the set of spectral feature coding vectors of the known wood types, the spectral feature coding vector of the wooden furniture material to be tested and the spectral feature coding vector will have an edge enhancement structure when mapped to the cosine similarity domain due to the presence of noise, which affects the calculation accuracy of the cosine similarity.

[0066] Based on this, in a preferred embodiment of the present application, the process of calculating the cosine similarity between the spectral feature coding vector of the wooden furniture material to be tested and each spectral feature coding vector in the set of spectral feature coding vectors of the known wood types includes:

[0067] Cascading the spectral feature coding vector of the wooden furniture material to be tested and the spectral feature coding vector of the known wood type to obtain a spectral feature cascade coding vector of the wooden furniture material to be tested and known;

[0068] A random permutation operator is used to perform dynamic topological reorganization on the cascade coding vector of the spectral feature of the to-be-tested-known wooden furniture material to generate a dynamic reorganized coding tensor of the spectral feature of the to-be-tested-known wooden furniture material, representing for ,in, is a random permutation operator, represents the cascade coding vector of the spectrum characteristics of the known wooden furniture material to be tested, represents matrix multiplication, Represents the dynamic reorganization encoding tensor of the spectral characteristics of the known wooden furniture material to be tested;

[0069] The complete linear field coding tensor is constructed by the outer product tensor generation operation of the cascade coding vector of the spectral characteristics of the to-be-tested and known wooden furniture materials, which is expressed as: ,in, represents the transposed vector of the cascade coding vector of the spectral characteristics of the known wooden furniture material to be tested, represents a complete linear field encoding tensor;

[0070] The dynamic reorganized coding tensor of the spectrum characteristics of the to-be-tested-known wooden furniture material and the complete linear field coding tensor are subjected to field projection to generate a scalar associated distribution heterogeneous coding tensor of the spectrum characteristics of the to-be-tested-known wooden furniture material, which is expressed as: ,in, Represents the heterogeneous encoding tensor of the scalar correlation distribution of the spectral characteristics of the tested and known wooden furniture materials;

[0071] The topological fragmentation analysis is performed on the conjugate tensor of the spectral characteristic scalar associated distribution heterogeneous coding tensor of the to-be-tested-known wooden furniture material and the spectral characteristic cascade coding vector of the to-be-tested-known wooden furniture material to obtain the spectral characteristic heterogeneous topological tensor of the to-be-tested-known wooden furniture material, which is expressed as: ,in, Before retaining The truncated decomposition of singular values, Represents the heterogeneous topological tensor of the spectral characteristics of the known wooden furniture material to be tested;

[0072] Performing field fusion on the spectral feature cascade coding vector of the to-be-tested-known wooden furniture material and the spectral feature heterogeneous topological tensor of the to-be-tested-known wooden furniture material to output an optimized spectral feature cascade coding vector of the to-be-tested-known wooden furniture material;

[0073] Decomposing the optimized spectral feature cascade coding vector of the to-be-tested-known wooden furniture material into an optimized spectral feature coding vector of the to-be-tested wooden furniture material and an optimized spectral feature coding vector of the known wood type according to a cascade mode;

[0074] The cosine similarity between the optimized spectral feature encoding vector of the wooden furniture material to be tested and the optimized spectral feature encoding vector of the known wood type is calculated.

[0075] Here, by dynamically topologically reorganizing the cascaded coding vector of the spectral features of the known wooden furniture material to be tested through a random permutation operator and embedding it into a complete linear field generated by the outer product tensor, a resolvable measurement of the scalar correlation distribution of the feature set in the heterogeneous topological structure can be achieved. Based on this, the heterogeneous topological tensor is constructed using conjugate tensor transformation, and the reduced-dimensional manifold of the feature space is reconstructed within the field in combination with topological piecewise analysis to avoid the compression effect of the edge enhancement architecture on the probability of feature characterization. In this way, the edge architecture smoothness of the spectral feature coding vector of the wooden furniture material to be tested and the spectral feature coding vector when mapped to the cosine similarity domain is improved to improve the calculation accuracy of the cosine similarity between the two.

[0076] In summary, a rapid detection system for wooden furniture materials based on the embodiment of the present application is explained, which uses a portable spectral scanner to collect spectral scanning images of wooden furniture materials to be tested from multiple positions, and introduces spectral image processing technology based on deep learning to extract spectral features of spectral scanning images at multiple positions, perform correlation analysis between spectral features at each position, and perform complementary information enhanced aggregation processing to construct a global spectral feature description of the wooden furniture materials to be tested, and then achieve rapid and accurate detection of wooden furniture materials based on matching analysis between the spectral features of the wooden furniture materials to be tested and the spectral features of known wood types. In this way, the limitations of traditional wood identification methods and the shortcomings of existing spectral detection methods can be effectively overcome, and the accuracy and efficiency of wooden furniture material detection can be improved.

[0077] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0078] In the above embodiments, the description of each embodiment has its own emphasis. For the parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0079] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0080] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0081] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A rapid detection system for wooden furniture materials, characterized in that: include: A multi-position spectral scanning image acquisition module, used to acquire spectral scanning images of multiple positions of the wooden furniture material to be tested collected by a portable spectral scanner; A multi-position spectral feature encoding module, used for calculating the spectral feature encoding vector of the wooden furniture material to be tested based on the spectral scanning images of the multiple positions; A reference feature extraction module is used to extract a set of spectral feature encoding vectors of known wood types from a wood material spectral feature database; A query matching module, used for querying and matching the spectral feature coding vector of the wooden furniture material to be tested with the set of spectral feature coding vectors of the known wood types to obtain a matching analysis result, wherein the matching analysis result is the closest wood type; The multi-position spectral feature encoding module comprises: A spectral feature extraction unit, used for respectively extracting the spectral features of the spectral scanning images at each position to obtain spectral feature encoding vectors of multiple positions; A feature dynamic compensation aggregation unit, used for performing attention-driven feature dynamic compensation aggregation on the plurality of position spectral feature coding vectors to obtain the spectral feature coding vector of the wooden furniture material to be tested; The characteristic dynamic compensation aggregation unit comprises: A hub feature extraction subunit, used for inputting the plurality of position spectral feature encoding vectors into a hub feature extraction network to obtain a multi-position spectral hub feature encoding vector; A saliency scoring subunit, configured to perform saliency scoring on each of the plurality of position spectral feature coding vectors based on complementary information of each position spectral feature coding vector relative to the multi-position spectral hub feature coding vector, so as to obtain attention weights of complementary information of spectral features of a plurality of position nodes; The feature compensation aggregation subunit is used to dynamically compensate and aggregate the spectral feature coding vectors of the multiple position nodes based on the attention weights of the complementary information of the spectral features of the multiple position nodes to obtain the spectral feature coding vector of the wooden furniture material to be tested.

2. The rapid detection system for wooden furniture materials according to claim 1 is characterized in that: The spectral feature extraction unit is used to: The spectral scanning images of the various positions are input into a spectral feature extractor based on the FPT model to obtain spectral feature encoding vectors of the multiple positions.

3. The rapid detection system for wooden furniture materials according to claim 2 is characterized in that: The significance scoring subunit comprises: A complementary information extraction secondary subunit is used to calculate the complementary information of each position spectrum feature coding vector in the multiple position spectrum feature coding vectors relative to the multiple position spectrum hub feature coding vector to obtain multiple position node-spectrum hub feature complementary information embedded coding vectors; The saliency identification secondary sub-unit is used to input each position node-spectral hub feature complementary information embedding coding vector in the multiple position node-spectral hub feature complementary information embedding coding vector into the complementary information saliency identification module based on the attention mechanism to obtain the attention weights of the multiple position node spectral feature complementary information.

4. The rapid detection system for wooden furniture materials according to claim 3 is characterized in that: The complementary information extraction secondary subunit is used for: Performing point convolution encoding based on Sigmoid activation function on the position spectrum feature encoding vector and the multi-position spectrum hub feature encoding vector respectively to obtain a standardized position spectrum feature encoding vector and a standardized multi-position spectrum hub feature encoding vector; Calculating the differential feature between the standardized position spectrum feature coding vector and the standardized multi-position spectrum hub feature coding vector to obtain a position node-spectrum hub feature differential feature coding vector; Based on the position node-spectral hub feature differential feature coding vector, the position spectral feature coding vector and the multi-position spectral hub feature coding vector are subjected to complementary feature enhancement modulation aggregation coding to obtain the position node-spectral hub feature complementary information embedding coding vector.

5. The rapid detection system for wooden furniture materials according to claim 4 is characterized in that: The characteristic compensation aggregation subunit is used for: Based on the attention weights of the complementary information of the spectral features of the multiple position nodes, attention modulation is performed on the embedded coding vectors of the complementary information of the multiple position nodes-spectral hub features to obtain multiple embedded coding vectors of the complementary information of the salient position nodes-spectral hub features; The multi-position spectral hub feature coding vector and the plurality of significant position node-spectral hub feature complementary information embedding coding vectors are fused to obtain the spectral feature coding vector of the wooden furniture material to be tested.

6. The rapid detection system for wooden furniture materials according to claim 5, characterized in that: The query matching module is used to: The cosine similarities between the spectral feature coding vector of the wooden furniture material to be tested and each spectral feature coding vector in the set of spectral feature coding vectors of the known wood types are calculated respectively, so as to determine the wood type corresponding to the spectral feature coding vector corresponding to the maximum cosine similarity as the matching analysis result.

Images