Building material surface defect intelligent detection method and system

Abstract

The invention discloses a building material surface defect intelligent detection method and system, and relates to the technical field of data processing. The method comprises the steps of obtaining K batches and the number of the K batches of a target building material; k initial extracted sample sets are obtained, surface defect detection is carried out, and K first-batch reliability factors are obtained; carrying out volume analysis, and determining K second batch reliability factors; obtaining K conditional sampling sample sets; carrying out surface defect detection to obtain a surface defect detection result set of K conditional samples; carrying out similar aggregation to obtain K clustering detection result clusters; determining a value set of K clustering defect feature sets; and performing cross-batch weighted calculation according to a preset weight to obtain a weighted clustering defect feature set value set, and taking the weighted clustering defect feature set value set as K surface defect detection results. The technical problem that the building material surface defect detection efficiency and accuracy are low in the prior art is solved, and the technical effect of improving the detection efficiency and accuracy is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an intelligent detection method and system for surface defects of building materials. Background Art

[0002] In the production and application of building materials, surface defects are important factors affecting product quality and performance. For example, surface defects such as cracks, pores, and scratches may cause the structural strength of the material to decrease, and the protective performance to decrease, thus affecting its service life and safety. At present, traditional surface defect detection methods mainly rely on manual detection or automated detection methods based on a single algorithm. On the one hand, manual detection is inefficient and has strong subjectivity due to the experience and judgment of the inspectors, making it difficult to ensure the objectivity and consistency of the detection results; on the other hand, automated detection technology based on a single algorithm is not adaptable enough when facing different batches or diversified types of building materials, and it is difficult to balance detection accuracy and generalization capabilities. Summary of the invention

[0003] The present application provides a method and system for intelligent detection of building material surface defects, which solves the technical problem of low efficiency and accuracy in detection of building material surface defects in the prior art.

[0004] In view of the above problems, the present application provides a method and system for intelligent detection of surface defects of building materials.

[0005] In a first aspect of the present application, a method for intelligently detecting surface defects of building materials is provided, the method comprising: Obtain K batches and K batch quantities of target building materials, where K is an integer greater than or equal to 1; extract a preset number of initial samples from the K batches in a random sampling manner to obtain K initial sample sets, and perform surface defect detection on the K initial sample sets, and obtain K first batch reliability factors according to the detection results; perform volume analysis based on the K batch quantities to determine K second batch reliability factors; perform conditional sampling on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets; traverse the K conditional sampling sample sets to perform surface defect detection to obtain K conditional sampling sample surface defect detection result sets The method comprises the following steps: performing similar aggregation on the K conditional sampling sample surface defect detection result sets to obtain K cluster detection result clusters, each cluster detection result cluster includes multiple cluster conditional sampling sample surface defect detection result sets, and each cluster conditional sampling sample surface defect detection result set includes multiple conditional sampling sample surface defect detection results belonging to the same surface defect type; performing defect feature value recognition on the K cluster detection result clusters, and performing feature value centralized analysis on the recognition results to determine K cluster defect feature centralized value sets; performing cross-batch weighted calculation on the K cluster defect feature centralized value sets according to preset weights to obtain weighted cluster defect feature centralized value sets, and using the weighted cluster defect feature centralized value sets as the K surface defect detection results.

[0006] The second aspect of the present application provides an intelligent detection system for surface defects of building materials, the system comprising: a data acquisition module, the data acquisition module is used to obtain K batches and K batch quantities of target building materials, wherein K is an integer greater than or equal to 1; a first sampling module, the first sampling module is used to extract a preset number of initial samples from the K batches in a random sampling manner to obtain K initial extracted sample sets, and perform surface defect detection on the K initial extracted sample sets, and obtain K first batch reliability factors according to the detection results; an analysis module, the analysis module is used to perform volume analysis based on the K batch quantities to determine K second batch reliability factors; a second sampling module, the second sampling module is used to perform conditional sampling on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets; a defect detection module, the defect detection module is used to traverse the K conditional sampling sample sets to perform surface defect detection The invention relates to a method for detecting a surface defect of a plurality of conditional sampling samples by a plurality of conditional sampling samples, wherein the method comprises: detecting a surface defect of the plurality of conditional sampling samples by a plurality of conditional sampling samples; ...

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, obtain K batches and K batch quantities of target building materials, where K is an integer greater than or equal to 1. Then, extract the preset number of initial samples from the K batches in a random sampling manner to obtain K initial sample sets, and perform surface defect detection on the K initial sample sets, and obtain K first batch reliability factors according to the detection results; at the same time, perform volume analysis based on the K batch quantities to determine K second batch reliability factors. Further, perform conditional sampling on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets; and traverse the K conditional sampling sample sets to perform surface defect detection to obtain K conditional sampling sample surface defect detection result sets. Then, perform homogeneous aggregation on the K conditional sampling sample surface defect detection result sets to obtain K cluster detection result clusters, each cluster detection result cluster includes multiple cluster conditional sampling sample surface defect detection result sets, and each cluster conditional sampling sample surface defect detection result set includes multiple conditional sampling sample surface defect detection results belonging to the same surface defect type. Next, the defect feature values ​​of the K cluster detection result clusters are identified, and the feature value concentration analysis is performed on the identification results to determine the K cluster defect feature concentration value sets. Finally, the K cluster defect feature concentration value sets are cross-batch weighted calculations according to preset weights to obtain weighted cluster defect feature concentration value sets, and the weighted cluster defect feature concentration value sets are used as K surface defect detection results. The technical problem of low efficiency and accuracy of building material surface defect detection in the prior art is solved, and the technical effect of improving detection efficiency and accuracy is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A schematic flow chart of a method for intelligent detection of surface defects of building materials provided in an embodiment of the present application.

[0010] Figure 2 A schematic diagram of the structure of an intelligent detection system for surface defects of building materials provided in an embodiment of the present application.

[0011] Explanation of the reference numerals: data acquisition module 11 , first sampling module 12 , analysis module 13 , second sampling module 14 , defect detection module 15 , aggregation module 16 , identification module 17 , calculation module 18 . DETAILED DESCRIPTION

[0012] The present application solves the technical problem of low efficiency and accuracy in detecting surface defects of building materials in the prior art by providing an intelligent method and system for detecting surface defects of building materials.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0015] Embodiment 1, as Figure 1 As shown, the present application provides an intelligent detection method for surface defects of building materials, wherein the method comprises: Obtain K batches and K batch quantities of target building materials, where K is an integer greater than or equal to 1.

[0016] When performing intelligent detection of surface defects of building materials, it is necessary to determine K batches of target building materials to be detected and the number of K batches, where K is an integer greater than or equal to 1.

[0017] A preset number of initial samples are drawn from the K batches in a random sampling manner to obtain K initial sample sets, and surface defect detection is performed on the K initial sample sets to obtain K first batch reliability factors based on the detection results.

[0018] For each batch (there are K batches in total), random sampling is performed according to the preset number of draws, that is, a certain number of building material samples are randomly selected from each batch to form an initial draw sample set. Through the random sampling process, K initial draw sample sets are obtained. For the initial sample set of each batch, surface defect detection equipment is used for detection, and the detection results of each sample (such as defect type, defect severity, etc.) are recorded. According to the results of surface defect detection, the quality of the initial draw sample set of each batch is evaluated, and the proportion of defective samples in the batch, the severity of defects and other indicators can be calculated. Then, based on these indicators, a first batch reliability factor is obtained. The first batch reliability factor reflects the reliability of the surface quality of the building materials in this batch.

[0019] Further, K initial sample sets are obtained, and surface defect detection is performed on the K initial sample sets, and K first batch reliability factors are obtained according to the detection results, including: Perform surface defect detection on the K initial sample sets to obtain K initial sample detection result sets; traverse the K initial sample detection result sets to perform result stability analysis to obtain K stability factors; and take the ratios of the K stability factors to the sum of the K stability factors as the K first batch reliability factors.

[0020] Preferably, each sample in the K initial sample collections is tested for surface defects, and K initial sample collections are obtained based on the test results; the K initial sample collections are traversed, and stability analysis is performed on the test results in each collection. The variance of all sample test results in each initial sample collection is calculated, and the inverse of the variance is used as the stability factor. The larger the variance, the greater the difference in the test results of the samples in the batch, and the lower the stability. Conversely, the smaller the variance, the higher the stability. In order to quantify the reliability of each batch, the stability factor of each batch is divided by the sum of the stability factors of all batches, thereby obtaining K first batch reliability factors.

[0021] A volume analysis is performed based on the K batch quantities to determine K second batch reliability factors.

[0022] K second batch reliability factors are obtained by dividing the number of K batches by the sum of the number of K batches. The second batch reliability factor reflects the importance of each batch. The larger the batch number, the higher the second batch reliability factor.

[0023] Conditional sampling is performed on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets.

[0024] According to the needs of practical applications, weights are assigned to the reliability factors of the first batch and the reliability factors of the second batch. These weights reflect the importance of the two factors to the final sampling decision during the sampling process. According to the assigned weights, the comprehensive reliability factor of each batch is calculated by multiplying the reliability factors of the first batch and the second batch by their weights respectively, and then adding the results. According to the comprehensive reliability factors, the sampling ratio of each batch is determined, and the sampling ratio is used to allocate the number of samples of each batch in the overall sample; according to the preset total sampling number, the actual sampling number of each batch is calculated; according to the calculated sampling number, samples are randomly selected from the material set of each batch to form a conditional sampling sample set.

[0025] Further, conditional sampling is performed on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets, including: The K first batch reliability factors and the K second batch reliability factors are weightedly calculated to obtain K conditional sampling probability coefficients; the areas of the K roulette wheel partitions in the conditional sampling roulette wheel are determined based on the sizes of the K conditional sampling probability coefficients, and the conditional sampling roulette wheel is turned multiple times according to a preset number of draws to obtain the K conditional sampling sample sets; wherein each roulette wheel partition corresponds to a batch.

[0026] Preferably, the weights of the first batch reliability factor and the second batch reliability factor are preset, the K first batch reliability factors and the second batch reliability factors are weighted, and the conditional sampling probability coefficient of each batch is calculated; according to the conditional sampling probabilities of the K batches, a conditional sampling roulette is designed, the roulette is divided into K areas, each area corresponds to a batch, and the area of ​​the area is proportional to the conditional sampling probability of the batch; according to actual needs, the total number of samples to be drawn is set; according to the set number of draws, the conditional sampling roulette is dialed multiple times, and after each dialing, the selected batch is determined according to the area where the pointer is located when the roulette stops; a sample is randomly drawn from the selected batch and added to the corresponding conditional sampling sample set; this process is repeated until the set number of draws is reached; after multiple dialing and drawing, the obtained samples are classified according to batches to form K conditional sampling sample sets.

[0027] Traversing the K conditional sampling sample sets to perform surface defect detection, and obtaining a surface defect detection result set of the K conditional sampling samples; Traverse the K conditional sampling sample sets and perform surface defect detection on the samples in each set one by one. Specifically, use the detection equipment to identify the surface defects of each sample, and record the detection results of each sample, including the defect type (such as cracks, pores or scratches, etc.) and the degree of the defect (such as slight, severe, etc.); for each batch of conditional sampling samples, all the test results are summarized to form the surface defect detection result set of the conditional sampling samples of that batch; by testing the samples of K batches in turn, finally obtain the surface defect detection result set of K conditional sampling samples.

[0028] The K conditional sampling sample surface defect detection result sets are aggregated in the same category to obtain K cluster detection result clusters, each cluster detection result cluster includes multiple cluster conditional sampling sample surface defect detection result sets, and each cluster conditional sampling sample surface defect detection result set includes multiple conditional sampling sample surface defect detection results belonging to the same surface defect type.

[0029] The samples are classified according to the defect type (such as cracks, pores, scratches, etc.) of each sample in the test results, and samples of the same defect type are classified into one category. For each batch of test result sets, all sample test results with the same defect type are extracted to form multiple cluster conditional sampling sample surface defect test result sets, each set corresponding to a specific defect type. These similar test results are further aggregated to generate clusters of clustered test results for each batch, and each cluster contains a set of clustered results of all different defect types detected in the batch. By processing the test results of K batches in sequence, K clusters of clustered test results are finally obtained.

[0030] Furthermore, the K conditional sampling surface defect detection result sets are aggregated into the same category to obtain K cluster detection result clusters, including: Extract K first conditional sampling sample surface defect detection result sets from the K conditional sampling sample surface defect detection result sets respectively, and store them in K first aggregation nodes; perform binary classification on the K conditional sampling sample surface defect detection result sets according to a preset similarity threshold at the K first aggregation nodes, add the detection results whose similarity with the K first conditional sampling sample surface defect detection result sets meets the preset similarity threshold into the K first clustering conditional sampling sample surface defect detection result sets, and add the remaining detection results into the K first to-be-divided sets; extract K from the K first to-be-divided sets again The second conditional sampling sample surface defect detection result set is stored in K second aggregation nodes; the K first sets to be divided are binary-classified at the K second aggregation nodes according to a preset similarity threshold, and the detection results whose similarity with the K second conditional sampling sample surface defect detection result sets meets the preset similarity threshold are added to the K second clustering conditional sampling sample surface defect detection result sets, and the remaining detection results are added to the K second sets to be divided; after multiple aggregation analyses, until the detection results in the K conditional sampling sample surface defect detection result sets are all divided, the K clustering detection result clusters are obtained.

[0031] Specifically, a representative test result is extracted from each batch of conditional sampling sample surface defect test result sets to form K first conditional sampling sample surface defect test result sets, and these sets are stored in K first aggregation nodes respectively; according to the preset similarity threshold, each batch of test result sets are classified into two categories: the test results whose similarity with the test result set in the first aggregation node meets the threshold are classified into K first cluster conditional sampling sample surface defect test result sets; the remaining test results are classified into K first sets to be divided. For the K first sets to be divided, a new representative test result is extracted again to form K second conditional sampling sample surface defect test result sets, and they are stored in K second aggregation nodes; in a similar manner, the sets to be divided are classified into two categories: the test results that meet the similarity threshold are classified into K second cluster conditional sampling sample surface defect test result sets; the remaining test results are classified into K second sets to be divided. According to the above steps, the remaining test results in the set to be divided are extracted, stored in new aggregation nodes, classified and divided in turn; as the number of iterations increases, the number of undivided test results gradually decreases, and the set of clustering condition samples gradually expands; multiple rounds of aggregation analysis are continued until all test results are classified into the corresponding clustering condition sampling sample sets, and the set to be divided is empty; finally, K clustering test result clusters are formed, each cluster contains the complete set of all test results in the batch after aggregation according to defect type.

[0032] Defect feature value identification is performed on the K cluster detection result clusters, and feature value centralized analysis is performed on the identification results to determine K cluster defect feature centralized value sets.

[0033] Extract features related to surface defects from each cluster of cluster detection results, such as defect size, shape, location, color, etc.; quantify the extracted features using machine learning algorithms or statistical methods to obtain specific values ​​for each feature, which will be used as defect feature values. For the feature value set in each cluster detection result cluster, calculate the central statistical indicators of the feature values, such as mean, variance, extreme value, etc., to comprehensively reflect the overall distribution of sample features in the cluster; after completing the centralized analysis of all feature values ​​in each cluster detection result cluster, form K cluster defect feature centralized value sets, each set includes the centralized value statistical results of all features in the cluster.

[0034] Further, including: Extract K first cluster conditional sampling sample surface defect detection result sets from the K cluster detection result clusters; use a semantic analyzer to identify defect feature values ​​on the K first cluster conditional sampling sample surface defect detection result sets to obtain K first cluster defect feature value sets; traverse the K first cluster defect feature value sets to perform feature value centralized analysis to determine the K first cluster defect feature centralized values; perform defect feature value centralized analysis on the K cluster detection result clusters to obtain the K cluster defect feature centralized value sets.

[0035] Specifically, a certain number of representative conditional sampling sample surface defect detection result sets are selected from each cluster detection result cluster as the first cluster conditional sampling sample surface defect detection result set; the extracted first cluster conditional sampling sample surface defect detection result set is preprocessed, such as data cleaning and formatting, to ensure the accuracy and consistency of the data; a semantic analyzer is used to identify defect feature values ​​for each first cluster conditional sampling sample surface defect detection result set, and key feature values ​​related to surface defects are extracted. For example, the semantic analyzer can automatically identify the geometric size of the defect based on specific rules or models, Position distribution, morphological characteristics, etc., to form K first cluster defect feature value sets; traverse each first cluster defect feature value set, perform statistical analysis on all eigenvalues ​​in the set, calculate the concentration index of the eigenvalues, such as mean, variance or median, etc., so as to determine the K first cluster defect feature concentration values; based on the results of the first step of feature value concentration analysis, further perform a more extensive defect feature value analysis on the K cluster detection result clusters, covering the eigenvalues ​​of all samples in each cluster, calculate the feature concentration value set of all samples, including the comprehensive statistical indicators of all features, and finally form K cluster defect feature concentration value sets.

[0036] Further, including: Calculate the mean of the K first cluster defect feature value sets to obtain the K first cluster defect feature means; use the K first cluster defect feature means as indexes, iterate in the K first cluster defect feature value sets according to a preset iteration step length, and obtain K first iteration cluster defect feature values; determine whether the concentration of the K first iteration cluster defect feature values ​​is greater than or equal to the concentration of the K first cluster defect feature means, and if so, use the K first iteration cluster defect feature values ​​as indexes, iterate in the K first cluster defect feature value sets according to a preset iteration step length until a preset number of iterations is met, obtain K first target cluster defect feature values, and use the K first target cluster defect feature values ​​as the K first cluster defect feature concentration values.

[0037] The defect feature value set of each cluster is averaged, and the mean of all feature values ​​in the K first cluster defect feature value sets is calculated to obtain the K first cluster defect feature means; the K first cluster defect feature means are used as indexes, and iterative updates are performed in the defect feature value set of each cluster according to the preset iteration step; through each round of iteration, the feature value set will gradually approach the true features of the cluster, thereby obtaining K first iteration cluster defect feature values; compare the K first iteration cluster defect feature values ​​with their corresponding means to determine whether their concentration is greater than or equal to the concentration of the initial mean; if If not, the iteration is stopped, and the means of the K first cluster defect features are taken as the concentrated value sets of the K first cluster defect features; if the concentrated value of the K first iteration cluster defect feature values ​​meets the conditions, these iteration values ​​are used as new indexes, and iteration is performed again in the feature value set according to the same iteration step; when the iteration process reaches the preset number of times and the concentrated value of the feature value set meets the stability condition, K first target cluster defect feature values ​​are finally obtained, and these values ​​are taken as the final concentrated value of the feature for each cluster; the K first target cluster defect feature values ​​are taken as the final concentrated value of the first cluster defect features.

[0038] Furthermore, the number of first cluster defect feature values ​​whose distance to the K first cluster defect feature means is the preset iteration step in the K first cluster defect feature value sets is counted to obtain the concentration of the K first iteration cluster defect feature values.

[0039] For each sample in the first cluster defect feature value set, calculate the distance from the sample to the corresponding cluster feature mean, which can be measured using Euclidean distance or Manhattan distance to indicate the degree of difference between each feature value and its corresponding mean. After calculating the distance from each sample to the mean, count the number of samples whose distance is less than or equal to the preset iteration step. This number indicates how many samples in the feature value set change within the preset step range in one iteration. The number of samples that meet the conditions obtained by counting is used as the concentration of the cluster. The concentration reflects the stability and consistency of the samples in the current iteration. The larger the concentration, the more concentrated the feature values ​​in the current cluster and the more stable the clustering results. The concentration can usually be expressed as a ratio or absolute value of the number.

[0040] The K clustered defect feature concentrated value sets are weighted calculated across batches according to preset weights to obtain a weighted clustered defect feature concentrated value set, and the weighted clustered defect feature concentrated value set is used as the K surface defect detection results.

[0041] The K cluster defect feature concentration value sets are weighted according to the preset weights. First, a weight is set for each batch. These weights can be determined according to the importance of the batch or other relevant factors. Then, the defect feature concentration values ​​of each batch are weighted according to these weights. The specific method is to multiply the feature concentration value of each batch by its corresponding weight, and sum them to obtain the weighted cluster defect feature concentration value set. Finally, the obtained weighted cluster defect feature concentration value set is used as the K surface defect detection results. Through this weighted calculation, the defect characteristics of different batches can be combined and adjusted according to the importance of each batch, so as to more accurately reflect the defect situation and quality level of the overall material.

[0042] In summary, the embodiments of the present application have at least the following technical effects: First, obtain K batches and K batch quantities of target building materials, where K is an integer greater than or equal to 1. Then, extract the preset number of initial samples from the K batches in a random sampling manner to obtain K initial sample sets, and perform surface defect detection on the K initial sample sets, and obtain K first batch reliability factors according to the detection results; at the same time, perform volume analysis based on the K batch quantities to determine K second batch reliability factors. Further, perform conditional sampling on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets; and traverse the K conditional sampling sample sets to perform surface defect detection to obtain K conditional sampling sample surface defect detection result sets. Then, perform homogeneous aggregation on the K conditional sampling sample surface defect detection result sets to obtain K cluster detection result clusters, each cluster detection result cluster includes multiple cluster conditional sampling sample surface defect detection result sets, and each cluster conditional sampling sample surface defect detection result set includes multiple conditional sampling sample surface defect detection results belonging to the same surface defect type. Next, the defect feature values ​​of the K cluster detection result clusters are identified, and the feature value concentration analysis is performed on the identification results to determine the K cluster defect feature concentration value sets. Finally, the K cluster defect feature concentration value sets are cross-batch weighted calculations according to preset weights to obtain weighted cluster defect feature concentration value sets, and the weighted cluster defect feature concentration value sets are used as K surface defect detection results. The technical problem of low efficiency and accuracy of building material surface defect detection in the prior art is solved, and the technical effect of improving detection efficiency and accuracy is achieved.

[0043] Embodiment 2, based on the same inventive concept as the intelligent detection method for surface defects of building materials in the above embodiment, Figure 2 As shown, the present application provides an intelligent detection system for surface defects of building materials, wherein the system comprises: A data acquisition module 11, the data acquisition module 11 is used to obtain K batches and K batch quantities of target building materials, wherein K is an integer greater than or equal to 1; a first sampling module 12, the first sampling module 12 is used to extract a preset number of initial samples from the K batches in a random sampling manner, obtain K initial extracted sample sets, and perform surface defect detection on the K initial extracted sample sets, and obtain K first batch reliability factors according to the detection results; an analysis module 13, the analysis module 13 is used to perform volume analysis based on the K batch quantities to determine K second batch reliability factors; a second sampling module 14, the second sampling module 14 is used to perform conditional sampling on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets; a defect detection module 15, the defect detection module 15 is used to traverse the K conditional sampling sample sets to perform surface defect detection on the K initial sample sets Surface defect detection, obtaining K conditional sampling sample surface defect detection result sets; aggregation module 16, the aggregation module 16 is used to perform similar aggregation on the K conditional sampling sample surface defect detection result sets to obtain K cluster detection result clusters, each cluster detection result cluster includes multiple cluster conditional sampling sample surface defect detection result sets, and each cluster conditional sampling sample surface defect detection result set includes multiple conditional sampling sample surface defect detection results belonging to the same surface defect type; identification module 17, the identification module 17 is used to identify defect feature values ​​of the K cluster detection result clusters, and perform feature value centralized analysis on the identification results to determine K cluster defect feature centralized value sets; calculation module 18, the calculation module 18 is used to perform cross-batch weighted calculation on the K cluster defect feature centralized value sets according to preset weights to obtain weighted cluster defect feature centralized value sets, and use the weighted cluster defect feature centralized value sets as K surface defect detection results.

[0044] Furthermore, the second sampling module 14 is used to perform the following method: The K first batch reliability factors and the K second batch reliability factors are weightedly calculated to obtain K conditional sampling probability coefficients; the areas of the K roulette wheel partitions in the conditional sampling roulette wheel are determined based on the sizes of the K conditional sampling probability coefficients, and the conditional sampling roulette wheel is turned multiple times according to a preset number of draws to obtain the K conditional sampling sample sets; wherein each roulette wheel partition corresponds to a batch.

[0045] Furthermore, the identification module 17 is used to execute the following method: Extract K first cluster conditional sampling sample surface defect detection result sets from the K cluster detection result clusters; use a semantic analyzer to identify defect feature values ​​on the K first cluster conditional sampling sample surface defect detection result sets to obtain K first cluster defect feature value sets; traverse the K first cluster defect feature value sets to perform feature value centralized analysis to determine the K first cluster defect feature centralized values; perform defect feature value centralized analysis on the K cluster detection result clusters to obtain the K cluster defect feature centralized value sets.

[0046] Furthermore, the identification module 17 is used to execute the following method: Calculate the mean of the K first cluster defect feature value sets to obtain the K first cluster defect feature means; use the K first cluster defect feature means as indexes, iterate in the K first cluster defect feature value sets according to a preset iteration step length, and obtain K first iteration cluster defect feature values; determine whether the concentration of the K first iteration cluster defect feature values ​​is greater than or equal to the concentration of the K first cluster defect feature means, and if so, use the K first iteration cluster defect feature values ​​as indexes, iterate in the K first cluster defect feature value sets according to a preset iteration step length until a preset number of iterations is met, obtain K first target cluster defect feature values, and use the K first target cluster defect feature values ​​as the K first cluster defect feature concentration values.

[0047] Furthermore, the identification module 17 is used to execute the following method: The number of first cluster defect feature values ​​whose distance from the K first cluster defect feature means is the preset iteration step in the K first cluster defect feature value sets is counted to obtain the concentration of the K first iteration cluster defect feature values.

[0048] Furthermore, the aggregation module 16 is used to perform the following method: Extract K first conditional sampling sample surface defect detection result sets from the K conditional sampling sample surface defect detection result sets respectively, and store them in K first aggregation nodes; perform binary classification on the K conditional sampling sample surface defect detection result sets according to a preset similarity threshold at the K first aggregation nodes, add the detection results whose similarity with the K first conditional sampling sample surface defect detection result sets meets the preset similarity threshold into the K first clustering conditional sampling sample surface defect detection result sets, and add the remaining detection results into the K first to-be-divided sets; extract K from the K first to-be-divided sets again The second conditional sampling sample surface defect detection result set is stored in K second aggregation nodes; the K first sets to be divided are binary-classified at the K second aggregation nodes according to a preset similarity threshold, and the detection results whose similarity with the K second conditional sampling sample surface defect detection result sets meets the preset similarity threshold are added to the K second clustering conditional sampling sample surface defect detection result sets, and the remaining detection results are added to the K second sets to be divided; after multiple aggregation analyses, until the detection results in the K conditional sampling sample surface defect detection result sets are all divided, the K clustering detection result clusters are obtained.

[0049] Furthermore, the first sampling module 12 is used to perform the following method: Perform surface defect detection on the K initial sample sets to obtain K initial sample detection result sets; traverse the K initial sample detection result sets to perform result stability analysis to obtain K stability factors; and take the ratios of the K stability factors to the sum of the K stability factors as the K first batch reliability factors.

[0050] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0051] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0052] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. An intelligent detection method for surface defects of building materials, characterized in that: The method comprises: Obtaining K batches and K batch quantities of target building materials, where K is an integer greater than or equal to 1; Extracting a preset number of initial samples from the K batches in a random sampling manner to obtain K initial sample sets, and performing surface defect detection on the K initial sample sets to obtain K first batch reliability factors according to the detection results; Performing volume analysis based on the K batch numbers to determine K second batch reliability factors; Perform conditional sampling on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets; Traversing the K conditional sampling sample sets to perform surface defect detection, and obtaining a surface defect detection result set of the K conditional sampling samples; Performing similar aggregation on the K conditional sampling sample surface defect detection result sets to obtain K cluster detection result clusters, each cluster detection result cluster includes multiple cluster conditional sampling sample surface defect detection result sets, and each cluster conditional sampling sample surface defect detection result set includes multiple conditional sampling sample surface defect detection results belonging to the same surface defect type; Perform defect feature value identification on the K cluster detection result clusters, and perform feature value centralized analysis on the identification results to determine a set of K cluster defect feature centralized values; The K clustered defect feature concentrated value sets are weighted calculated across batches according to preset weights to obtain a weighted clustered defect feature concentrated value set, and the weighted clustered defect feature concentrated value set is used as the K surface defect detection results.

2. The intelligent detection method for surface defects of building materials according to claim 1, characterized in that: Conditionally sampling the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets, including: Performing weighted calculation on the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling probability coefficients; The areas of K roulette wheel partitions in the conditional sampling roulette are determined based on the sizes of the K conditional sampling probability coefficients, and the conditional sampling roulette wheel is turned multiple times according to a preset number of draws to obtain the K conditional sampling sample sets; wherein each roulette wheel partition corresponds to a batch.

3. The intelligent detection method for surface defects of building materials according to claim 1, characterized in that: include: Extracting K first cluster conditional sampling sample surface defect detection result sets from the K cluster detection result clusters; Using a semantic analyzer to identify defect feature values ​​on the K first cluster conditional sampling sample surface defect detection result sets, to obtain K first cluster defect feature value sets; Traversing the K first cluster defect feature value sets to perform feature value concentration analysis, and determining K first cluster defect feature concentration values; Perform defect feature value concentration analysis on the K cluster detection result clusters to obtain the K cluster defect feature concentration value sets.

4. The intelligent detection method for surface defects of building materials according to claim 3, characterized in that: include: Calculating the mean of the K first cluster defect feature value sets to obtain the K first cluster defect feature means; Taking the mean values ​​of the K first clustering defect features as indexes, iterating in the K first clustering defect feature value sets according to a preset iteration step length to obtain K first iteration clustering defect feature values; Determine whether the concentration of the K first iterative clustering defect feature values ​​is greater than or equal to the concentration of the K first clustering defect feature mean values. If so, use the K first iterative clustering defect feature values ​​as indexes, and iterate in the K first clustering defect feature value sets according to a preset iteration step until a preset number of iterations is met, so as to obtain K first target clustering defect feature values, and use the K first target clustering defect feature values ​​as the concentration values ​​of the K first clustering defect features.

5. The intelligent detection method for surface defects of building materials according to claim 4, characterized in that: The number of first cluster defect feature values ​​whose distance from the K first cluster defect feature means is the preset iteration step in the K first cluster defect feature value sets is counted to obtain the concentration of the K first iteration cluster defect feature values.

6. The intelligent detection method for surface defects of building materials according to claim 1, characterized in that: The K conditional sampling sample surface defect detection result sets are aggregated into the same category to obtain K cluster detection result clusters, including: Extracting K first conditional sampling sample surface defect detection result sets from the K conditional sampling sample surface defect detection result sets respectively, and storing them in K first aggregation nodes; Binary classification is performed on the K conditional sampling sample surface defect detection result sets according to a preset similarity threshold at the K first aggregation nodes, and detection results whose similarity with the K first conditional sampling sample surface defect detection result sets meets the preset similarity threshold are added into the K first clustering conditional sampling sample surface defect detection result sets, and the remaining detection results are added into the K first to-be-divided sets; Extracting K second conditional sampling sample surface defect detection result sets from the K first to-be-divided sets again, and storing them in K second aggregation nodes; Binary classification is performed on the K first to-be-divided sets at the K second aggregation nodes according to a preset similarity threshold, and the detection results whose similarity with the K second conditional sampling sample surface defect detection result sets meets the preset similarity threshold are added into the K second clustering conditional sampling sample surface defect detection result sets, and the remaining detection results are added into the K second to-be-divided sets; After multiple aggregation analyses, until all the detection results in the K conditional sampling sample surface defect detection result sets are divided, the K cluster detection result clusters are obtained.

7. The intelligent detection method for surface defects of building materials according to claim 1, characterized in that: Obtain K initial sample sets, perform surface defect detection on the K initial sample sets, and obtain K first batch reliability factors according to the detection results, including: Performing surface defect detection on the K initial sample sets to obtain K initial sample detection result sets; Traversing the K initial sample test result sets to perform result stability analysis and obtain K stability factors; The ratios of the K stability factors to the sum of the K stability factors are respectively used as the K first batch reliability factors.

8. Intelligent detection system for building material surface defects, characterized in that: For implementing the intelligent detection method for surface defects of building materials according to any one of claims 1 to 7, the system comprises: A data acquisition module, the data acquisition module is used to obtain K batches and K batch quantities of target building materials, wherein K is an integer greater than or equal to 1; A first sampling module, the first sampling module is used to extract a preset number of initial samples from the K batches in a random sampling manner to obtain K initial sample sets, and perform surface defect detection on the K initial sample sets to obtain K first batch reliability factors according to the detection results; An analysis module, the analysis module is used to perform volume analysis based on the K batch quantities to determine K second batch reliability factors; A second sampling module, the second sampling module is used to perform conditional sampling on the K batches according to the K first batch reliability factors and the K second batch reliability factors to obtain K conditional sampling sample sets; A defect detection module, the defect detection module is used to traverse the K conditional sampling sample sets to perform surface defect detection and obtain a surface defect detection result set of the K conditional sampling samples; An aggregation module, wherein the aggregation module is used to perform similar aggregation on the K conditional sampling sample surface defect detection result sets to obtain K cluster detection result clusters, each cluster detection result cluster includes multiple cluster conditional sampling sample surface defect detection result sets, and each cluster conditional sampling sample surface defect detection result set includes multiple conditional sampling sample surface defect detection results belonging to the same surface defect type; An identification module, the identification module is used to identify defect feature values ​​for the K cluster detection result clusters, and perform feature value centralized analysis on the identification results to determine a set of K cluster defect feature centralized values; A calculation module is used to perform cross-batch weighted calculation on the K clustered defect feature concentrated value sets according to preset weights to obtain a weighted clustered defect feature concentrated value set, and use the weighted clustered defect feature concentrated value set as the K surface defect detection results.

Images