An artificial intelligence-based quality classification method for building panels

Abstract

The present invention relates to the field of data processing technology, and in particular to a method for classifying building board quality based on artificial intelligence, comprising: obtaining feature data corresponding to each feature of N groups of building boards; calculating the standard deviation corresponding to each feature according to the feature data corresponding to each feature in the i-th group of building boards in the N groups of building boards; for the i-th decision tree corresponding to the i-th group of building boards, determining the classification condition of the i-th decision tree according to the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and obtaining the classification result of the i-th decision tree for the i-th group of building boards; analyzing the classification results of the N decision trees for the N groups of building boards, and obtaining the quality classification result of the building boards; wherein N and i are positive integers. The present invention effectively improves the efficiency and accuracy of quality classification of building boards.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a building board quality classification method based on artificial intelligence. Background Art

[0002] During the production process of building boards, various factors may affect the production of building boards, resulting in different defects. Different degrees of defect severity have different effects on the life and safety of building boards. Therefore, quality classification of building boards can improve the efficiency of judging the qualification of building boards and ensure the safety of building boards during use.

[0003] At present, the quality classification method of building boards mainly relies on the experience of quality inspectors to judge, which is time-consuming and laborious to implement, and is highly subjective, and the classification results may be inaccurate. The classification method using defect detection methods can usually only detect the severity of a certain defect, and cannot be comprehensively classified according to multiple quality characteristics. Therefore, when it is necessary to classify the quality of a large number of building boards, the efficiency of using the above method to classify building boards is low, and the classification accuracy is poor. Summary of the invention

[0004] In order to solve the technical problems of low efficiency and poor accuracy of building board classification, the purpose of the present invention is to provide a building board quality classification method based on artificial intelligence. The technical solution adopted is as follows:

[0005] Obtain feature data corresponding to each feature of N groups of building panels;

[0006] Calculate the standard deviation corresponding to each feature according to the feature data corresponding to each feature in the i-th group of building boards in the N groups of building boards;

[0007] For the i-th decision tree corresponding to the i-th group of building boards, determine the classification condition of the i-th decision tree according to the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and obtain the classification result of the i-th decision tree for the i-th group of building boards;

[0008] The classification results of N groups of building boards by N decision trees are analyzed to obtain the quality classification results of the building boards; wherein N and i are positive integers.

[0009] Preferably, for the i-th decision tree corresponding to the i-th group of building boards, the classification condition of the i-th decision tree is determined according to the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and the classification result of the i-th decision tree for the i-th group of building boards is obtained, including:

[0010] When the feature with the largest standard deviation is determined for the jth time, the jth classification condition of the i-th decision tree is determined based on the feature with the largest standard deviation;

[0011] When the feature data corresponding to the feature with the largest standard deviation all meet the jth classification condition, the i-th decision tree outputs j classification results for the i-th group of building boards;

[0012] When the feature data corresponding to the feature with the largest standard deviation does not meet the j-th classification condition, the feature with the largest standard deviation is reselected from the features for which the classification condition has not been determined, and the j+1-th classification condition of the ith decision tree is determined based on the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the determined classification condition.

[0013] Preferably, the classification condition of the i-th decision tree is determined according to the feature with the largest standard deviation, including:

[0014] According to the degree of discreteness of the feature data corresponding to the feature with the largest standard deviation, the classification threshold corresponding to the feature with the largest standard deviation is determined; and the classification condition of the i-th decision tree is determined to be greater than the classification threshold.

[0015] Preferably, determining the classification threshold corresponding to the feature with the largest standard deviation according to the degree of dispersion of the feature data corresponding to the feature with the largest standard deviation includes:

[0016] For each feature data in the feature data corresponding to the feature with the largest standard deviation, the distance between each feature data and its adjacent feature data is calculated to obtain the classification threshold corresponding to the feature with the largest standard deviation.

[0017] Preferably, obtaining feature data corresponding to each feature of N groups of building boards includes:

[0018] Obtaining the detection values ​​of each feature of N groups of building boards;

[0019] The difference between the preset expected value and the detected value corresponding to each feature is determined as the feature data corresponding to each feature.

[0020] Preferably, the method further comprises:

[0021] When the feature data corresponding to the feature with the largest standard deviation does not meet the j-th classification condition, if the maximum standard deviation is reselected from the features whose classification conditions have not been determined and is not greater than the preset standard deviation threshold, the building boards in the i-th group of building boards that do not meet the j-th classification condition will be determined as the j+1-th classification result.

[0022] Preferably, the classification results of N groups of building boards by N decision trees are analyzed to obtain the quality classification results of the building boards, including:

[0023] Determine the weight coefficient of each classification result in the N decision trees according to the occurrence of the classification conditions used by the classification results at the same position in the N decision trees;

[0024] According to each classification result in N decision trees and the corresponding weight coefficients, the quality classification result of the building board is obtained.

[0025] Preferably, according to the occurrence of the classification conditions used by the classification results at the same position in the N decision trees, the weight coefficients of the classification results in the N decision trees are determined, including:

[0026] Determine the first classification credibility of the m-th classification result according to the number of classification conditions used by the m-th classification result on each decision tree in the N decision trees;

[0027] According to the classification conditions used by the m-th classification result on the s-th decision tree, the number of occurrences of the same position in the N decision trees, the second classification credibility of the m-th classification result on the s-th decision tree is determined;

[0028] The first classification credibility and the second classification credibility are calculated to obtain the weight coefficient of the mth classification result on the sth decision tree.

[0029] Preferably, determining the second classification credibility of the mth classification result on the sth decision tree according to the classification condition used by the mth classification result on the sth decision tree and the number of occurrences of the same position in the N decision trees includes:

[0030] The ratio of the number of occurrences of the classification condition used for the mth classification result on the sth decision tree at the same position in the N decision trees to the total number N of all decision trees is determined as the second classification credibility of the mth classification result on the sth decision tree.

[0031] Preferably, according to each classification result in the N decision trees and the corresponding weight coefficients, a quality classification result of the building board is obtained, including:

[0032] Perform weighted calculation on each classification result in N decision trees according to the weight coefficient to obtain the classification value corresponding to each classification result;

[0033] The classification result with the largest classification value is determined as the quality classification result of the building board.

[0034] The present invention has the following beneficial effects:

[0035] Obtain feature data corresponding to each feature of N groups of building boards; calculate the standard deviation corresponding to each feature based on the feature data corresponding to each feature in the i-th group of building boards in the N groups of building boards; for the i-th decision tree corresponding to the i-th group of building boards, determine the classification condition of the i-th decision tree based on the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and obtain the classification result of the i-th decision tree for the i-th group of building boards; analyze the classification results of the N decision trees for the N groups of building boards, and obtain the quality classification result of the building boards. In this way, the building boards are automatically classified using multiple decision trees and the feature data of the building boards, which improves the efficiency and accuracy of the quality classification of the building boards. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. 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 paying creative work.

[0037] Figure 1 A method flow chart of a method for classifying building board quality based on artificial intelligence provided by one embodiment of the present invention;

[0038] Figure 2 A schematic diagram of determining adjacent feature data on a coordinate axis using feature data provided by an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of a classification result of a decision tree provided by an embodiment of the present invention;

[0040] Figure 4 A schematic diagram of another classification result of a decision tree provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, characteristics and effects of a building board quality classification method based on artificial intelligence proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The specific scheme of the building board quality classification method based on artificial intelligence provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] See also Figure 1 , which shows a method flow chart of a method for classifying building board quality based on artificial intelligence provided by an embodiment of the present invention. In an exemplary embodiment, a method for classifying building board quality based on artificial intelligence is provided, comprising:

[0045] S110, obtaining feature data corresponding to each feature of N groups of building boards;

[0046] S120, calculating the standard deviation corresponding to each feature according to the feature data corresponding to each feature in the i-th group of building boards in the N groups of building boards;

[0047] S130, for the i-th decision tree corresponding to the i-th group of building boards, determine the classification condition of the i-th decision tree according to the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and obtain the classification result of the i-th decision tree for the i-th group of building boards;

[0048] S140, analyzing the classification results of N groups of building boards by N decision trees to obtain quality classification results of the building boards.

[0049] In step S110, illustratively, the building board includes wood board, metal board, cement board, etc. N groups of building boards belong to the same batch of building boards, and N in the N groups of building boards is set according to actual needs. It is generally determined by the number of decision trees, and each decision tree corresponds to a set of data. In order to improve the calculation efficiency, the number of decision trees should not be too large, but if the number of decision trees is too small, it will be greatly affected by the randomly drawn samples, and the final classification result may be inaccurate. In this embodiment, 10 decision trees are set, and N is 10.

[0050] The number of boards in each group of building boards is the same, and the number of boards is set according to actual needs. The boards in each group of building boards can be randomly selected, and the specific extraction method is not limited. For example, 50 boards are randomly selected from each group. Optionally, the boards between each group of building boards can be repeated, but cannot be exactly the same. In this embodiment, in order to obtain a wider coverage of the extracted data and better reflect the overall quality differences of the boards, a sampling method without replacement is used to extract the boards to obtain 10 groups of building boards. In order to obtain a wider coverage of the extracted data and better reflect the overall quality differences of the boards, a sampling method without replacement is used to randomly extract 50 boards from each group.

[0051] Different types of building boards can be provided with the same features or different features, which are not limited here. In this embodiment, taking cement board as an example, its features include surface flatness, color consistency, edge integrity, length, width, thickness, internal density, compressive strength, tensile strength, bending strength, etc. The feature data corresponding to each feature can be directly detected or calculated based on the detected detection value.

[0052] Preferably, step 110 includes: obtaining detection values ​​of various features of N groups of building boards;

[0053] The difference between the preset expected value and the detected value corresponding to each feature is determined as the feature data corresponding to each feature.

[0054] Specifically, taking cement board as an example, the detection values ​​of various features of cement board can be obtained by collecting cement board images using image processing technology to obtain the surface flatness, color consistency and edge integrity of the cement board. The color consistency can also be determined by calculating the degree of color difference in the image, and by extracting edge defects and obtaining the proportion of defects in the cement board. The length, width and thickness can be obtained by direct measurement using various types of rulers. The internal density can be obtained by calculating the density consistency of different positions inside the cement board using ultrasonic imaging technology. The compressive strength, tensile strength and flexural strength can be obtained by testing with a special testing machine.

[0055] A corresponding preset expected value is set for each feature and stored. The preset expected value is generally the production standard of this type of building board. After obtaining the detection value of each feature, the corresponding preset expected value is extracted, the absolute value of the difference between the detection value of the feature and the preset expected value is calculated, and the absolute value of the difference is used as the feature data of the feature, so as to calculate the feature data of each feature. For example, the feature data corresponding to the length of the cement board is the absolute value of the difference between the length detection value of the cement board and the standard length (preset expected value). The feature data corresponding to the color consistency of the cement board is the difference between the grayscale of all pixels in the grayscale image of the cement board surface image and the standard grayscale image. In this embodiment, the feature data corresponding to the building board can be processed.

[0056] In step S120, for example, in the random sampling process, the selected boards may have similarities in certain features, such as the size of the selected cement boards may all meet the specified size standards. Then, when the boards are classified by size features, the classification results cannot effectively reflect the quality differences of the cement boards. Therefore, the standard deviation corresponding to each feature is calculated to determine the feature with the largest quality difference among the features.

[0057] Specifically, taking cement board as an example, all feature data under the surface flatness of the i-th group of building boards are obtained, the mean of all the above feature data is first calculated, and then the variance is calculated based on the mean and all the feature data, and finally the square root of the variance is calculated to obtain the standard deviation corresponding to the surface flatness. According to the above calculation method, the color consistency, edge integrity, length, width, thickness, internal density, compressive strength, tensile strength, bending strength and other characteristics of the cement board are calculated respectively to obtain the standard deviation corresponding to each feature of the i-th group of building boards. Since N groups of building boards can be calculated at the same time, the feature data corresponding to each feature of the remaining groups of building boards are calculated according to the method of calculating the standard deviation of each feature of the i-th group of building boards, and the calculation of the standard deviation corresponding to each feature of the N groups of building boards can be completed.

[0058] In step S130, illustratively, the feature with the largest standard deviation is the feature that reflects the quality difference of the i-th group of building boards. Optionally, a classification threshold may be set in advance for each feature, and the classification threshold may be set according to actual conditions. In this way, after determining the feature with the largest standard deviation among the standard deviations of the various features of the i-th group of building boards, the classification threshold corresponding to the feature is searched, and whether it exceeds the classification threshold is used as the corresponding classification condition. Optionally, the categories of the decision tree are divided into the first category, the second category...the Nth category, and the quality of the building boards in the later categories is better, that is, the farther the distance between the leaf node corresponding to the category of the building board and the root node is, the better the quality. The classification result includes the category and the number of building boards under the category.

[0059] Preferably, the classification condition of the i-th decision tree is determined according to the feature with the largest standard deviation, including:

[0060] According to the degree of dispersion of the feature data corresponding to the feature with the largest standard deviation, the classification threshold corresponding to the feature with the largest standard deviation is determined.

[0061] For example, the degree of dispersion indicates the dispersion of feature data. The densely distributed part is the building board data with similar feature data. It can be considered that these building boards have similar quality levels under this feature. Therefore, the position with the largest degree of dispersion can be selected for classification. For example, the degree of dispersion of feature data is determined by the distance between feature data, and the first value with the largest distance is used as the classification threshold. The classification condition of the i-th decision tree is greater than the first value.

[0062] Furthermore, according to the degree of discreteness of the feature data corresponding to the feature with the largest standard deviation, the classification threshold corresponding to the feature with the largest standard deviation is determined, including: for each feature data in the feature data corresponding to the feature with the largest standard deviation, the distance between each feature data and its adjacent feature data is calculated to obtain the classification threshold corresponding to the feature with the largest standard deviation.

[0063] Specifically, when any feature data has two adjacent feature data, the distances between any feature data and the two adjacent feature data are calculated to obtain a first distance and a second distance, and the sum of the first distance and the second distance is determined as the classification threshold. Figure 2 As shown, all feature data corresponding to the feature with the largest standard deviation are arranged on the coordinate axis, the distance between each feature data and its two adjacent feature data on the coordinate axis is calculated, and then the calculated distances are compared, and the second value with the largest distance is used as the classification threshold, then the classification condition is greater than the second value.

[0064] The distance between each feature data and its two adjacent feature data is calculated as follows:

[0065]

[0066] in, represents the distance between the i-th feature data and its nearest feature data, represents the value of the i-th feature data, Represents the value of the first feature data closest to the i-th feature data, Indicates the value of the second feature data that is closest to the i-th feature data.

[0067] Furthermore, after determining the classification condition, it is determined whether all feature data corresponding to the feature with the largest standard deviation meet the classification condition. If so, the classification result of the i-th decision tree for the i-th group of building boards is generated, that is, the classification result of the i-th group of building boards is determined. If not, continue to determine the feature with the largest standard deviation, calculate the classification threshold based on all feature data corresponding to the feature with the largest standard deviation, and thus determine the next classification condition of the i-th decision tree. The above process is executed repeatedly until all feature data corresponding to the feature with the largest standard deviation meet the classification condition, stop the classification, and output the decision tree.

[0068] In step S140, illustratively, since the N groups of building boards belong to the same batch of building boards and there are N decision trees at this time, the classification results of the N decision trees are needed to be analyzed to determine the quality classification results of the batch of building boards.

[0069] Specifically, taking cement board as an example, due to the randomness of each sampling, the quality of the cement board extracted may be very good or the characteristic value of each quality feature may be close, so that the classification result cannot show the quality difference of cement board. As the number of sampling increases, the classification results of the decision tree will gradually tend to be consistent, that is, the number of times cement boards of different quality levels are extracted will be closer, so the number of decision trees that can better show the overall quality difference of cement board will appear more. Therefore, by analyzing the occurrence of the classification conditions used in the classification results at the same position in N decision trees, for example, the occurrence can be the number of occurrences, and the corresponding weight coefficients are set in advance for different numbers of occurrences. In this way, after counting the number of occurrences of the classification conditions used in the classification results at the same position, the corresponding weight coefficient can be found by the number of occurrences. According to the classification results and corresponding weight coefficients in the N decision trees, the quality classification results of the building board are obtained, thereby determining the quality of the building board.

[0070] In the technical solution of the present application, feature data corresponding to each feature of N groups of building boards are obtained; based on the feature data corresponding to each feature in the i-th group of building boards in the N groups of building boards, the standard deviation corresponding to each feature is calculated; for the i-th decision tree corresponding to the i-th group of building boards, the classification condition of the i-th decision tree is determined based on the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and the classification result of the i-th decision tree for the i-th group of building boards is obtained; the classification results of the N decision trees for the N groups of building boards are analyzed to obtain the quality classification results of the building boards. In this way, the building boards are automatically classified using multiple decision trees and the feature data of the building boards, which improves the classification efficiency and accuracy of the building boards, and at the same time gets rid of the reliance on manual classification of the boards.

[0071] In one embodiment, for the i-th decision tree corresponding to the i-th group of building boards, the classification condition of the i-th decision tree is determined according to the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and the classification result of the i-th decision tree for the i-th group of building boards is obtained, step S130 includes:

[0072] When the feature with the largest standard deviation is determined for the jth time, the jth classification condition of the i-th decision tree is determined based on the feature with the largest standard deviation;

[0073] When the feature data corresponding to the feature with the largest standard deviation all meet the jth classification condition, the i-th decision tree outputs j classification results for the i-th group of building boards;

[0074] When the feature data corresponding to the feature with the largest standard deviation does not satisfy the j-th classification condition, the feature with the largest standard deviation is reselected from the features for which the classification condition has not been determined, and the j+1-th classification condition of the ith decision tree is determined based on the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all satisfy the determined classification condition; wherein j is a positive integer.

[0075] Specifically, Figure 3 As shown, the building board is cement board, the classification condition is greater than the classification threshold, and j is 1. If the feature with the largest standard deviation is surface flatness, when all feature data under the surface flatness of the cement board Both greater than When the surface flatness is less than 0.000, the cement board is classified into the first category, which is considered to be the worst quality category, and the classification is stopped and no further classification is performed. In the case of , the number of panels greater than the classification threshold is recorded under the first category of the i-th decision tree and used as the classification result corresponding to the first category of the i-th decision tree. Then continue to classify the i-th group of building panels.

[0076] Since the surface flatness of the cement board has been considered in the first classification, the classification conditions that have appeared will not be considered in the subsequent classification. In the second classification, the feature with the largest standard deviation is determined among color consistency, edge integrity, length, width, thickness, internal density, compressive strength, tensile strength, and flexural strength. If it is thickness, the classification threshold is calculated based on all feature data corresponding to the thickness, thereby determining the second classification condition, that is, greater than When the thickness of cement board is Both greater than When the thickness is less than 0.000, it is classified into the second category and the classification is stopped. In the case of , the above processing method is repeated to continue to classify the i-th group of building boards until the feature data corresponding to the feature with the largest standard deviation all meet the classification conditions determined by it. In this way, the classification of the i-th decision tree is realized, so that the above method can realize the classification of N decision trees.

[0077] Furthermore, the method further comprises:

[0078] When the feature data corresponding to the feature with the largest standard deviation does not meet the j-th classification condition, if the maximum standard deviation is reselected from the features whose classification conditions have not been determined and is not greater than the preset standard deviation threshold, the building boards in the i-th group of building boards that do not meet the j-th classification condition will be determined as the j+1-th classification result.

[0079] Exemplarily, the preset standard deviation threshold is set according to actual conditions and is not limited here. For example, the preset standard deviation threshold can be set to 1.

[0080] Specifically, Figure 4 As shown, taking cement board as an example, in the process of producing cement board, defects may only be caused by individual parameter settings, environmental influences and other factors, and there will not be large quality differences in each feature of cement board. Therefore, when classifying, the feature with the largest standard deviation corresponds to When , the quality level of the remaining cement boards is considered to be the same, and they are directly classified into one category (j+1 category), and the classification is stopped to obtain the final decision tree. In this way, the generation efficiency of the decision tree is improved, thereby improving the classification efficiency of building boards.

[0081] In one embodiment, the classification results of N groups of building boards by N decision trees are analyzed to obtain the quality classification results of the building boards. Step S140 includes:

[0082] S1410, determining a weight coefficient of each classification result in the N decision trees according to the occurrence of the classification conditions used by the classification results at the same position in the N decision trees;

[0083] S1420. Obtain quality classification results of building boards according to the classification results in the N decision trees and the corresponding weight coefficients.

[0084] Preferably, step S1420 includes: performing weighted calculation on each classification result in the N decision trees according to the weight coefficient to obtain a classification value corresponding to each classification result; and determining the classification result with the largest classification value as the quality classification result of the building board.

[0085] Specifically, after determining the weight coefficient of each classification result in N decision trees, the classification value calculation is performed on all the first categories in the N decision trees, and the classification value calculation is performed on all the second categories in the N decision trees..., and the calculation is performed until the last category. Among them, the specific method of classification value calculation is to multiply the weight coefficient corresponding to the classification result of the first category of each decision tree by the number of building boards under the classification result of the first category, to obtain the classification value of the first category under each decision tree, and to add all the above classification values ​​to obtain the final classification value corresponding to the first category. According to the above method, the classification value calculation is performed on each classification result to obtain the final classification value corresponding to each classification result. In this embodiment, the final classification value is recorded in a table as a voting value. Finally, the final classification values ​​corresponding to each classification result are compared, and the classification result with the largest value is used as the quality classification result of the building board. The higher the classification result, the worse the quality of the building board, that is, the quality of the first category is worse than the quality of the second category, the quality of the second category is worse than the quality of the third category, and so on. In this way, the quality of the building board is automatically classified to improve the accuracy of the board quality classification. At the same time, when it is necessary to classify new building panels of the same type, it is only necessary to put the corresponding feature values ​​into the generated decision tree to obtain the classification results.

[0086] In one embodiment, according to the occurrence of the classification conditions used by the classification results at the same position in the N decision trees, the weight coefficients of the classification results in the N decision trees are determined, and step S1410 includes:

[0087] Determine the first classification credibility of the m-th classification result according to the number of classification conditions used by the m-th classification result on each decision tree in the N decision trees;

[0088] According to the classification conditions used by the m-th classification result on the s-th decision tree, the number of occurrences of the same position in the N decision trees, the second classification credibility of the m-th classification result on the s-th decision tree is determined;

[0089] The first classification credibility and the second classification credibility are calculated to obtain the weight coefficient of the mth classification result on the sth decision tree.

[0090] Specifically, since different features have different degrees of expression of building board quality differences, when a feature is more obvious in the expression of building board quality differences, the feature will appear repeatedly as a classification condition in the same classification position of all decision trees, and the result obtained by using the feature to judge the quality difference of building boards will have a higher degree of discrimination. The first classification credibility indicates the degree of significance of each classification position in the expression of quality differences.

[0091] The calculation formula of the first category credibility is as follows:

[0092]

[0093] in, Indicates the first classification credibility of the m-th classification result (i.e., the significance of the quality difference of the m-th classification node). Indicates the number of classification conditions that have appeared in the mth classification node (i.e., classification result).

[0094] Further, according to the classification condition used by the m-th classification result on the s-th decision tree and the number of occurrences of the same position in the N decision trees, the second classification credibility of the m-th classification result on the s-th decision tree is determined, including:

[0095] The ratio of the number of occurrences of the classification condition used for the mth classification result on the sth decision tree at the same position in the N decision trees to the total number N of all decision trees is determined as the second classification credibility of the mth classification result on the sth decision tree.

[0096] Specifically, the calculation formula of the second category credibility is as follows:

[0097]

[0098] in, Indicates the second classification credibility of the m-th classification result on the s-th decision tree, It indicates the number of times the feature condition corresponding to the mth classification result on the sth decision tree appears at that position on all decision trees. Represents the total number of decision trees.

[0099] Due to the randomness of sampling, the classification results on some decision trees may not be consistent with the significance of the classification position on the quality difference of building boards. That is, when the credibility of the classification result on the decision tree is low, even if the classification position has a significant effect on the quality difference of building boards, the weight of the impact of the classification result on the final classification result of cement boards should be relatively small. For example, if the surface flatness of cement boards is always used as the classification condition in the first classification, the significance of the position on the quality difference of cement boards is relatively large. When a decision tree uses the color consistency of cement boards as the classification condition in the first classification, the classification result does not conform to the overall quality classification of cement boards, so the impact on the final classification result is relatively small. Therefore, the weight of the impact of each classification result on the final classification result on each decision tree is obtained:

[0100]

[0101] in, Represents the weight coefficient of the mth classification result on the sth decision tree, Indicates the second classification credibility of the m-th classification result on the s-th decision tree, Indicates the first classification confidence of the mth classification result.

[0102] In other embodiments, a device is provided, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device executes the above-mentioned building board quality classification method based on artificial intelligence. The device can be a chip, a component or a module, and the chip can include a connected processor and a memory; wherein the memory is used to store instructions, and when the processor calls and executes the instructions, the chip can execute the building board quality classification method based on artificial intelligence provided in the above embodiment.

[0103] In other embodiments, a computer program product is also provided. When the computer program product is run on a computer, the computer is enabled to execute the above-mentioned related steps to implement an artificial intelligence-based building board quality classification method provided in the above-mentioned embodiment.

[0104] In other embodiments, a computer-readable storage medium is also provided, in which a computer program code is stored. When the computer program code is run on a computer, the computer executes the above-mentioned related method steps to implement an artificial intelligence-based building board quality classification method provided in the above embodiment.

[0105] Among them, the provided systems, devices, computer program products, and computer-readable storage media are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above and will not be repeated here.

[0106] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. 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.

[0107] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A building board quality classification method based on artificial intelligence, characterized in that: The method comprises: Obtaining feature data corresponding to each feature of N groups of building boards; the feature data at least includes surface flatness, color consistency, edge integrity, length, width, thickness, internal density, compressive strength, tensile strength, and bending strength; Calculate the standard deviation corresponding to each feature according to the feature data corresponding to each feature in the i-th group of building boards in the N groups of building boards; For the i-th decision tree corresponding to the i-th group of building boards, determine the classification condition of the i-th decision tree according to the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and obtain the classification result of the i-th decision tree for the i-th group of building boards; Analyze the classification results of N groups of building boards by N decision trees to obtain the quality classification results of building boards; wherein N and i are positive integers; The i-th decision tree corresponding to the i-th group of building boards determines the classification condition of the i-th decision tree according to the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all meet the classification condition, and obtains the classification result of the i-th decision tree for the i-th group of building boards, including: When the feature with the largest standard deviation is determined for the jth time, determining the jth classification condition of the i-th decision tree according to the feature with the largest standard deviation; When the feature data corresponding to the feature with the largest standard deviation all satisfy the j-th classification condition, the i-th decision tree outputs j classification results for the i-th group of building boards; In the case where the feature data corresponding to the feature with the largest standard deviation does not satisfy the j-th classification condition, the feature with the largest standard deviation is reselected from the features for which the classification condition has not been determined, and the j+1-th classification condition of the i-th decision tree is determined based on the feature with the largest standard deviation, until the feature data corresponding to the feature with the largest standard deviation all satisfy the determined classification condition; wherein j is a positive integer; The determining the classification condition of the i-th decision tree according to the feature with the largest standard deviation includes: According to the degree of discreteness of the feature data corresponding to the feature with the largest standard deviation, the classification threshold corresponding to the feature with the largest standard deviation is determined; and the classification condition of the i-th decision tree is determined to be greater than the classification threshold.

2. The method for classifying building board quality based on artificial intelligence according to claim 1, characterized in that: The step of determining the classification threshold corresponding to the feature with the largest standard deviation according to the degree of dispersion of the feature data corresponding to the feature with the largest standard deviation includes: For each feature data in the feature data corresponding to the feature with the largest standard deviation, the distance between each feature data and its adjacent feature data is calculated to obtain the classification threshold corresponding to the feature with the largest standard deviation.

3. The method for classifying building board quality based on artificial intelligence according to claim 1, characterized in that: The step of obtaining feature data corresponding to each feature of N groups of building boards includes: Obtaining detection values ​​of various features of the N groups of building boards; The difference between the preset expected value and the detected value corresponding to each feature is determined as the feature data corresponding to each feature.

4. The method for classifying building board quality based on artificial intelligence according to claim 1, characterized in that: The method further comprises: When the feature data corresponding to the feature with the largest standard deviation does not satisfy the j-th classification condition, if the features with the maximum standard deviation not greater than the preset standard deviation threshold are reselected from the features for which the classification condition has not been determined, the building boards in the i-th group of building boards that do not satisfy the j-th classification condition are determined as the j+1-th classification result.

5. The method for classifying building board quality based on artificial intelligence according to claim 1, characterized in that: The classification results of N groups of building boards by N decision trees are analyzed to obtain the quality classification results of the building boards, including: Determine the weight coefficient of each classification result in the N decision trees according to the occurrence of the classification conditions used by the classification results at the same position in the N decision trees; According to each classification result in the N decision trees and the corresponding weight coefficients, a quality classification result of the building board is obtained.

6. The method for classifying building board quality based on artificial intelligence according to claim 5, characterized in that: Determining the weight coefficient of each classification result in the N decision trees according to the occurrence of the classification conditions used by the classification results at the same position in the N decision trees includes: Determining a first classification credibility of the m-th classification result according to the number of classification conditions used by the m-th classification result on each decision tree in the N decision trees; Determine the second classification credibility of the mth classification result on the sth decision tree according to the number of occurrences of the same position in the N decision trees based on the classification condition used by the mth classification result on the sth decision tree; The first classification credibility and the second classification credibility are calculated to obtain a weight coefficient of the mth classification result on the sth decision tree; wherein s and m are positive integers.

7. The method for classifying building board quality based on artificial intelligence according to claim 6, characterized in that: Determining the second classification credibility of the mth classification result on the sth decision tree according to the classification condition used by the mth classification result on the sth decision tree and the number of occurrences of the same position in the N decision trees includes: The ratio of the number of occurrences of the classification condition used by the mth classification result on the sth decision tree at the same position in the N decision trees to the total number N of all decision trees is determined as the second classification credibility of the mth classification result on the sth decision tree.

8. A method for classifying building board quality based on artificial intelligence according to any one of claims 5 to 7, characterized in that: The quality classification result of the building board is obtained according to each classification result in the N decision trees and the corresponding weight coefficient, including: Performing weighted calculation on each classification result in the N decision trees according to the weight coefficient to obtain a classification value corresponding to each classification result; The classification result with the largest classification value is determined as the quality classification result of the building board.

Images