Strip steel surface cleanliness detection method, detection device and system
Through the analysis and classification of the spectral data of the strip surface, the spectral line offset problem caused by the vibration of the conveyor belt is solved, and the accuracy of the surface cleanliness detection of the strip surface is improved.
Patent Information
- Application Number
- CN202411267711.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-09-11
AI Technical Summary
In the surface cleanliness detection of strip steel, the spectral line obtained by the spectral instrument is offset due to the vibration of the conveyor belt, which affects the accuracy of the detection results.
By collecting spectral data on the surface of the strip at multiple preset wavelengths, the peaks of each data point are obtained, and divided into different types of collections and substances. Combining the peak offset degree and clustering clusters, we can determine whether substances are affected by cleaning. Form, calculate surface cleanliness.
The influence of spectral line deviation caused by conveyor belt vibration is eliminated, and the accuracy of strip surface cleanliness detection is improved.
Smart Images

Figure CN118794898B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of strip steel spectral analysis, and in particular to a strip steel surface cleanliness detection method, detection device and system. Background Art
[0002] Strip steel, as an important industrial material, is specially produced by various steel rolling enterprises to meet the diverse needs of different industrial sectors for metal or mechanical products in industrial production. The surface cleanliness of strip steel refers to the degree of adhesion of pollutants such as grease, dirt, rust and scale on the surface of strip steel, and is one of the important indicators for measuring the surface quality of strip steel. If the surface cleanliness of strip steel does not meet the standard, defects such as granulation and scarring may be formed during the production process. Especially in the production of coated and plated plates, the residual carbon content after annealing is too high, which will also affect the surface quality of the finished product and cause flatness and adhesion.
[0003] With the rapid development of spectroscopy technology, the excellent performance of spectroscopy technology in object composition detection has become increasingly prominent, especially when analyzing residues on the surface of objects through spectroscopy technology, showing fast and accurate detection capabilities. However, in practical applications, materials such as strip steel are usually located on a conveyor belt. The movement of the conveyor belt will cause the strip steel to vibrate, resulting in the deviation of the spectral lines obtained by the spectrometer, which will cause errors in identifying substances through spectral data, thereby affecting the accuracy of the strip steel surface cleanliness detection results. Summary of the invention
[0004] In view of the above, it is necessary to provide a strip steel surface cleanliness detection method, a detection device and a system, which improve the accuracy of strip steel surface cleanliness detection compared with the traditional strip steel surface cleanliness detection method:
[0005] In a first aspect, an embodiment of the present application provides a method for detecting the cleanliness of a steel strip surface, the method comprising the following steps:
[0006] Collecting spectral data images of the strip surface at multiple preset wavelengths, wherein one pixel is recorded as one strip data point, and collecting spectral data of each substance at the multiple preset wavelengths, wherein there is a cleaning-affecting substance;
[0007] Obtaining the peak value of the spectral data corresponding to each steel strip data point at all wavelengths; and dividing all steel strip data points into sets of steel strip types based on the number of the peak values of each steel strip data point;
[0008] Calculate the average level of wavelength difference between any two strip steel data points in each strip steel type set at all peak points with the same sequence number, and divide the strip steel data points in each strip steel type set into classes with similar peak positions based on the distribution of all the average levels corresponding to each strip steel type set;
[0009] Based on the peak difference at all the same sequence numbers between any two strip steel data points in any peak position similarity class and the degree of dispersion of the peak difference, determine the peak shift degree between any two strip steel data points;
[0010] Obtaining clusters of all the peak shift degrees corresponding to each class with similar peak positions, combining the difference between the peak shift degrees in different clusters of each class with similar peak positions and the fluctuation degree of the peak shift degree in each cluster, dividing all the strip steel data points in each class with similar peak positions into various material classes;
[0011] Determining whether the steel strip data points in any substance class are formed by a cleaning-influencing substance based on similarities between all steel strip data points in any substance class and the spectral data of each substance at the same wavelength;
[0012] The surface cleanliness of the steel strip is determined based on the proportion of steel strip data points belonging to cleanliness-affecting substances in the spectral data image.
[0013] In one embodiment, the method of dividing all the steel strip data points into sets of steel strip types is: dividing all the steel strip data points with the same number of peak values into the same set of steel strip types.
[0014] In one embodiment, the process of dividing the strip data points in each strip type set into classes with similar peak positions is:
[0015] Arrange the wavelengths of all peak points of each steel strip data point in ascending order to form a peak position sequence of each steel strip spectrum;
[0016] The average value of the difference of all elements at the same position in the strip steel spectrum peak position sequence of the arbitrary two strip steel data points is used as the strip steel vibration offset value between the arbitrary two strip steel data points;
[0017] Perform threshold segmentation on all the strip steel vibration offset values corresponding to any strip steel type set, and form a co-located strip steel vibration offset set with all the strip steel vibration offset values in any strip steel type set that are less than or equal to the segmentation threshold;
[0018] The two strip data points corresponding to each strip vibration offset value in any iso-position strip vibration offset set are taken as one class, the classes with at least one identical strip data point are merged, and the classes obtained after the merger are taken as classes with similar peak positions.
[0019] In one embodiment, the process of determining the peak shift degree is as follows:
[0020] Arrange all the peak values of each steel strip data point in ascending order of wavelength to form a peak value sequence of each steel strip spectrum;
[0021] The differences between all elements at the same position in the strip steel spectrum peak sequence of any two strip steel data points are used to form a strip steel peak deviation sequence according to the element positions;
[0022] The expression of the peak deviation degree between any two strip data points is:
[0023] ; In the formula, Indicates the peak deviation degree between the xth and yth strip data points in any peak position similar class; , Respectively represent the strip steel spectrum peak sequences of the xth and yth strip steel data points in any of the similar peak position classes; represents the strip peak deviation sequence between the xth and yth strip data points in any peak position similar class; Cos() represents the cosine similarity function; MAD() represents the degree of dispersion function; Indicates a preset value greater than 0.
[0024] In one embodiment, the process of dividing all the strip data points in each class with similar peak positions into each material class is as follows:
[0025] A clustering algorithm is used to obtain clusters of all the peak shift degrees corresponding to any class with similar peak positions, and a cluster with the largest mean value of the peak shift degree and a cluster with the smallest mean value are obtained, which are recorded as the largest cluster and the smallest cluster respectively;
[0026] The expression for the determination value of the substance type of any peak position similar class is:
[0027] ; Wherein, GF represents the substance type determination value of any peak position similar class; P and Q represent the maximum cluster and the minimum cluster of any peak position similar class respectively; , Respectively represent the discrete degree of the peak shift degree in the largest cluster and the smallest cluster of any peak position similar class; max() and min() represent the maximum value function and the minimum value function respectively;
[0028] When the material type judgment value of any of the classes with similar peak positions is less than or equal to 0, any of the classes with similar peak positions is regarded as a material class; otherwise, the two strip steel data points corresponding to the peak deviation degrees in the largest cluster are divided into one class, and the classes with at least one identical strip steel data point are merged, and each class obtained after the merger is recorded as a material class.
[0029] In one embodiment, the process of determining whether any of the substance classes corresponds to a cleaning-affecting substance is:
[0030] Calculate the average values of the spectral data of all the strip data points in any material class at the same wavelength, and arrange the average values in ascending order of wavelength to form a material sequence to be determined;
[0031] Arrange the spectrum data of each substance in ascending order of wavelength to form each spectrum sequence;
[0032] The similarity between the undetermined material sequence of any material class and the spectral sequences of various materials is calculated. When the material corresponding to the spectral sequence with the greatest similarity is a cleaning-affecting substance, it is determined that the strip data points in any material class are formed by the cleaning-affecting substance.
[0033] In one embodiment, the expression of the surface cleanliness is:
[0034] ; Wherein, B represents the surface cleanliness of the steel strip; M represents the number of steel strip data points in the spectral data image; N represents the number of steel strip data points belonging to cleanliness-affecting substances in the spectral data image.
[0035] In one embodiment, when the surface cleanliness of the steel strip is greater than a preset cleanliness threshold, the quality of the steel strip is determined to be qualified; otherwise, the quality of the steel strip is determined to be unqualified.
[0036] In a second aspect, an embodiment of the present application further provides a strip steel surface cleanliness detection device, the device comprising:
[0037] A spectral data acquisition module is used to acquire spectral data images of the strip surface at multiple preset wavelengths, wherein one pixel is recorded as one strip data point, and spectral data of each substance at the multiple preset wavelengths are respectively acquired, wherein there are substances that affect cleaning;
[0038] A spectral data classification module is used to obtain the peak value of the spectral data corresponding to each strip steel data point at all wavelengths; based on the number of the peak values of each strip steel data point, all strip steel data points are divided into sets of each strip steel type;
[0039] Calculate the average level of wavelength difference between any two strip steel data points in each strip steel type set at all peak points with the same sequence number, and divide the strip steel data points in each strip steel type set into classes with similar peak positions based on the distribution of all the average levels corresponding to each strip steel type set;
[0040] Based on the peak difference at all the same sequence numbers between any two strip steel data points in any peak position similarity class and the degree of dispersion of the peak difference, determine the peak shift degree between any two strip steel data points;
[0041] Obtaining clusters of all the peak shift degrees corresponding to each class with similar peak positions, combining the difference between the peak shift degrees in different clusters of each class with similar peak positions and the fluctuation degree of the peak shift degree in each cluster, dividing all the strip steel data points in each class with similar peak positions into various material classes;
[0042] A cleanliness detection module, for determining whether a strip data point in any substance class is formed by a cleanliness-affecting substance based on similarities between all strip data points in any substance class and spectral data of each substance at the same wavelength;
[0043] The surface cleanliness of the steel strip is determined based on the proportion of steel strip data points belonging to cleanliness-affecting substances in the spectral data image.
[0044] In a third aspect, an embodiment of the present application also provides a strip surface cleanliness detection system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above-mentioned strip surface cleanliness detection methods when executing the computer program.
[0045] This application has at least the following beneficial effects:
[0046] The present application performs a first classification of the spectral data points of the steel strip by analyzing the number of peaks in the spectral data of the steel strip data points. Secondly, the first classification result is secondary classified by using the difference in peak positions. Thirdly, the secondary classification result is classified in combination with the difference between the peaks to obtain each material class. The steel strip data points in each material class are formed by the same material. Further, based on the average level of the spectral data of the steel strip data points in each material class and the similarity of the spectral data of each material, it is judged whether the steel strip data points in each material class are formed by cleaning-affecting materials, thereby eliminating the influence of the spectral line deviation caused by the vibration of the conveyor belt on the material identification result, thereby improving the accuracy of detecting the cleanliness of the steel strip surface. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0048] Figure 1 A flow chart of the steps of a method for detecting the cleanliness of a steel strip surface provided in one embodiment of the present application;
[0049] Figure 2 It is a detection schematic diagram of the AMPD algorithm;
[0050] Figure 3 It is a schematic diagram of the synthesis of the classes with similar peak positions;
[0051] Figure 4 Schematic diagram of the process of obtaining peaks with similar positions. DETAILED DESCRIPTION
[0052] In the description of the embodiments of the present application, words such as "exemplary", "or", "for example" and the like are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "or", "for example" and the like is intended to present related concepts in a concrete manner.
[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. It should be understood that, unless otherwise specified, " / " means or.
[0054] It should also be noted that the terms "first" and "second" in the present application are used to distinguish similar objects rather than to describe a specific order or sequence.
[0055] The specific scheme of the strip surface cleanliness detection method, detection device and system provided in the present application is described in detail below with reference to the accompanying drawings.
[0056] See also Figure 1 , which shows a flow chart of the steps of a strip steel surface cleanliness detection method provided by an embodiment of the present application, the method comprising the following steps:
[0057] Step 1, collecting spectral data images of the strip surface at multiple preset wavelengths, wherein one pixel is recorded as a strip data point, and collecting spectral data of each substance at the multiple preset wavelengths respectively, among which there are cleaning-affecting substances.
[0058] A multi-spectral imager is installed vertically above the conveyor belt of the conveyor steel strip, ensuring that its shooting angle is 90 degrees to the horizontal plane, so as to directly aim at the surface of the strip steel and obtain accurate spectral data. Since the main residues on the surface of the strip steel are rolling oil and carbon elements, these two substances have a strong reaction in the wavelength range of 200~1000nm, so the multi-spectral imager is used to collect the spectral absorption rate data of the strip steel surface in the wavelength range of 200~1000nm, which is recorded as spectral data. For the collected spectral data image, one pixel is recorded as a strip steel data point.
[0059] A spectral imaging instrument was used to collect spectral data of each substance in the wavelength range of 200-1000nm. The substances include rolling oil, carbon element, emulsion, iron powder and strip steel, among which rolling oil, carbon element, emulsion and iron powder are cleaning-affecting substances.
[0060] Since spectral data may be lost during the collection or transmission process, the missing value filling method is used to fill the missing spectral data. For the completed spectral data, all spectral data of each strip data point are arranged in order from small to large wavelength to form each spectral sequence; all spectral data of various substances are arranged in order from small to large wavelength to form each spectral sequence.
[0061] In this embodiment, the mean filling method is used for filling. As other implementation methods, on the basis of achievable data filling, the implementer can use other existing technologies for data filling, such as the median filling method, interpolation method, etc., and this application does not make any special restrictions.
[0062] Step 2, obtaining the peak value of the spectral data corresponding to each steel strip data point at all wavelengths; based on the number of the peak values of each steel strip data point, dividing all steel strip data points into sets of each steel strip type.
[0063] The main production process of strip steel is pickling-rolling-alkali washing-annealing-coating-slitting. Before coating the strip steel, the surface must be clean to ensure the quality and adhesion of the coating and produce higher quality strip steel products. Since the strip steel products are large in size and weight, they cannot be transported manually, so they are generally transported by conveyor belts. When the strip steel is transported to the inspection area via a conveyor belt, the vibration of the conveyor belt causes the spectral line of the strip steel spectrum data obtained by the multi-spectral imager to shift, so the characteristic peaks of the strip steel spectrum data shift. However, since the spectral data is determined by the absorption of the wavelength by the substance, although the characteristic peaks will shift, the order of multiple characteristic peaks of a substance will not change.
[0064] The spectrum sequence of each strip data point is used as the input of the automatic multiscale-based peak detection (AMPD) algorithm, and the peak value in the spectrum sequence of each strip data point and the wavelength corresponding to the peak value are output. The wavelengths corresponding to all peak values in the spectrum sequence of each strip data point are arranged in ascending order to form a spectrum peak position sequence for each strip. The AMPD algorithm is a well-known technology and will not be described in detail in this application. The detection schematic diagram of the AMPD algorithm is shown in FIG. Figure 2 As shown in the figure, the horizontal axis is wavelength and the vertical axis is absorbance.
[0065] Since different substances have different absorption characteristics, they produce unique peak combinations in specific bands. Therefore, substances with different numbers of peaks in a certain band are different types of substances. Therefore, for all strip data points, they are classified according to the length of the strip spectral peak sequence, and all strip data points with the same strip spectral peak sequence length are classified into the same strip type set.
[0066] Step 3, calculate the average level of wavelength difference between any two strip data points in each strip type set at all peak points with the same serial number, and based on the distribution of all the average levels corresponding to each strip type set, divide the strip data points in each strip type set into classes with similar peak positions.
[0067] Taking any set of strip steel types as an example, although the spectral lines are shifted due to the vibration of the conveyor belt, the peaks of the spectral data of the same material are close to each other.
[0068] Based on the difference between the strip steel spectrum peak position sequences of any two strip steel data points in the strip steel type set, the strip steel vibration offset value between the any two strip steel data points is determined, and the expression is:
[0069] ; In the formula, represents the strip vibration offset value between the i-th and j-th strip data points in the strip type set; n represents the length of the strip spectrum peak position sequence of any strip data point in the strip type set; , They respectively represent the kth element in the strip steel spectrum peak position sequence of the i-th and j-th strip steel data points in the strip steel type set.
[0070] It should be noted that: ideally, for two strip data points of the same material, if the spectral data are not offset, the elements at the same position in the strip spectral peak sequence of the two strip data points are the same, so that the absolute value of the difference between the elements in the two strip spectral peak sequences is The value of is 0, and the strip vibration offset value between the strip data points is 0; when there is a shift in the spectral data, the elements in the strip spectral peak sequence of the same substance are less different than those of different substances, that is, Small, so that the strip vibration offset value between strip data points When the value of is small, when all the strip vibration offset values are small, it means that the strip data points in the strip type set correspond to the same substance. If the substance is a substance that affects the cleanliness of the strip, such as rolling oil, carbon powder or iron powder, the more elements in the strip type set, the lower the surface cleanliness of the strip.
[0071] If the strip type set is composed of strip data points of different materials, since the strip vibration offset values between strip data points of different materials are larger, and the strip vibration offset values between strip data points of the same material are smaller, all strip vibration offset values are used as input of the threshold segmentation algorithm, and the segmentation threshold of the strip vibration offset value is output. All strip vibration offset values that are less than or equal to the segmentation threshold are grouped into a same-position strip vibration offset set, and all strip vibration offset values that are greater than the segmentation threshold are grouped into a different-position strip vibration offset set.
[0072] In this embodiment, the Otsu threshold segmentation algorithm is used to obtain the segmentation threshold. As other implementation methods, on the basis of being able to obtain the segmentation threshold, the implementer may adopt other existing technologies to obtain the segmentation threshold, such as global threshold segmentation, iterative threshold segmentation, etc., and this application does not impose any special restrictions.
[0073] The two strip data points corresponding to each strip vibration offset value in the same-position strip vibration offset set are taken as one class, the classes with at least one identical strip data point are merged, and the classes obtained after the merger are taken as classes with similar peak positions.
[0074] The synthesis diagram of the peak positions is as follows Figure 3 As shown, for example, the binary groups corresponding to each element in the vibration offset set of the same-position strip are (1,2), (2,5), (5,8), (3,7), (9,3), and (7,9). A binary group consists of two strip data points and is considered as a class. Since (1,2) and (2,5) have the same strip data point 2, the two classes are merged into (1,2,5). Since (5,8) also has strip data point 5, (1,2,5) is merged with (5,8) to obtain (1,2,5,8). Similarly, (3,7), (9,3), and (7,9) are merged into (3,7,9). Then (1,2,5,8) and (3,7,9) are both classes with similar peak positions. The schematic diagram of the process of obtaining classes with similar peak positions is shown in the figure. Figure 4 shown.
[0075] Step 4: Determine the peak deviation degree between any two strip data points based on the peak difference between any two strip data points in any peak position similarity class at all positions with the same sequence number and the discrete degree of the peak difference.
[0076] Taking any peak position similar category as an example, when calculating the surface cleanliness of the strip, rolling oil, carbon powder, iron powder and other substances are not evenly distributed on the surface of the strip, and may have different concentrations in different areas. Since different concentrations of the same substance have different absorption rates for the same wavelength, the higher the concentration of the substance, the stronger the absorption capacity for the wavelength, and vice versa, the lower the concentration, the lower the absorption capacity for the wavelength. This property exists for all wavelengths. Therefore, at different concentrations of the same substance, the peak value of the spectral characteristic peak will have a certain difference, which makes it easy to make mistakes when identifying substances through spectral data, and then leads to errors when calculating the surface cleanliness of the strip through the material identification results. Therefore, it is necessary to eliminate the problem of errors in material identification of strip data points due to concentration differences.
[0077] All peaks in the spectrum sequence of each strip data point are arranged in order of wavelength from small to large to form each strip spectrum peak sequence.
[0078] Based on the above analysis, the difference between the elements at the same position in the strip steel spectrum peak sequence of any two strip steel data points in the class with similar peak positions is calculated, which is recorded as the strip steel peak deviation; when two strip steel spectrum peak sequences are formed by the same substance with the same concentration, the substance has the same absorption rate at the same wavelength. Therefore, the difference between the elements at the same position in the two strip steel spectrum peak sequences is small, that is, the value of the strip steel peak deviation is small.
[0079] In this embodiment, the difference between the elements at the same position in the strip spectral peak sequence is the absolute value of the difference. As other implementation methods, on the basis of being able to measure the difference between the elements at the same position in the strip spectral peak sequence, the implementer may adopt other calculation methods to measure the difference between the elements at the same position in the strip spectral peak sequence, such as ratio, square of difference, etc., and this application does not impose any special restrictions.
[0080] The strip peak deviations between all elements at the same position in the strip spectrum peak sequence of any two strip data points are combined into a strip peak deviation sequence according to the element positions.
[0081] Although concentration differences will cause differences between peaks, the differences between peaks show a regular change, that is, the positions of the peaks shift with changes in concentration without changing the peak shape or introducing new peaks.
[0082] Therefore, based on the similarity of the strip steel spectrum peak sequence between any two strip steel data points and the discrete degree of the strip steel peak deviation in the strip steel peak deviation sequence between them, the peak deviation degree between any two strip steel data points is determined, and the expression is:
[0083] ; In the formula, Indicates the degree of peak deviation between the xth and yth strip data points in the class with similar peak positions; , They represent the strip spectrum peak sequences of the xth and yth strip data points in the class with similar peak positions respectively; represents the strip peak deviation sequence between the xth and yth strip data points in the peak position similar class; Cos() represents the cosine similarity function; MAD() represents the discrete degree function; Indicates that the value is preset to be greater than 0, in order to avoid the denominator being 0. The value of is preset by humans and can be set by the implementer. The value is 1.
[0084] In this embodiment, the degree of discreteness of the strip peak deviation in the strip peak deviation sequence is the mean absolute deviation. As other implementation methods, on the basis of being able to measure the uneven distribution of the strip peak deviation, the implementer may use other existing technologies to measure the uneven distribution of the strip peak deviation, such as variance, standard deviation, coefficient of variation, etc., and this application does not impose any special restrictions.
[0085] It should be noted that when two strip steel data points are formed by the same substance, such as rolling oil and carbon, although the peaks in the strip steel spectrum peak sequence of the two strip steel data points may not be in the same position due to the different concentrations of the substance, the peak position shifts as a whole with the change of concentration, so the cosine similarity between the two strip steel spectrum peak sequences The difference between the peak deviations of the two strip steel spectrum peak sequences at different positions is small, that is, the degree of dispersion is small, which makes the peak deviation degree The value of is larger.
[0086] Step 5, obtain the clusters of all the peak shift degrees corresponding to each peak position similar class, combine the differences between the peak shift degrees in different clusters of each peak position similar class, and the fluctuation degree of the peak shift degree in each cluster, and divide all the strip data points in each peak position similar class into various material classes.
[0087] If a strip data point is formed by a cleaning-affecting substance, the strip data point with a greater peak deviation from the strip data point is more likely to be formed by an impurity substance. In order to distinguish which substance the strip data point is formed by, it is necessary to classify the strip data points in the class with similar peak positions to determine the area of the substance on the strip surface.
[0088] Based on the above analysis, the peak shift values obtained for the same material are large and close, and have certain concentrated characteristics in the distribution of values. If all the strip data points in the peak-similar class are formed by the same material on the strip surface, then all the peak shifts have the same fluctuation characteristics. Therefore, all the peak shifts corresponding to the peak-similar class are used as the input of the clustering algorithm, and each cluster is output. The mean of all the peak shifts in each cluster is calculated, and the cluster with the largest mean is recorded as the largest cluster, and the cluster with the smallest mean is recorded as the smallest cluster.
[0089] In this embodiment, K-means (k-means clustering algorithm) clustering algorithm is used for clustering. As other implementation methods, on the basis of being able to cluster the peak shift degree, the implementer can use other existing technologies to cluster the peak shift degree, such as peak clustering algorithm, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, etc., and this application does not make any special restrictions.
[0090] Combined with the difference between the peak shift degrees in the largest cluster and the smallest cluster of the class with similar peak positions, and the fluctuation degree of the peak shift degrees in the largest cluster and the smallest cluster, the substance type judgment value of the class with similar peak positions is determined, and the expression is:
[0091] ; In the formula, GF represents the judgment value of the substance type of the class with similar peak positions; P and Q represent the largest cluster and the smallest cluster of the class with similar peak positions, respectively; , They respectively represent the degree of discreteness of the peak shift in the largest cluster and the smallest cluster of clusters with similar peak positions; max() and min() represent the maximum value function and the minimum value function respectively.
[0092] In this embodiment, the degree of discreteness of the peak offset degree is the standard deviation. As other implementation methods, on the basis of being able to measure the uneven distribution of the peak offset degree, the implementer may use other existing technologies to measure the uneven distribution of the peak offset degree, such as variance, coefficient of variation, etc., and this application does not impose any special restrictions.
[0093] It should be noted that when the strip data points in the peak position-similar class are formed by the same material on the strip surface, the fluctuation degree of all peak shift degrees corresponding to the peak position-similar class is small, that is, the difference between all peak shift degrees is small. Therefore, after the peak shift degrees are classified, the minimum value in the largest cluster The maximum value in the smallest cluster The difference should be smaller than the maximum fluctuation of the peak shift degree in the two clusters. Therefore, the smaller the material type judgment value is, the greater the possibility that the strip data points in the peak position similar category are formed by the same material.
[0094] When the material type judgment value of the peak-close class is less than or equal to 0, the strip steel data points in the peak-close class are formed by the same material, and the peak-close class is regarded as a material class; if the material type judgment value of the peak-close class is greater than 0, the strip steel data points in the peak-close class are formed by different materials. Since the peak shift degree between different materials is small, the two strip steel data points corresponding to each peak shift degree in cluster P are divided into one class, according to Figure 3 In the merging method shown, classes having at least one identical strip data point are merged into the same class, and each class obtained after the merging is recorded as a material class.
[0095] For example, a, b and c are strip data points of the same material, forming class A, and d, e and f are strip data points of the same material, forming class B. The peak shift degrees between the strip data points in class A will be clustered into cluster P, and the peak shift degrees between the strip data points in class B will also be clustered into cluster P. However, the peak shift degrees between the strip data points in class A and the strip data points in class B will be clustered into cluster Q. After clustering the peak shift degrees corresponding to classes A and B, the classes of strip data points corresponding to each peak shift degree in cluster P are ab, ac, bc, de, df and ef respectively. After merging the classes with the same strip data points, two material classes abc and def are obtained.
[0096] Step 6: Determine whether the steel strip data points in any substance class are formed by cleaning-influencing substances based on the similarity between all steel strip data points in any substance class and the spectral data of each substance at the same wavelength.
[0097] The average values of the spectral data of all the strip data points in any material class at the same wavelength are calculated, and the average values are arranged in order from small to large wavelength to form a material sequence to be determined.
[0098] The cosine similarity between the undetermined material sequence of any material class and the spectral sequences of various materials is calculated. When the material corresponding to the spectral sequence with the largest cosine similarity is a cleaning-affecting substance, it is determined that the strip data points in the any material class are formed by the cleaning-affecting substance.
[0099] Step 7: Determine the surface cleanliness of the steel strip based on the proportion of steel strip data points belonging to cleanliness-affecting substances in the spectral data image.
[0100] Based on the proportion of strip data points belonging to cleanliness-affecting substances in the spectral data image, the surface cleanliness of the strip is determined, and the expression is:
[0101] ; Wherein, B represents the surface cleanliness of the steel strip; M represents the number of steel strip data points in the spectral data image; N represents the number of steel strip data points belonging to cleanliness-affecting substances in the spectral data image.
[0102] Furthermore, a cleanliness threshold is set, and the quality of the strip is detected by comparing the surface cleanliness of the strip with the cleanliness threshold. The specific method is as follows:
[0103] When the surface cleanliness of the steel strip is greater than the cleanliness threshold, the quality of the steel strip is determined to be qualified; when the surface cleanliness of the steel strip is less than the cleanliness threshold, the quality of the steel strip is determined to be unqualified and the steel strip needs to be reprocessed.
[0104] In this embodiment, the cleanliness threshold value is 67%. The cleanliness threshold value is preset manually and can be set by the implementer according to actual conditions. This application does not impose any special restrictions.
[0105] Based on the same inventive concept as the above method, the embodiment of the present application also provides a strip surface cleanliness detection device, comprising:
[0106] A spectral data acquisition module is used to acquire spectral data images of the strip surface at multiple preset wavelengths, wherein one pixel is recorded as one strip data point, and spectral data of each substance at the multiple preset wavelengths are respectively acquired, wherein there are substances that affect cleaning;
[0107] A spectral data classification module is used to obtain the peak value of the spectral data corresponding to each strip steel data point at all wavelengths; based on the number of the peak values of each strip steel data point, all strip steel data points are divided into sets of each strip steel type;
[0108] Calculate the average level of wavelength difference between any two strip steel data points in each strip steel type set at all peak points with the same sequence number, and divide the strip steel data points in each strip steel type set into classes with similar peak positions based on the distribution of all the average levels corresponding to each strip steel type set;
[0109] Based on the peak difference at all the same sequence numbers between any two strip steel data points in any peak position similarity class and the degree of dispersion of the peak difference, determine the peak shift degree between any two strip steel data points;
[0110] Obtaining clusters of all the peak shift degrees corresponding to each class with similar peak positions, combining the difference between the peak shift degrees in different clusters of each class with similar peak positions and the fluctuation degree of the peak shift degree in each cluster, dividing all the strip steel data points in each class with similar peak positions into various material classes;
[0111] A cleanliness detection module, for determining whether a strip data point in any substance class is formed by a cleanliness-affecting substance based on similarities between all strip data points in any substance class and spectral data of each substance at the same wavelength;
[0112] The surface cleanliness of the steel strip is determined based on the proportion of steel strip data points belonging to cleanliness-affecting substances in the spectral data image.
[0113] Based on the same inventive concept as the above method, an embodiment of the present application also provides a strip surface cleanliness detection system, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, the steps of any one of the above-mentioned strip surface cleanliness detection methods are implemented.
[0114] To summarize, the present application performs a first classification of the spectral data points of the steel strip by analyzing the number of peaks in the spectral data of the steel strip data points. Secondly, the first classification result is secondary classified by using the difference in peak positions. Thirdly, the secondary classification result is classified in combination with the difference between the peaks to obtain each material class. The steel strip data points in each material class are formed by the same material. Further, based on the average level of the spectral data of the steel strip data points in each material class and the similarity of the spectral data of each material, it is judged whether the steel strip data points in each material class are formed by cleaning-affecting materials, thereby eliminating the influence of the spectral line deviation caused by the vibration of the conveyor belt on the material identification results, thereby improving the accuracy of detecting the cleanliness of the steel strip surface.
[0115] The flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and the block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in a different order from the order disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.
[0116] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the basic features of the present application. Therefore, no matter from which point of view, the above embodiments of the present application should be regarded as exemplary and non-restrictive.
Claims
1. A method for detecting the cleanliness of a steel strip surface, characterized in that: The method comprises the following steps: Collecting spectral data images of the strip steel surface at multiple preset wavelengths, wherein the spectral absorption rate data of the strip steel surface within the wavelength range of 200-1000nm is collected by a multi-spectral imager and recorded as spectral data, one pixel is recorded as one strip steel data point, and the spectral data of each substance at the multiple preset wavelengths are collected respectively, wherein there are substances affecting cleaning, and the substances include rolling oil, carbon element, emulsion, iron powder and strip steel, wherein rolling oil, carbon element, emulsion and iron powder are substances affecting cleaning; Obtaining the peak value of the spectral data corresponding to each steel strip data point at all wavelengths; and dividing all steel strip data points into sets of steel strip types based on the number of the peak values of each steel strip data point; Calculate the average level of wavelength difference between any two strip steel data points in each strip steel type set at all peak points with the same sequence number, and divide the strip steel data points in each strip steel type set into classes with similar peak positions based on the distribution of all the average levels corresponding to each strip steel type set; Based on the peak difference at all the same sequence numbers between any two strip steel data points in any peak position similarity class and the degree of dispersion of the peak difference, determine the peak shift degree between any two strip steel data points; Obtaining clusters of all the peak shift degrees corresponding to each class with similar peak positions, combining the difference between the peak shift degrees in different clusters of each class with similar peak positions and the fluctuation degree of the peak shift degree in each cluster, dividing all the strip steel data points in each class with similar peak positions into various material classes; Determining whether the steel strip data points in any substance class are formed by a cleaning-influencing substance based on similarities between all steel strip data points in any substance class and the spectral data of each substance at the same wavelength; The surface cleanliness of the steel strip is determined based on the proportion of steel strip data points belonging to cleanliness-affecting substances in the spectral data image.
2. The method for detecting the surface cleanliness of a steel strip according to claim 1, characterized in that: The method for dividing all the strip steel data points into sets of different strip steel types is as follows: dividing all the strip steel data points with the same number of peak values into the same strip steel type set.
3. The method for detecting the surface cleanliness of a steel strip according to claim 1, characterized in that: The process of dividing the strip steel data points in each strip steel type set into classes with similar peak positions is as follows: Arrange the wavelengths of all peak points of each steel strip data point in ascending order to form a peak position sequence of each steel strip spectrum; The average value of the differences of all elements at the same position in the strip steel spectrum peak position sequence of the arbitrary two strip steel data points is used as the strip steel vibration offset value between the arbitrary two strip steel data points; Perform threshold segmentation on all the strip steel vibration offset values corresponding to any strip steel type set, and form a co-located strip steel vibration offset set with all the strip steel vibration offset values in any strip steel type set that are less than or equal to the segmentation threshold; The two strip data points corresponding to each strip vibration offset value in any iso-position strip vibration offset set are taken as one class, the classes with at least one identical strip data point are merged, and the classes obtained after the merger are taken as classes with similar peak positions.
4. The method for detecting the surface cleanliness of a steel strip according to claim 1, characterized in that: The process of determining the peak shift degree is as follows: Arrange all the peak values of each steel strip data point in ascending order of wavelength to form a peak value sequence of each steel strip spectrum; The differences between all elements at the same position in the strip steel spectrum peak sequence of any two strip steel data points are used to form a strip steel peak deviation sequence according to the element positions; The expression of the peak deviation degree between any two strip data points is: ; In the formula, Indicates the peak deviation degree between the xth and yth strip data points in any peak position similar class; , Respectively represent the strip steel spectrum peak sequences of the xth and yth strip steel data points in any of the similar peak position classes; represents the strip peak deviation sequence between the xth and yth strip data points in any peak position similar class; Cos() represents the cosine similarity function; MAD() represents the degree of dispersion function; Indicates a preset value greater than 0.
5. The method for detecting the surface cleanliness of a steel strip according to claim 1, characterized in that: The process of dividing all the strip data points in each class with similar peak positions into each material class is as follows: A clustering algorithm is used to obtain clusters of all the peak shift degrees corresponding to any class with similar peak positions, and a cluster with the largest mean value of the peak shift degree and a cluster with the smallest mean value are obtained, which are recorded as the largest cluster and the smallest cluster respectively; The expression for the determination value of the substance type of any peak position similar class is: ; Wherein, GF represents the substance type determination value of any peak position similar class; P and Q represent the maximum cluster and the minimum cluster of any peak position similar class respectively; , Respectively represent the discrete degree of the peak shift degree in the largest cluster and the smallest cluster of any peak position similar class; max() and min() represent the maximum value function and the minimum value function respectively; When the material type judgment value of any of the classes with similar peak positions is less than or equal to 0, any of the classes with similar peak positions is regarded as a material class; otherwise, the two strip steel data points corresponding to the peak deviation degrees in the largest cluster are divided into one class, and the classes with at least one identical strip steel data point are merged, and each class obtained after the merger is recorded as a material class.
6. The method for detecting the surface cleanliness of a steel strip according to claim 1, characterized in that: The process of determining whether any of the substance classes corresponds to a cleaning-affecting substance is as follows: Calculate the average values of the spectral data of all the strip data points in any material class at the same wavelength, and arrange the average values in ascending order of wavelength to form a material sequence to be determined; Arrange the spectrum data of each substance in ascending order of wavelength to form each spectrum sequence; The similarity between the undetermined material sequence of any material class and the spectral sequences of various materials is calculated. When the material corresponding to the spectral sequence with the greatest similarity is a cleaning-affecting substance, it is determined that the strip data points in any material class are formed by the cleaning-affecting substance.
7. The method for detecting the surface cleanliness of a steel strip according to claim 1, characterized in that: The expression of surface cleanliness is: ; In the formula, B represents the surface cleanliness of the strip; M represents the number of steel strip data points in the spectral data image; N represents the number of steel strip data points belonging to cleaning-affecting substances in the spectral data image.
8. The method for detecting the surface cleanliness of a steel strip according to claim 7, characterized in that: When the surface cleanliness of the steel strip is greater than a preset cleanliness threshold, the quality of the steel strip is determined to be qualified; otherwise, the quality of the steel strip is determined to be unqualified.
9. Strip steel surface cleanliness detection device, characterized in that: The device comprises: A spectral data acquisition module is used to collect spectral data images of the strip steel surface at multiple preset wavelengths, wherein the spectral absorption rate data of the strip steel surface within the wavelength range of 200-1000nm is collected by a multi-spectral imager and recorded as spectral data, one pixel is recorded as one strip steel data point, and the spectral data of each substance at the multiple preset wavelengths is collected respectively, wherein there are substances affecting cleaning, and the substances include rolling oil, carbon element, emulsion, iron powder and strip steel, wherein rolling oil, carbon element, emulsion and iron powder are substances affecting cleaning; A spectral data classification module is used to obtain the peak value of the spectral data corresponding to each strip steel data point at all wavelengths; based on the number of the peak values of each strip steel data point, all strip steel data points are divided into sets of each strip steel type; Calculate the average level of wavelength difference between any two strip steel data points in each strip steel type set at all peak points with the same sequence number, and divide the strip steel data points in each strip steel type set into classes with similar peak positions based on the distribution of all the average levels corresponding to each strip steel type set; Based on the peak difference at all the same sequence numbers between any two strip steel data points in any peak position similarity class and the degree of dispersion of the peak difference, determine the peak shift degree between any two strip steel data points; Obtaining clusters of all the peak shift degrees corresponding to each class with similar peak positions, combining the difference between the peak shift degrees in different clusters of each class with similar peak positions and the fluctuation degree of the peak shift degree in each cluster, dividing all the strip steel data points in each class with similar peak positions into various material classes; A cleanliness detection module, for determining whether a strip data point in any substance class is formed by a cleanliness-affecting substance based on similarities between all strip data points in any substance class and spectral data of each substance at the same wavelength; The surface cleanliness of the steel strip is determined based on the proportion of steel strip data points belonging to cleanliness-affecting substances in the spectral data image.
10. A strip steel surface cleanliness detection system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the strip surface cleanliness detection method as described in any one of claims 1-8 are implemented.
Citation Information
Patent Citations
Classifying image features
CN101151623A
Hyperspectral image classification method based on convolutional network and cyclic neural network
CN108460342A
Vision and analog sensing scrap sorting system and method
CN112243392A
Strip steel surface defect detection method and system based on image processing
CN115311292A
Method and device for detecting a microscope object in image data
CN115439398A
