A component defect detection and marking system and method based on visual recognition

Abstract

The present invention discloses a component defect detection and marking system and method based on visual recognition, and relates to the technical field of defect detection and marking. The marking method comprises the following steps: constructing a component defect detection and management system to generate detection records; performing abnormal evaluation and defect type identification on any detection record; judging the rationality of the defect mark in the detection record and identifying characteristic behavior according to the abnormal evaluation situation and defect judgment situation presented by any detection record, and extracting characteristic factors in the characteristic behavior; correcting the abnormal detection record, and establishing a matching mechanism to match corresponding marks with different abnormal evaluation results; performing abnormal evaluation on the component whenever a defect detection is performed on the component; capturing the characteristic behavior and correcting the abnormal evaluation result, setting a corresponding mark for the situation where the component has defects and performing abnormal reminder.

Description

Technical Field

[0001] The present invention relates to the technical field of defect detection marking, and in particular to a component defect detection marking system and method based on visual recognition. Background Art

[0002] Parts defect detection is to identify and evaluate defects or unevenness in mechanical parts without damaging or affecting the performance of the object being tested, to ensure the quality and reliability of the parts, and to prevent failures or safety hazards during use. In the defect detection process, visual recognition is usually used to identify defects on the surface of parts, and then marking the defect location can facilitate later viewing and tracing.

[0003] In the process of marking the defect location, various marking types are usually used to distinguish the severity of the defect. However, since defect recognition is based on the collected image, the accuracy of defect recognition is easily affected by external factors, and the severity of the defect is highly subjective, which is difficult to unify among different staff members. This will lead to differences in the defect recognition of parts and components, resulting in incorrect marking. When the defects are subsequently repaired, the repair time will be delayed due to incorrect marking. Summary of the invention

[0004] The purpose of the present invention is to provide a component defect detection and marking system and method based on visual recognition to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solution: a component defect detection and marking method based on visual recognition, the marking method comprising the following steps:

[0006] Step S100: constructing a component defect detection management system to record the defect detection process and marking conditions of any component and generate corresponding detection records; based on the component images collected in any detection record and the recorded marking conditions, performing abnormality assessment and defect type identification on the detection record;

[0007] Step S200: according to the abnormal evaluation and defect judgment presented by any test record, the rationality of the defect mark in the test record is judged; the characteristic behavior of the abnormal test record with unreasonable judgment result is identified, and the characteristic factors in the characteristic behavior are extracted;

[0008] Step S300: by processing the characteristic factors in any abnormality detection record, correcting the characteristic behavior of the abnormality detection record, and re-evaluating the abnormality and determining the defect; based on the defect detection situation presented after the correction, establishing a matching mechanism to match the corresponding marks to different abnormality evaluation results;

[0009] Step S400: Whenever a component is subjected to defect detection, an abnormality assessment is performed on the component, relevant data is extracted for any characteristic factor during the defect detection process, and characteristic behavior of the component is captured; based on the captured characteristic behavior, the abnormality assessment result is corrected, and a corresponding mark is set for the situation where the component has defects and an abnormality reminder is issued.

[0010] Furthermore, step S100 includes the following steps:

[0011] Step S101: Whenever defect detection is performed on a component, an image of the component is collected by visual recognition technology, and the image of the component is transmitted to the component defect detection management system to generate a corresponding detection record; in the component defect detection management system, several dimensional evaluation rules are set for the component image to obtain the characteristic value of each dimension; the dimensions set for the component image include the size of the component, the number of defects, the size of the defects, the detection accuracy and environmental factors, etc.; by setting multiple dimensions, the defect situation of the component can be obtained more accurately;

[0012] Step S102: preset an expected feature range for each dimension of the component image, and set the expected feature range amplitude of the i-th dimension to ΔE i , according to the formula:

[0013] ;

[0014] Where j is a positive integer and j∈(1,m), m is the number of dimensions set, ΔE j is the expected feature range of the jth dimension; the influence ratio of the i-th dimension is calculated as α i ; Set the eigenvalue of the component in the i-th dimension to E i , according to the formula:

[0015] ;

[0016] Where k is a positive integer and k∈(1,m), α kis the influence proportion of the kth dimension; the accurate value Z of the detection record is calculated; the characteristic value of each dimension is obtained by comparing with the expected value, and in the final accuracy calculation process, the influence proportion of different dimensions is different, so it is necessary to consider the influence proportion. This problem can be well solved by analyzing the expected value range. The larger the expected range, the more errors are allowed, and the smaller the impact on the final accuracy;

[0017] Step S103: Setting an accurate threshold Z th , if Z<Z th , it is determined that the component has defects after inspection, and each abnormal dimension whose characteristic value is not within the expected characteristic range is extracted from the inspection record, and features are extracted for each abnormal dimension to obtain an abnormal feature set of the inspection record; if there is a defect mark in the inspection record, the mark type of the defect mark is obtained, and the abnormal feature set is matched with the mark type; firstly, the method of judging the abnormal situation and abnormal features of each inspection record is explained, which can directly lead to the subsequent process of detailed analysis of the defect situation.

[0018] Further, step S200 includes the following steps:

[0019] Step S201: setting the detection record with defect marks as abnormal detection record, obtaining the accurate value of any abnormal detection record and the mark type of the recorded defect mark, and obtaining the accurate value range corresponding to each mark type; if there are overlaps between the accurate value ranges of several mark types, the several mark types are all set as abnormal mark types; it is expected to reflect the degree of defects through the accurate value range, but the overlapping area can directly indicate that there are problems in the original accurate value analysis process, providing a basis for the subsequent optimization process;

[0020] Step S202: arbitrarily select an abnormal mark type, obtain the overlapping range between the abnormal mark type and the other abnormal mark types, and set one of the overlapping ranges to (Z1, Z2); select an abnormal detection record from each abnormal detection record of the abnormal mark type, wherein the accurate value Z∈(Z1, Z2) of the abnormal detection record is selected, and each abnormal dimension of the abnormal detection record is extracted;

[0021] Step S203: randomly select a reference detection record from the remaining abnormal detection records of the abnormal mark type, and the accurate value Z of the reference detection record ’∉(Z1, Z2), extract each abnormal dimension of the reference detection record; compare each abnormal dimension of the abnormal detection record with each abnormal dimension of the reference detection record to obtain a difference dimension set; by analyzing the abnormal detection records in the overlapping range, the cause of the overlap can be directly obtained, and the overlapping range can be removed by only checking each factor later;

[0022] Step S204: randomly select a difference dimension from the difference dimension set, and obtain the characteristic values ​​of the difference dimension in the two detection records respectively, according to the formula:

[0023] ;

[0024] Among them, α total is the sum of the impact proportions of all dimensions, and α total =α1+α2+α3+…+α m , m is the number of dimensions set, α w is the influence ratio of the wth difference dimension, E w is the characteristic value of the w-th difference dimension in the anomaly detection record, (E w ) ’ is the characteristic value of the w-th difference dimension in the reference detection record, ΔE w is the expected feature range of the w-th difference dimension; the difference amplitude η of the w-th difference dimension is calculated w If η w <1, the wth difference dimension is set as the feature dimension; the difference value recorded under the proportion presented by the difference dimension is compared with the feature difference value obtained by direct calculation. If the difference between the two is large, it directly reflects the degree of defect of the parts. If the difference is small, it reflects the accuracy range and marking situation, which is also the reason for the marking error. Therefore, it is necessary to check the difference features with small differences;

[0025] Step S205: extract and merge all feature dimensions in the anomaly detection record to obtain feature behaviors of the anomaly detection record, perform feature extraction on the data presented in the anomaly detection record by each feature dimension, and obtain a feature factor set of the anomaly detection record.

[0026] Further, step S300 includes the following steps:

[0027] Step S301: randomly selecting a characteristic factor from a characteristic factor set of any abnormality detection record, obtaining the value of the characteristic factor in the abnormality detection record, obtaining the average value of the characteristic factor from the remaining detection records, and obtaining the offset amplitude of the characteristic factor in the abnormality detection record;

[0028] Step S302: Obtain the characteristic value of the characteristic dimension where the characteristic factor is located as E1, and set the offset amplitude of the uth characteristic factor in the characteristic dimension as f u , according to the formula:

[0029] ;

[0030] Among them, E1 min is the minimum expected eigenvalue of the feature dimension, τ u is the correction coefficient of the uth characteristic factor, and x is the number of characteristic factors in the characteristic dimension; the characteristic factors contained in each detection record of the characteristic dimension are substituted into the formula to obtain the correction coefficient of each characteristic factor; the correction coefficient is the influence of the characteristic factor on the characteristic value of the characteristic dimension. By obtaining the correction coefficient of each characteristic factor, the accurate value of each detection record can be corrected, and the characteristic factor is the reason for the overlapping range, so correcting the characteristic factor is equivalent to correcting the overlapping range, which helps each mark type to have an independent value range;

[0031] Step S303: Set the correction coefficient of the vth characteristic factor in the characteristic factor set of the abnormal detection record to τ v , according to the formula:

[0032] ;

[0033] Wherein, Z is the exact value of the anomaly detection record, and x is the number of characteristic factors in the characteristic factor set; the exact value Z of the anomaly detection record after correction is calculated. ’ ;

[0034] Step S304: extract the abnormal dimension and feature dimension in any abnormal detection record, remove the feature dimension from the abnormal dimension, and obtain the actual abnormal dimension set of the abnormal detection record; compare the actual abnormal dimension sets in each abnormal detection record with each other, classify several abnormal detection records with the same actual abnormal dimension set into the same category, obtain the corrected accurate values ​​of the several abnormal detection records respectively, obtain an accurate value range of the actual abnormal dimension set and randomly assign a label type;

[0035] Step S305: After correcting the exact value of each anomaly detection record, obtain the exact value range of each tag type. If there is no overlap between the exact value ranges of each tag type, match each tag type with the corresponding exact value range. If there is an overlap, continue to extract the feature dimension and correct it again until there is no overlap between the tag types.

[0036] Furthermore, step S400 includes the following steps:

[0037] Step S401: collecting an image of a component to generate a corresponding real-time detection record, extracting feature values ​​of the real-time detection record in each dimension, and obtaining the accuracy of the real-time detection record;

[0038] Step S402: When the obtained accuracy is less than the set accuracy threshold, the abnormal dimension and feature dimension of the real-time detection record are extracted to obtain the actual abnormal dimension set of the real-time detection record; a number of feature factors under each feature dimension are obtained to correct the real-time detection record; the corrected accuracy is compared with the accuracy range of each marking type to determine the marking type of the real-time detection record, mark the certain component and send an abnormal reminder to the staff.

[0039] In order to better implement the above method, a component defect detection and marking system is also proposed, the marking system includes a historical detection analysis module, a defect feature recognition module, a defect marking correction module and a defect recognition and marking module;

[0040] The historical detection analysis module is used to build a component defect detection management system to record the defect detection process and marking conditions of any component and generate corresponding detection records; based on the component images collected and the recorded marking conditions in any detection record, the detection record is evaluated for abnormalities and defect type identification;

[0041] The defect feature recognition module is used to judge the rationality of the defect mark in any detection record according to the abnormal evaluation and defect judgment presented in the detection record; to identify the characteristic behavior of the abnormal detection record with unreasonable judgment results, and to extract the characteristic factors in the characteristic behavior;

[0042] The defect mark correction module is used to correct the characteristic behavior of any abnormal detection record by processing the characteristic factors in the abnormal detection record, and re-evaluate the abnormality and determine the defect; based on the defect detection situation presented after the correction, a matching mechanism is established to match the corresponding marks of different abnormal evaluation results;

[0043] The defect identification and marking module is used to perform abnormal evaluation on a certain component whenever a defect detection is performed on the component, extract relevant data for any characteristic factor during the defect detection process, and capture the characteristic behavior of the component; based on the captured characteristic behavior, correct the abnormal evaluation result, set a corresponding mark for the situation where the component has defects, and issue an abnormal reminder.

[0044] Further, the historical inspection analysis module includes an inspection record construction unit and a defect inspection evaluation unit;

[0045] The detection record construction unit is used to construct a component defect detection management system to record the defect detection process and marking status of any component and generate corresponding detection records; the defect detection evaluation unit is used to perform abnormality evaluation and defect type identification on the detection record based on the component image collected in any detection record and the recorded marking status.

[0046] Furthermore, the defect feature recognition module includes a defect mark determination unit and a feature factor extraction unit;

[0047] The defect marking judgment unit is used to judge the rationality of the defect marking in any detection record based on the abnormal evaluation situation and defect judgment situation presented by the detection record; the characteristic factor extraction unit is used to identify the characteristic behavior of the abnormal detection record with unreasonable judgment results, and extract the characteristic factors in the characteristic behavior.

[0048] Further, the defect marking correction module includes an abnormality assessment correction unit and a marking matching setting unit;

[0049] The anomaly assessment correction unit is used to correct the characteristic behavior of any anomaly detection record by processing the characteristic factors in the anomaly detection record, and re-perform anomaly assessment and defect judgment; the mark matching setting unit is used to establish a matching mechanism to match corresponding marks to different anomaly assessment results based on the defect detection situation presented after correction.

[0050] Further, the defect identification and marking module includes a characteristic behavior capture unit and a detection anomaly analysis unit;

[0051] The characteristic behavior capturing unit is used to perform abnormal evaluation on a certain component whenever defect detection is performed on the component, extract relevant data for any characteristic factor during the defect detection process, and capture the characteristic behavior of the component; the detection abnormality analyzing unit is used to correct the abnormal evaluation result based on the captured characteristic behavior, set a corresponding mark for the situation where the component has defects, and issue an abnormal reminder.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. When the present invention detects defects in parts, it can automatically identify the severity of the defects and mark them accordingly, so as to help the staff to accurately and quickly identify the defects and formulate relevant repair measures; at the same time, marking the defects also helps to facilitate later viewing and tracing;

[0054] 2. The present invention solves the problem of inconsistent identification of defects of different degrees by different personnel, and formulates an accurate and scientific matching method to help staff accurately grasp the defect situation according to the marking type in various situations, thereby improving the recognition efficiency and facilitating the subsequent repair work;

[0055] 3. The present invention optimizes the errors generated in visual recognition. Since it is difficult to have a clear dividing line between different defect degrees, it is easier to distinguish the defect degrees by eliminating the error factors with small influence, thereby improving the recognition and marking efficiency, and helping the staff to distinguish more easily, which is convenient for later identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic diagram of the steps of a component defect detection and marking method based on visual recognition;

[0057] Figure 2 This is a structural diagram of a component defect detection and marking system based on visual recognition. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] Example: Figure 1 to Figure 2 As shown, the present invention provides a component defect detection and marking method based on visual recognition, and the marking method comprises the following steps:

[0060] Step S100: constructing a component defect detection management system to record the defect detection process and marking conditions of any component and generate corresponding detection records; based on the component images collected in any detection record and the recorded marking conditions, performing abnormality assessment and defect type identification on the detection record;

[0061] Wherein, step S100 includes the following steps:

[0062] Step S101: Whenever defect detection is performed on a component, a component image is collected by visual recognition technology, and the component image is transmitted to a component defect detection management system to generate a corresponding detection record; in the component defect detection management system, evaluation rules of several dimensions are set for the component image to obtain feature values ​​of each dimension;

[0063] Step S102: preset an expected feature range for each dimension of the component image, and set the expected feature range amplitude of the i-th dimension to ΔE i , according to the formula:

[0064] ;

[0065] Where j is a positive integer and j∈(1,m), m is the number of dimensions set, ΔE j is the expected feature range of the jth dimension; the influence ratio of the i-th dimension is calculated as α i ; Set the eigenvalue of the component in the i-th dimension to E i , according to the formula:

[0066] ;

[0067] Where k is a positive integer and k∈(1,m), α k is the influence proportion of the kth dimension; the accurate value Z of the detection record is calculated;

[0068] Embodiment 1: Set four dimensions for the component image and the expected feature range amplitudes are 1, 2, 3, and 4 respectively, so the influence proportions of each dimension are 1 / 9=0.11, 2 / 8=0.25, 3 / 7=0.43, and 4 / 6=0.67 respectively, and the sum of the four influence proportions is 1.46; set the eigenvalues ​​under each dimension to 20, 20, 20, and 20 respectively; then obtain the accurate value Z=(0.11 / 1.46×20)+(0.25 / 1.46×20)+(0.43 / 1.46×20)+(0.67 / 1.46×20)=1.5+3.42+5.89+9.18=19.99;

[0069] Step S103: Setting an accurate threshold Z th , if Z<Z th , it is determined that the component has defects after inspection, and each abnormal dimension whose feature value is not within the expected feature range is extracted from the inspection record, and features are extracted for each abnormal dimension to obtain an abnormal feature set of the inspection record; if there is a defect mark in the inspection record, the mark type of the defect mark is obtained, and the abnormal feature set is matched with the mark type.

[0070] Step S200: according to the abnormal evaluation and defect judgment presented by any test record, the rationality of the defect mark in the test record is judged; the characteristic behavior of the abnormal test record with unreasonable judgment result is identified, and the characteristic factors in the characteristic behavior are extracted;

[0071] Wherein, step S200 includes the following steps:

[0072] Step S201: setting the detection record with defect marks as abnormal detection record, obtaining the accurate value of any abnormal detection record and the mark type of the recorded defect mark, and obtaining the accurate value range corresponding to each mark type; if there are several mark types whose accurate value ranges overlap with each other, then all of the several mark types are set as abnormal mark types;

[0073] Step S202: arbitrarily select an abnormal mark type, obtain the overlapping range between the abnormal mark type and the other abnormal mark types, and set one of the overlapping ranges to (Z1, Z2); select an abnormal detection record from each abnormal detection record of the abnormal mark type, wherein the accurate value Z∈(Z1, Z2) of the abnormal detection record is selected, and each abnormal dimension of the abnormal detection record is extracted;

[0074] Step S203: randomly select a reference detection record from the remaining abnormal detection records of the abnormal mark type, and the accurate value Z of the reference detection record ’ ∉(Z1, Z2), extracting each abnormal dimension of the reference detection record; comparing each abnormal dimension of the abnormal detection record with each abnormal dimension of the reference detection record to obtain a difference dimension set;

[0075] Step S204: randomly select a difference dimension from the difference dimension set, and obtain the characteristic values ​​of the difference dimension in the two detection records respectively, according to the formula:

[0076] ;

[0077] Among them, α total is the sum of the impact proportions of all dimensions, and α total =α1+α2+α3+…+α m , m is the number of dimensions set, α w is the influence ratio of the wth difference dimension, E w is the characteristic value of the w-th difference dimension in the anomaly detection record, (E w ) ’ is the characteristic value of the w-th difference dimension in the reference detection record, ΔE w is the expected feature range of the w-th difference dimension; the difference amplitude η of the w-th difference dimension is calculated w If η w <1, the wth difference dimension is set as the feature dimension;

[0078] Example 2: Set the influence ratio of a difference dimension to 0.2, and the sum of the influence ratios of all dimensions to 1, the accurate value of the anomaly detection record is 20, the accurate value of the reference detection record is 26, the characteristic value of the difference dimension in the anomaly detection record is 18, the accurate value in the reference detection record is 19, and the expected characteristic range amplitude of the difference dimension is 5, and the difference amplitude is (0.2×6-1) / 5=0.04. Since 0.04<1, it is set as the characteristic dimension;

[0079] Step S205: extract and merge all feature dimensions in the anomaly detection record to obtain feature behaviors of the anomaly detection record, perform feature extraction on the data presented in the anomaly detection record by each feature dimension, and obtain a feature factor set of the anomaly detection record.

[0080] Step S300: by processing the characteristic factors in any abnormality detection record, correcting the characteristic behavior of the abnormality detection record, and re-evaluating the abnormality and determining the defect; based on the defect detection situation presented after the correction, establishing a matching mechanism to match the corresponding marks to different abnormality evaluation results;

[0081] Wherein, step S300 includes the following steps:

[0082] Step S301: randomly selecting a characteristic factor from a characteristic factor set of any abnormality detection record, obtaining the value of the characteristic factor in the abnormality detection record, obtaining the average value of the characteristic factor from the remaining detection records, and obtaining the offset amplitude of the characteristic factor in the abnormality detection record;

[0083] Step S302: Obtain the characteristic value of the characteristic dimension where the characteristic factor is located as E1, and set the offset amplitude of the uth characteristic factor in the characteristic dimension as f u , according to the formula:

[0084] ;

[0085] Among them, E1 min is the minimum expected eigenvalue of the feature dimension, τ u is the correction coefficient of the uth characteristic factor, and x is the number of characteristic factors in the characteristic dimension; the characteristic factors contained in each detection record of the characteristic dimension are substituted into the formula to obtain the correction coefficient of each characteristic factor;

[0086] Step S303: Set the correction coefficient of the vth characteristic factor in the characteristic factor set of the abnormal detection record to τ v , according to the formula:

[0087] ;

[0088] Wherein, Z is the exact value of the anomaly detection record, and x is the number of characteristic factors in the characteristic factor set; the exact value Z of the anomaly detection record after correction is calculated. ’ ;

[0089] Step S304: extract the abnormal dimension and feature dimension in any abnormal detection record, remove the feature dimension from the abnormal dimension, and obtain the actual abnormal dimension set of the abnormal detection record; compare the actual abnormal dimension sets in each abnormal detection record with each other, classify several abnormal detection records with the same actual abnormal dimension set into the same category, obtain the corrected accurate values ​​of the several abnormal detection records respectively, obtain an accurate value range of the actual abnormal dimension set and randomly assign a label type;

[0090] Step S305: After correcting the exact value of each anomaly detection record, obtain the exact value range of each tag type. If there is no overlap between the exact value ranges of each tag type, match each tag type with the corresponding exact value range. If there is an overlap, continue to extract the feature dimension and correct it again until there is no overlap between the tag types.

[0091] Step S400: Whenever a defect detection is performed on a certain component, an abnormality assessment is performed on the component, relevant data is extracted for any characteristic factor during the defect detection process, and characteristic behavior of the component is captured; based on the captured characteristic behavior, the abnormality assessment result is corrected, and a corresponding mark is set for the situation where the component has a defect and an abnormality reminder is issued;

[0092] Wherein, step S400 includes the following steps:

[0093] Step S401: collecting an image of a component to generate a corresponding real-time detection record, extracting feature values ​​of the real-time detection record in each dimension, and obtaining the accuracy of the real-time detection record;

[0094] Step S402: When the obtained accuracy is less than the set accuracy threshold, the abnormal dimension and feature dimension of the real-time detection record are extracted to obtain the actual abnormal dimension set of the real-time detection record; a number of feature factors under each feature dimension are obtained to correct the real-time detection record; the corrected accuracy is compared with the accuracy range of each marking type to determine the marking type of the real-time detection record, mark the certain component and send an abnormal reminder to the staff.

[0095] A component defect detection and marking system, the marking system includes a historical detection and analysis module, a defect feature recognition module, a defect marking correction module and a defect recognition and marking module;

[0096] The historical detection analysis module is used to build a component defect detection management system to record the defect detection process and marking conditions of any component and generate corresponding detection records; based on the component images collected and the recorded marking conditions in any detection record, the detection record is evaluated for abnormalities and defect type identification;

[0097] The defect feature recognition module is used to judge the rationality of the defect mark in any detection record according to the abnormal evaluation and defect judgment presented in the detection record; to identify the characteristic behavior of the abnormal detection record with unreasonable judgment results, and to extract the characteristic factors in the characteristic behavior;

[0098] The defect mark correction module is used to correct the characteristic behavior of any abnormal detection record by processing the characteristic factors in the abnormal detection record, and re-evaluate the abnormality and determine the defect; based on the defect detection situation presented after the correction, a matching mechanism is established to match the corresponding marks of different abnormal evaluation results;

[0099] The defect identification and marking module is used to perform abnormal evaluation on a certain component whenever a defect detection is performed on the component, extract relevant data for any characteristic factor during the defect detection process, and capture the characteristic behavior of the component; based on the captured characteristic behavior, correct the abnormal evaluation result, set a corresponding mark for the situation where the component has defects, and issue an abnormal reminder.

[0100] Among them, the historical detection analysis module includes a detection record construction unit and a defect detection evaluation unit;

[0101] The detection record construction unit is used to construct a component defect detection management system to record the defect detection process and marking status of any component and generate corresponding detection records; the defect detection evaluation unit is used to perform abnormality evaluation and defect type identification on the detection record based on the component image collected in any detection record and the recorded marking status.

[0102] Among them, the defect feature recognition module includes a defect mark judgment unit and a feature factor extraction unit;

[0103] The defect marking judgment unit is used to judge the rationality of the defect marking in any detection record based on the abnormal evaluation situation and defect judgment situation presented by the detection record; the characteristic factor extraction unit is used to identify the characteristic behavior of the abnormal detection record with unreasonable judgment results, and extract the characteristic factors in the characteristic behavior.

[0104] Among them, the defect mark correction module includes an abnormal evaluation correction unit and a mark matching setting unit;

[0105] The anomaly assessment correction unit is used to correct the characteristic behavior of any anomaly detection record by processing the characteristic factors in the anomaly detection record, and re-perform anomaly assessment and defect judgment; the mark matching setting unit is used to establish a matching mechanism to match corresponding marks to different anomaly assessment results based on the defect detection situation presented after correction.

[0106] Among them, the defect recognition and marking module includes a feature behavior capture unit and a detection anomaly analysis unit;

[0107] The characteristic behavior capturing unit is used to perform abnormal evaluation on a certain component whenever defect detection is performed on the component, extract relevant data for any characteristic factor during the defect detection process, and capture the characteristic behavior of the component; the detection abnormality analyzing unit is used to correct the abnormal evaluation result based on the captured characteristic behavior, set a corresponding mark for the situation where the component has defects, and issue an abnormal reminder.

[0108] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A component defect detection and marking method based on visual recognition, characterized in that: The marking method comprises the following steps: Step S100: constructing a component defect detection management system to record the defect detection process and marking conditions of any component and generate corresponding detection records; based on the component images collected in any detection record and the recorded marking conditions, performing abnormality assessment and defect type identification on the detection record; Step S200: according to the abnormal evaluation and defect judgment presented by any test record, the rationality of the defect mark in the test record is judged; the characteristic behavior of the abnormal test record with unreasonable judgment result is identified, and the characteristic factors in the characteristic behavior are extracted; Step S300: by processing the characteristic factors in any abnormality detection record, correcting the characteristic behavior of the abnormality detection record, and re-evaluating the abnormality and determining the defect; based on the defect detection situation presented after the correction, establishing a matching mechanism to match the corresponding marks to different abnormality evaluation results; Step S400: Whenever a defect detection is performed on a certain component, an abnormality assessment is performed on the component, relevant data is extracted for any characteristic factor during the defect detection process, and characteristic behavior of the component is captured; based on the captured characteristic behavior, the abnormality assessment result is corrected, and a corresponding mark is set for the situation where the component has a defect and an abnormality reminder is issued; The step S200 includes the following steps: Step S201: setting the detection record with defect marks as abnormal detection record, obtaining the accurate value of any abnormal detection record and the mark type of the recorded defect mark, and obtaining the accurate value range corresponding to each mark type; if there are several mark types whose accurate value ranges overlap with each other, then all of the several mark types are set as abnormal mark types; Step S202: arbitrarily select an abnormal mark type, obtain the overlapping range between the abnormal mark type and the other abnormal mark types, and set one of the overlapping ranges to (Z1, Z2); select an abnormal detection record from each abnormal detection record of the abnormal mark type, wherein the accurate value Z∈(Z1, Z2) of the abnormal detection record is selected, and each abnormal dimension of the abnormal detection record is extracted; Step S203: randomly select a reference detection record from the remaining abnormal detection records of the abnormal mark type, and the accurate value Z of the reference detection record ’ ∉(Z1, Z2), extracting each abnormal dimension of the reference detection record; comparing each abnormal dimension of the abnormal detection record with each abnormal dimension of the reference detection record to obtain a difference dimension set; Step S204: randomly select a difference dimension from the difference dimension set, and obtain the characteristic values ​​of the difference dimension in the two detection records respectively, according to the formula: ； Among them, α total is the sum of the impact proportions of all dimensions, and α total =α1+α2+α3+…+α m , m is the number of dimensions set, α w is the influence ratio of the wth difference dimension, E w is the characteristic value of the w-th difference dimension in the anomaly detection record, (E w ) ’ is the characteristic value of the w-th difference dimension in the reference detection record, ΔE w is the expected feature range of the w-th difference dimension; the difference amplitude η of the w-th difference dimension is calculated w If η w <1, the wth difference dimension is set as the feature dimension; Step S205: extract and merge all feature dimensions in the anomaly detection record to obtain feature behaviors of the anomaly detection record, perform feature extraction on data presented in the anomaly detection record by each feature dimension, and obtain a feature factor set of the anomaly detection record; The step S300 includes the following steps: Step S301: randomly selecting a characteristic factor from a characteristic factor set of any abnormality detection record, obtaining the value of the characteristic factor in the abnormality detection record, obtaining the average value of the characteristic factor from the remaining detection records, and obtaining the offset amplitude of the characteristic factor in the abnormality detection record; Step S302: Obtain the characteristic value of the characteristic dimension where the characteristic factor is located as E1, and set the offset amplitude of the uth characteristic factor in the characteristic dimension as f u , according to the formula: ； Among them, E1 min is the minimum expected eigenvalue of the feature dimension, τ u is the correction coefficient of the uth characteristic factor, and x is the number of characteristic factors in the characteristic dimension; the characteristic factors contained in each detection record of the characteristic dimension are substituted into the formula to obtain the correction coefficient of each characteristic factor; Step S303: Set the correction coefficient of the vth characteristic factor in the characteristic factor set of the abnormal detection record to τ v , according to the formula: ； Wherein, Z is the exact value of the anomaly detection record, and x is the number of characteristic factors in the characteristic factor set; the exact value Z of the anomaly detection record after correction is calculated. ’ ; Step S304: extract the abnormal dimension and feature dimension in any abnormal detection record, remove the feature dimension from the abnormal dimension, and obtain the actual abnormal dimension set of the abnormal detection record; compare the actual abnormal dimension sets in each abnormal detection record with each other, classify several abnormal detection records with the same actual abnormal dimension set into the same category, obtain the corrected accurate values ​​of the several abnormal detection records respectively, obtain an accurate value range of the actual abnormal dimension set and randomly assign a label type; Step S305: After correcting the exact value of each anomaly detection record, obtain the exact value range of each tag type. If there is no overlap between the exact value ranges of each tag type, match each tag type with the corresponding exact value range. If there is an overlap, continue to extract the feature dimension and correct it again until there is no overlap between the tag types.

2. A component defect detection and marking method based on visual recognition according to claim 1, characterized in that: The step S100 includes the following steps: Step S101: Whenever defect detection is performed on a component, a component image is collected by visual recognition technology, and the component image is transmitted to a component defect detection management system to generate a corresponding detection record; in the component defect detection management system, evaluation rules of several dimensions are set for the component image to obtain feature values ​​of each dimension; Step S102: preset an expected feature range for each dimension of the component image, and set the expected feature range amplitude of the i-th dimension to ΔE i , according to the formula: ； Where j is a positive integer and j∈(1,m), m is the number of dimensions set, ΔE j is the expected feature range of the jth dimension; the influence ratio of the i-th dimension is calculated as α i ; Set the eigenvalue of the component in the i-th dimension to E i , according to the formula: ； Where k is a positive integer and k∈(1,m), α k is the influence proportion of the kth dimension; the accurate value Z of the detection record is calculated; Step S103: Setting an accurate threshold Z th , if Z<Z th , it is determined that the component has defects after inspection, and each abnormal dimension whose feature value is not within the expected feature range is extracted from the inspection record, and features are extracted for each abnormal dimension to obtain an abnormal feature set of the inspection record; if there is a defect mark in the inspection record, the mark type of the defect mark is obtained, and the abnormal feature set is matched with the mark type.

3. A component defect detection and marking method based on visual recognition according to claim 2, characterized in that: The step S400 includes the following steps: Step S401: collecting an image of a component to generate a corresponding real-time detection record, extracting feature values ​​of the real-time detection record in each dimension, and obtaining the accuracy of the real-time detection record; Step S402: When the obtained accuracy is less than the set accuracy threshold, the abnormal dimension and feature dimension of the real-time detection record are extracted to obtain the actual abnormal dimension set of the real-time detection record; a number of feature factors under each feature dimension are obtained to correct the real-time detection record; the corrected accuracy is compared with the accuracy range of each marking type to determine the marking type of the real-time detection record, mark the certain component and send an abnormal reminder to the staff.

4. A component defect detection and marking system, used to execute a component defect detection and marking method based on visual recognition as claimed in any one of claims 1 to 3, characterized in that: The marking system includes a historical detection and analysis module, a defect feature recognition module, a defect marking correction module and a defect recognition and marking module; The historical detection analysis module is used to construct a component defect detection management system to record the defect detection process and marking conditions of any component and generate corresponding detection records; based on the component images collected in any detection record and the recorded marking conditions, the detection record is evaluated for abnormalities and defect types are identified; The defect feature recognition module is used to judge the rationality of the defect mark in the detection record according to the abnormal evaluation situation and defect judgment situation presented by any detection record; to identify the characteristic behavior of the abnormal detection record with unreasonable judgment results, and to extract the characteristic factors in the characteristic behavior; The defect mark correction module is used to correct the characteristic behavior of any abnormal detection record by processing the characteristic factors in the abnormal detection record, and re-evaluate the abnormality and determine the defect; based on the defect detection situation presented after the correction, establish a matching mechanism to match the corresponding marks of different abnormal evaluation results; The defect identification and marking module is used to perform abnormal evaluation on a certain component whenever a defect detection is performed on the component, extract relevant data during the defect detection process for any characteristic factor, and capture the characteristic behavior of the certain component; The abnormal evaluation result is corrected based on the captured characteristic behavior, and a corresponding mark is set for the situation where a certain component has defects and an abnormal reminder is issued.

5. A component defect detection and marking system according to claim 4, characterized in that: The historical detection analysis module includes a detection record construction unit and a defect detection evaluation unit; The detection record construction unit is used to construct a component defect detection management system to record the defect detection process and marking situation of any component and generate corresponding detection records; The defect detection and evaluation unit is used to perform abnormality evaluation and defect type identification on any detection record based on the component image collected and the recorded marking conditions in the detection record.

6. A component defect detection and marking system according to claim 4, characterized in that: The defect feature recognition module includes a defect mark judgment unit and a feature factor extraction unit; The defect marking judgment unit is used to judge the rationality of the defect marking in any detection record based on the abnormal evaluation situation and defect judgment situation presented by the detection record; the characteristic factor extraction unit is used to identify the characteristic behavior of the abnormal detection record with unreasonable judgment results, and extract the characteristic factors in the characteristic behavior.

7. A component defect detection and marking system according to claim 4, characterized in that: The defect mark correction module includes an abnormality assessment correction unit and a mark matching setting unit; The anomaly assessment correction unit is used to correct the characteristic behavior of any anomaly detection record by processing the characteristic factors in the anomaly detection record, and re-perform anomaly assessment and defect judgment; the mark matching setting unit is used to establish a matching mechanism to match corresponding marks to different anomaly assessment results based on the defect detection situation presented after correction.

8. A component defect detection and marking system according to claim 4, characterized in that: The defect identification and marking module includes a feature behavior capture unit and a detection anomaly analysis unit; The characteristic behavior capturing unit is used to perform abnormal evaluation on a certain component whenever defect detection is performed on the component, extract relevant data for any characteristic factor during the defect detection process, and capture the characteristic behavior of the component; the detection abnormality analyzing unit is used to correct the abnormal evaluation result based on the captured characteristic behavior, set a corresponding mark for the situation where the component has defects, and issue an abnormal reminder.

Images