A visual detection method for the flattening effect of an ultra-high strength steel plate for a new energy vehicle

Abstract

The application provides a new energy automobile ultra-high strength steel plate flattening effect visual detection method, relates to the steel plate flattening detection technical field, and includes the following steps: real-time acquisition of the first target image of a plurality of to-be-detected areas on the surface of the to-be-detected steel plate; selecting a to-be-detected feature subset corresponding to each to-be-detected area from a feature set, including a plurality of feature types, the feature types including local curvature, area flatness, edge straightness and micro-defects, and the to-be-detected feature subset including at least one feature type in the feature set; extracting feature indicators of the feature types in the corresponding to-be-detected feature subset from the first target image of each to-be-detected area; obtaining a flattening effect detection result according to the feature indicators of each to-be-detected area and feeding back to a flattening machine control system. The application realizes the breakthrough of the detection accuracy and reliability of the ultra-high strength steel plate flattening effect by depositing the process experience into a special feature library and a dynamic feature selection mechanism, and significantly improves the yield.

Description

Technical Field

[0001] This application relates to the field of steel plate leveling and testing technology, specifically to a visual inspection method for the leveling effect of ultra-high strength steel plates used in new energy vehicles. Background Technology

[0002] Currently, the leveling and inspection of steel plates often adopts general machine vision inspection. For example, the invention patent with application number 202410405767.6 proposes an adaptive electromagnetic heating leveling device and method, which discloses a machine vision inspection flatness module to detect the flatness of the plane to be leveled.

[0003] Due to the high hardness and complex springback characteristics of ultra-high strength steel plates used in new energy vehicles, the quality of their leveling process cannot be judged solely by inspecting surface flatness. Therefore, although this traditional inspection method can objectively and quickly identify defects on the steel plate surface, it lacks a dedicated feature system for the leveling effect of ultra-high strength steel plates used in new energy vehicles. Consequently, it cannot reflect the core quality parameters of the leveling process and makes it difficult to accurately quantify the leveling effect of ultra-high strength steel plates. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in the prior art, this application aims to provide a visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles, so as to accurately quantify the leveling effect of ultra-high strength steel plates;

[0005] The method includes the following steps:

[0006] An industrial camera installed at the exit of the leveling machine is used to acquire first target images of multiple areas to be inspected on the surface of the steel plate under test in real time.

[0007] Select a subset of features to be tested corresponding to each region to be detected from the full feature set. The full feature set includes multiple feature types, including local curvature, region flatness, edge straightness, and micro-defects. The subset of features to be tested includes at least one of the feature types from the full feature set.

[0008] From the first target image of each of the regions to be detected, feature indices of the feature types in the corresponding subset of features to be tested are extracted; the feature index of local curvature is the amount of local curvature change, the feature index of regional flatness is the regional flatness deviation, the feature index of edge straightness is the edge straightness offset, and the feature index of micro-defects is the micro-defect distribution density.

[0009] Based on the characteristic indicators of each of the areas to be tested, the detection results of the straightening effect are obtained and fed back to the straightening machine control system.

[0010] According to the technical solution provided in this application, the step of selecting a subset of features to be detected corresponding to each region to be detected from the full feature set includes the following steps:

[0011] Obtain the real-time quality fluctuation index of the current production line, wherein the real-time quality fluctuation index is proportional to the number of defective areas in the nearest N steel plates;

[0012] If the real-time quality fluctuation index is greater than or equal to the first preset threshold, then the entire set of features is taken as the subset of features to be tested.

[0013] According to the technical solution provided in this application, after obtaining the real-time quality fluctuation index of the current production line, the method further includes the following steps:

[0014] If the real-time quality fluctuation index is less than the first preset threshold, the testing requirements of the steel plate to be tested are obtained.

[0015] Based on the detection requirements, a corresponding selection strategy is obtained, which includes selection based on location and selection based on function.

[0016] A corresponding selection strategy is adopted to select a subset of the features to be detected for each region to be detected from the full feature set.

[0017] The location-based selection is based on the position type of the area to be detected on the surface of the steel plate to be tested, and the function-based selection is based on the functional area to which the area to be detected belongs on the surface of the steel plate to be tested.

[0018] According to the technical solution provided in this application, the location types include edge area, center area, weld-affected area, and hole periphery area;

[0019] The selection based on the location type of the area to be detected on the surface of the steel plate to be tested includes the following steps:

[0020] If the location type of the area to be detected on the surface of the steel plate to be tested is an edge area, then the local curvature, area flatness, edge straightness and micro-defects are selected to form the corresponding subset of features to be tested;

[0021] If the location type of the area to be detected on the surface of the steel plate to be tested is the central area, then the local curvature, regional flatness and micro-defects are selected to form the corresponding subset of features to be tested;

[0022] If the location type of the area to be detected on the surface of the steel plate to be tested is the weld influence zone, then the local curvature, regional flatness and micro-defects are selected to form the corresponding subset of features to be tested;

[0023] If the area to be detected is located in the periphery of a hole on the surface of the steel plate to be tested, then the local curvature and micro-defects are selected to form the corresponding subset of features to be tested.

[0024] According to the technical solution provided in this application, the functional area includes a door frame reinforcement area, a battery pack mounting surface, and an A-pillar area;

[0025] The selection of the functional area on the surface of the steel plate to be tested based on the area to be tested includes the following steps:

[0026] Retrieve the feature requirement mapping table, and extract the corresponding requirement weight sequence from the feature requirement mapping table according to the functional area to which the region to be detected belongs, and use it as the target weight sequence; wherein, the feature requirement mapping table includes the requirement weight sequence corresponding to the functional area, and the requirement weight sequence includes the requirement weight corresponding to each feature type in the feature set.

[0027] Select the minimum feature combination whose cumulative required weight is greater than or equal to a second preset threshold from the target weight sequence of each region to be detected, and use the minimum feature combination as the subset of features to be tested corresponding to the region to be detected.

[0028] According to the technical solution provided in this application, after selecting the minimum feature combination with a cumulative required weight greater than or equal to a second preset threshold from the target weight sequence of each of the regions to be detected, the method further includes the following steps:

[0029] Determine whether the minimum feature combination contains the mandatory detection feature of the functional area to which the region to be detected belongs;

[0030] The step of using the minimum feature combination as the subset of features to be detected corresponding to the region to be detected includes the following steps:

[0031] If so, the minimum feature combination is taken as the subset of features to be tested corresponding to the region to be detected.

[0032] According to the technical solution provided in this application, the forced detection feature of the door frame reinforcement area is edge straightness, the forced detection feature of the battery pack mounting surface is micro-defects, and the forced detection feature of the A-pillar area is local curvature.

[0033] According to the technical solution provided in this application, after determining whether the minimum feature combination contains the mandatory detection feature of the functional area to which the region to be detected belongs, the method further includes the following steps:

[0034] If not, then the minimum feature combination and the forced detection features together form the subset of features to be tested corresponding to the region to be detected.

[0035] According to the technical solution provided in this application, the step of extracting feature indicators of the corresponding feature type from the first target image of each of the detection regions includes the following steps:

[0036] Based on the feature type in the subset of features to be tested, the corresponding feature extraction algorithm module is dynamically loaded to calculate the feature index of the feature type;

[0037] If the subset of features to be tested contains two or more feature types, the calculations are performed in ascending order of feature computation complexity.

[0038] According to the technical solution provided in this application, obtaining the detection result of the straightening effect based on the characteristic indicators of each of the regions to be detected includes the following steps:

[0039] Based on the feature indices of the feature types included in the subset of features to be tested corresponding to the region to be detected, a defect probability sequence of the region to be detected is calculated, wherein the defect probability sequence includes the defect probability of each feature type.

[0040] Based on the defect probability sequence of the area to be detected, the corresponding demand weight sequence in the feature demand mapping table is retrieved to obtain the comprehensive defect index of the area to be detected.

[0041] If there is a test area with a comprehensive defect index greater than or equal to the third preset threshold, the test result is unqualified.

[0042] If the comprehensive defect index of all areas to be tested is less than the fourth preset threshold, the test result is qualified; the fourth preset threshold is less than the third preset threshold.

[0043] Compared with existing technologies, the advantages of this invention are as follows: Based on a deep understanding of the material deformation laws during the leveling of ultra-high strength steel plates, this invention constructs a complete set of dedicated features (local curvature change, regional flatness deviation, edge straightness offset, and micro-defect distribution density) to accurately quantify core process quality indicators. After obtaining the complete set of dedicated features, a rule set can be formulated by combining historical process data (such as setting a curvature change threshold and a micro-defect density tolerance standard), transforming implicit knowledge into executable detection logic to ensure that the evaluation criteria meet actual production needs. Moreover, through a dynamic feature subset selection mechanism (such as matching key features for different detection areas), detection efficiency and targeting are improved, redundant calculations are avoided, and mature image processing technologies in the industrial field (such as edge detection and region measurement) are used to extract preset features, eliminating the need to develop cutting-edge algorithms and lowering the technical implementation threshold. In summary, this invention achieves a breakthrough in the detection accuracy and reliability of ultra-high strength steel plate leveling effects by accumulating process experience into a dedicated feature library and a dynamic feature selection mechanism, significantly improving the yield rate. Attached Figure Description

[0044] Figure 1 A flowchart illustrating the steps of the visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles provided in this application. Detailed Implementation

[0045] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0046] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0047] Example 1

[0048] As mentioned in the background section, in view of the problems in the prior art, this application proposes a visual inspection method for the leveling effect of ultra-high strength steel plates used in new energy vehicles, such as... Figure 1 As shown, it includes the following steps:

[0049] S1. The first target image of multiple areas to be inspected on the surface of the steel plate to be tested is acquired in real time by an industrial camera installed at the exit of the leveling machine.

[0050] S2. Select a subset of features to be tested corresponding to each region to be detected from the full set of features. The full set of features includes multiple feature types, including local curvature, region flatness, edge straightness and micro-defects. The subset of features to be tested includes at least one of the feature types in the full set of features.

[0051] S3. Extract feature indices of the feature types in the corresponding feature subsets to be tested from the first target images of each of the regions to be tested; the feature index of local curvature is the amount of local curvature change, the feature index of regional flatness is the regional flatness deviation, the feature index of edge straightness is the edge straightness offset, and the feature index of micro-defects is the micro-defect distribution density.

[0052] Further, the step of extracting feature indicators of the corresponding feature type from the first target image of each of the regions to be detected includes the following steps:

[0053] Based on the feature type in the subset of features to be tested, the corresponding feature extraction algorithm module is dynamically loaded to calculate the feature index of the feature type;

[0054] If the subset of features to be tested contains two or more feature types, the calculations are performed in ascending order of feature computation complexity.

[0055] Specifically, the feature computation complexity is ordered in ascending order: region flatness (lowest) → edge straightness → micro-defects → local curvature (highest); the feature extraction algorithm module corresponding to local curvature executes the improved curvature chain code method, specifically:

[0056] a) Extract point cloud data from the steel plate surface and calculate the angle θ between the normal vectors of adjacent points;

[0057] b) Multiply θ by the strength compensation coefficient α to obtain the curvature change, where α = 0.02 × (tensile strength of steel plate / 1000).

[0058] c) Use the standard deviation of the curvature changes at all adjacent points as the local curvature change;

[0059] The feature extraction algorithm module corresponding to the region flatness executes the partition projection difference method, specifically:

[0060] a) Divide the region to be detected into an m×n grid (m, n≥3);

[0061] b) Calculate the elevation value of each grid center point relative to the reference plane;

[0062] c) Use the standard deviation of all grid elevation values ​​as the area flatness deviation;

[0063] The feature extraction algorithm module corresponding to edge straightness executes the breakpoint compensation algorithm, specifically:

[0064] a) Extract the pixel set of the steel plate edge contour;

[0065] b) Connect the broken contour segments using linear interpolation;

[0066] c) Calculate the maximum offset distance between the connected contour point set and the ideal straight line as the edge straightness offset;

[0067] The feature extraction algorithm module for micro-defects executes a multi-scale morphological fusion detection method, specifically:

[0068] a) Perform morphological opening operations on the image using circular structural elements with sizes of 3px, 5px, and 7px;

[0069] b) Generate a binary image of the defect by fusing the results of the three operations;

[0070] c) Calculate the percentage of defective pixels per unit area as the micro-defect distribution density;

[0071] S4. Based on the characteristic indicators of each of the areas to be detected, obtain the detection results of the leveling effect and feed them back to the leveling machine control system.

[0072] Further, obtaining the detection result of the straightening effect based on the characteristic indicators of each of the regions to be detected includes the following steps:

[0073] Based on the feature indices of the feature types included in the subset of features to be tested corresponding to the region to be detected, a defect probability sequence of the region to be detected is calculated, wherein the defect probability sequence includes the defect probability of each feature type.

[0074] Specifically, for each feature index, the defect probability is calculated. ,in, p i This represents the defect probability of feature type i. s i The feature index representing feature type i (e.g., i represents edge straightness, s) i (Indicates edge straightness offset, unit: mm); t represents the dynamic defect threshold. Where t0 represents the baseline threshold, determined by the functional area type, such as t0=0.2mm for the battery pack mounting surface; h represents the actual thickness of the steel plate; and k represents the sensitivity adjustment parameter. D represents the strength grade of the steel plate (strength grade is 100MPa), D = tensile strength / 100.

[0075] Based on the defect probability sequence of the area to be detected, the corresponding demand weight sequence in the feature demand mapping table is retrieved to obtain the comprehensive defect index of the area to be detected.

[0076] Specifically, the corresponding demand weight sequence in the feature demand mapping table is retrieved to obtain the demand weight for each feature type, and then calculated using the formula... The comprehensive defect index of the area to be inspected is obtained, where P 区域 This represents the overall defect index of the area to be inspected. This represents the demand weight of feature type i in the corresponding functional area, and n represents the number of feature types. This indicates the risk amplification factor (β=0.9 for the battery pack mounting surface, β=0.7 for the A-pillar area, and β=0.5 for the door frame reinforcement area).

[0077] If there is a test area with a comprehensive defect index greater than or equal to the third preset threshold, the test result is unqualified.

[0078] If the comprehensive defect index of all areas to be tested is less than the fourth preset threshold, the test result is qualified; the fourth preset threshold is less than the third preset threshold.

[0079] Optionally, the third preset threshold is 0.95, and the fourth preset threshold is 0.7.

[0080] In a preferred embodiment, selecting a subset of features to be detected corresponding to each of the regions to be detected from the full feature set includes the following steps:

[0081] Obtain the real-time quality fluctuation index of the current production line, wherein the real-time quality fluctuation index is proportional to the number of defective areas in the nearest N steel plates;

[0082] Specifically, the real-time quality fluctuation index is an indicator that dynamically reflects the recent quality stability of the production line, and its calculation method is as follows: Where k represents the proportionality coefficient, and N is set to 50-100 pieces according to the production line cycle time. The execution process is as follows: the vision system records the inspection results of each steel plate and marks the defective areas (such as warping and dents). The system automatically counts the total number of defective areas of the most recent N steel plates, calculates the real-time quality fluctuation index according to the formula, and then uploads it to the MES (Manufacturing Execution System) dashboard via the Industrial Internet of Things.

[0083] If the real-time quality fluctuation index is greater than or equal to the first preset threshold, then the entire set of features is taken as the subset of features to be tested.

[0084] Specifically, the first preset threshold is a critical value (e.g., 0.15) set based on historical quality data, representing the upper limit of allowable quality fluctuation. Implementation steps: Retrieve the quality database of the past 1000 steel plates, calculate the standard deviation of the non-conforming rate, set "mean + 3 times standard deviation" as the threshold, and write it into the system configuration parameters. When the real-time quality fluctuation index is greater than or equal to the first preset threshold, the system enables "feature set" for all inspection areas of the steel plate.

[0085] In this implementation method, when the quality of the production line fluctuates greatly, full feature detection can cover potential defect patterns and avoid missed detections.

[0086] Furthermore, after obtaining the real-time quality fluctuation index of the current production line, the following steps are also included:

[0087] If the real-time quality fluctuation index is less than the first preset threshold, the testing requirements of the steel plate to be tested are obtained.

[0088] Based on the detection requirements, a corresponding selection strategy is obtained, which includes selection based on location and selection based on function.

[0089] A corresponding selection strategy is adopted to select a subset of the features to be detected for each region to be detected from the full feature set.

[0090] The location-based selection is based on the position type of the area to be detected on the surface of the steel plate to be tested, and the function-based selection is based on the functional area to which the area to be detected belongs on the surface of the steel plate to be tested.

[0091] Specifically, the testing requirements are based on the application scenario requirements of the steel plate. By linking the steel plate ID to MES order data, the customer's technical requirements document is retrieved, and document keywords (such as "collision safety" and "battery sealing") are parsed to generate emphasized target requirements. Selection is based on location: The steel plate CAD model is used to categorize location types (e.g., edge area, center area). Selection is also based on function: Functional areas are mapped from automotive component drawings (e.g., impact resistance requirements corresponding to door frame reinforcement areas). If the emphasized target requirement is "process adaptability" (e.g., stamping), selection is initiated based on location. If the emphasized target requirement is "service performance" (e.g., collision safety), selection is initiated based on function.

[0092] The dynamic strategy of this implementation can reduce the amount of computation for invalid features and achieve a quality-efficiency balance through demand grading.

[0093] Furthermore, the location types include edge areas, center areas, weld-affected areas, and hole periphery areas;

[0094] The selection based on the location type of the area to be detected on the surface of the steel plate to be tested includes the following steps:

[0095] If the location type of the area to be detected on the surface of the steel plate to be tested is an edge area, then the local curvature, area flatness, edge straightness and micro-defects are selected to form the corresponding subset of features to be tested;

[0096] If the location type of the area to be detected on the surface of the steel plate to be tested is the central area, then the local curvature, regional flatness and micro-defects are selected to form the corresponding subset of features to be tested;

[0097] If the location type of the area to be detected on the surface of the steel plate to be tested is the weld influence zone, then the local curvature, regional flatness and micro-defects are selected to form the corresponding subset of features to be tested;

[0098] If the area to be detected is located in the periphery of a hole on the surface of the steel plate to be tested, then the local curvature and micro-defects are selected to form the corresponding subset of features to be tested.

[0099] Specifically, the edge zone refers to a 10mm wide area around the outer edge of the steel plate, the center zone refers to the area within ±20% of the geometric center of the steel plate, the weld-affected zone refers to the 5mm extension of the weld heat-affected zone, and the hole periphery zone refers to the 8mm radius annular area around the positioning hole / bolt hole. The characteristic subset of the edge zone is local curvature + area flatness + edge straightness + micro-defects. Key considerations include: Edge straightness: During the leveling process, the edges of the steel plate are easily affected by uneven roller stress, resulting in "wavy edges," which directly affects the positioning accuracy of the stamping die. Local curvature: The edge area (width ≤ 10mm) is a stress release zone; sudden changes in curvature can lead to abnormal wear of the cutting tool. Micro-defects: When the steel plate coil is uncoiled, the edges are prone to scratches due to friction with the guide rollers. Feature subset of the central area: local curvature + regional flatness + micro-defects. Core considerations: Regional flatness: The central area is the visual master plane of the body panel. Flatness deviation > 0.5mm / m will cause optical distortion of the paint surface (orange peel effect). Local curvature: It is crucial for stamping. Curvature deviation (e.g. > design value ±10%) will cause uneven material flow and cracking (especially for ultra-high strength steel above 980MPa). In addition, since the central area is far from the boundary constraints, the edge straightness is meaningless, and there is no stress concentration caused by punching, so there is no need to inspect the features around the hole. The characteristic subset of the weld-affected zone (HAZ) consists of local curvature, regional flatness, and micro-defects. Key considerations include: increased hardness in the HAZ after laser welding, leading to "saddle-shaped" deformation (abnormal curvature) on both sides of the weld during straightening; and the release of thermal stress easily triggering micro-cracks (concentrated within ±5mm of the weld center). Curvature and flatness detection can capture saddle-shaped deformation, while micro-defect detection enhances the detection rate of protrusions or depressions. Edge straightness is excluded because the weld is usually located inside the steel plate and is unrelated to edge geometry. The characteristic subset of the hole periphery consists of local curvature and micro-defects. Key considerations include: work hardening of the material due to punching, increasing the yield strength of the periphery; and the tendency for springback warping (abnormal local curvature) at the hole edge during straightening. Micro-defects are considered to be punching burrs and micro-cracks, which are the main causes of failure in battery pack mounting bolt holes. The scientific basis for discarding flatness is: finite element analysis shows that flatness within 5mm of the hole periphery is dominated by the punching process and is difficult to improve with straightening. Engineering verification: Adding flatness inspection only improves yield by 0.7%, but the time consumption is significantly increased, making the cost-effectiveness too low.

[0100] This implementation method reduces the number of detection items and improves overall efficiency by using differentiated feature subsets while ensuring key quality.

[0101] In a preferred embodiment, the functional area includes a door frame reinforcement area, a battery pack mounting surface, and an A-pillar area;

[0102] The selection of the functional area on the surface of the steel plate to be tested based on the area to be tested includes the following steps:

[0103] Retrieve the feature requirement mapping table, and extract the corresponding requirement weight sequence from the feature requirement mapping table according to the functional area to which the region to be detected belongs, and use it as the target weight sequence; wherein, the feature requirement mapping table includes the requirement weight sequence corresponding to the functional area, and the requirement weight sequence includes the requirement weight corresponding to each feature type in the feature set.

[0104] Specifically, the feature requirement mapping table stores the weight requirements of each functional area for feature types. The construction process of the feature requirement mapping table is as follows: collect finite element analysis data, such as the curvature tolerance of column A in a collision being ≤0.03 / mm. Determine the weight values ​​through DOE (Design of Experiments) and write them into the system knowledge base. An example feature requirement mapping table is shown in Table 1:

[0105] Table 1

[0106]

[0107] Select the minimum feature combination whose cumulative required weight is greater than or equal to a second preset threshold from the target weight sequence of each region to be detected, and use the minimum feature combination as the subset of features to be tested corresponding to the region to be detected.

[0108] Specifically, the functional area of ​​the steel plate is identified by linking the BOM information to the QR code, the corresponding weight sequence is retrieved, and the features are arranged in descending order of weight. The features are accumulated starting from the high-weight features until the cumulative weight is ≥0.85 (the second preset threshold is selectable). For example, in the A-pillar area, {local curvature, micro-defects} is selected as the corresponding subset of features to be tested, and the cumulative weight is 0.95+0.7=1.65>0.85.

[0109] Furthermore, after selecting the minimum feature combination with a cumulative required weight greater than or equal to a second preset threshold from the target weight sequence of each of the regions to be detected, the method further includes the following steps:

[0110] Determine whether the minimum feature combination contains the mandatory detection feature of the functional area to which the region to be detected belongs;

[0111] Furthermore, the forced detection feature of the door frame reinforcement area is edge straightness, the forced detection feature of the battery pack mounting surface is micro-defects, and the forced detection feature of the A-pillar area is local curvature.

[0112] Specifically, the engineering basis for determining the mandatory inspection features is as follows: the door frame reinforcement area is the edge straightness, as door frame deformation > 0.5 mm / m will cause the sealing strip to fail. The battery pack mounting surface is a micro-defect, as scratches > 0.2 mm may puncture the battery pack insulation layer. The A-pillar area is a local curvature, as the curvature tolerance is ±0.03 mm; exceeding this tolerance will cause collision energy absorption failure.

[0113] The step of using the minimum feature combination as the subset of features to be detected corresponding to the region to be detected includes the following steps:

[0114] If so, the minimum feature combination is taken as the subset of features to be tested corresponding to the region to be detected.

[0115] Furthermore, after determining whether the minimum feature combination contains the mandatory detection feature of the functional area to which the region to be detected belongs, the method further includes the following steps:

[0116] If not, then the minimum feature combination and the forced detection features together form the subset of features to be tested corresponding to the region to be detected.

[0117] Specifically, mandatory detection features are features that must be detected in a specific functional area, determined by automotive safety standards or material failure analysis, and cannot be excluded by weighted optimization algorithms. The system reads the functional area identifier of the current detection area, retrieves the list of mandatory detection features for that functional area, and compares whether the subset of features to be selected contains mandatory features.

[0118] For example, in the detection of the battery pack mounting surface, a weighted subset is generated: {regional flatness (weight 0.9), local curvature (weight 0.7)}. The cumulative weight 1.6 > the threshold 0.85. Forced verification: the subset is missing micro-defects. Therefore, the final subset of features to be tested corresponding to the battery pack mounting surface = {flatness, local curvature, micro-defects}.

[0119] This implementation takes into account the pursuit of efficiency through weight accumulation, but security features may be eliminated due to non-high weights. Therefore, mandatory features are required items that are independent of weights.

[0120] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are only preferred embodiments of this application. It should be noted that due to the limitations of written expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. A visual inspection method for the leveling effect of ultra-high strength steel plates used in new energy vehicles, characterized in that, Includes the following steps: An industrial camera installed at the exit of the leveling machine is used to acquire first target images of multiple areas to be inspected on the surface of the steel plate under test in real time. Select a subset of features to be tested corresponding to each region to be detected from the full feature set. The full feature set includes multiple feature types, including local curvature, region flatness, edge straightness, and micro-defects. The subset of features to be tested includes at least one of the feature types from the full feature set. From the first target image of each of the regions to be detected, feature indices of the feature types in the corresponding subset of features to be tested are extracted; the feature index of local curvature is the amount of local curvature change, the feature index of regional flatness is the regional flatness deviation, the feature index of edge straightness is the edge straightness offset, and the feature index of micro-defects is the micro-defect distribution density. Based on the characteristic indicators of each of the areas to be tested, the detection results of the straightening effect are obtained and fed back to the straightening machine control system.

2. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 1, characterized in that: The step of selecting a subset of features to be detected corresponding to each region from the full feature set includes the following steps: Obtain the real-time quality fluctuation index of the current production line, wherein the real-time quality fluctuation index is proportional to the number of defective areas in the nearest N steel plates; If the real-time quality fluctuation index is greater than or equal to the first preset threshold, then the entire set of features is taken as the subset of features to be tested.

3. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 2, characterized in that: After obtaining the real-time quality fluctuation index of the current production line, the following steps are also included: If the real-time quality fluctuation index is less than the first preset threshold, the testing requirements of the steel plate to be tested are obtained. Based on the detection requirements, a corresponding selection strategy is obtained, which includes selection based on location and selection based on function. A corresponding selection strategy is adopted to select a subset of the features to be detected for each region to be detected from the full feature set. The location-based selection is based on the position type of the area to be detected on the surface of the steel plate to be tested, and the function-based selection is based on the functional area to which the area to be detected belongs on the surface of the steel plate to be tested.

4. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 3, characterized in that: The location types include edge area, center area, weld-affected area, and hole periphery area; The selection based on the location type of the area to be detected on the surface of the steel plate to be tested includes the following steps: If the location type of the area to be detected on the surface of the steel plate to be tested is an edge area, then the local curvature, area flatness, edge straightness and micro-defects are selected to form the corresponding subset of features to be tested; If the location type of the area to be detected on the surface of the steel plate to be tested is the central area, then the local curvature, regional flatness and micro-defects are selected to form the corresponding subset of features to be tested; If the location type of the area to be detected on the surface of the steel plate to be tested is the weld influence zone, then the local curvature, regional flatness and micro-defects are selected to form the corresponding subset of features to be tested; If the area to be detected is located in the periphery of a hole on the surface of the steel plate to be tested, then the local curvature and micro-defects are selected to form the corresponding subset of features to be tested.

5. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 3, characterized in that: The functional area includes the door frame reinforcement area, the battery pack mounting surface, and the A-pillar area; The selection of the functional area on the surface of the steel plate to be tested based on the area to be tested includes the following steps: Retrieve the feature requirement mapping table, and extract the corresponding requirement weight sequence from the feature requirement mapping table according to the functional area to which the region to be detected belongs, and use it as the target weight sequence; wherein, the feature requirement mapping table includes the requirement weight sequence corresponding to the functional area, and the requirement weight sequence includes the requirement weight corresponding to each feature type in the feature set. Select the minimum feature combination whose cumulative required weight is greater than or equal to a second preset threshold from the target weight sequence of each region to be detected, and use the minimum feature combination as the subset of features to be tested corresponding to the region to be detected.

6. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 5, characterized in that: After selecting the minimum feature combination with a cumulative required weight greater than or equal to a second preset threshold from the target weight sequence of each of the regions to be detected, the method further includes the following steps: Determine whether the minimum feature combination contains the mandatory detection feature of the functional area to which the region to be detected belongs; The step of using the minimum feature combination as the subset of features to be detected corresponding to the region to be detected includes the following steps: If so, the minimum feature combination is taken as the subset of features to be tested corresponding to the region to be detected.

7. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 6, characterized in that: The forced detection feature of the door frame reinforcement area is edge straightness, the forced detection feature of the battery pack mounting surface is micro-defects, and the forced detection feature of the A-pillar area is local curvature.

8. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 6, characterized in that: After determining whether the minimum feature combination contains the mandatory detection feature of the functional area to which the region to be detected belongs, the method further includes the following steps: If not, then the minimum feature combination and the forced detection features together form the subset of features to be tested corresponding to the region to be detected.

9. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 1, characterized in that: The step of extracting feature indices of the corresponding feature type from the first target image of each of the detection regions includes the following steps: Based on the feature type in the subset of features to be tested, the corresponding feature extraction algorithm module is dynamically loaded to calculate the feature index of the feature type; If the subset of features to be tested contains two or more feature types, the calculations are performed in ascending order of feature computation complexity.

10. The visual inspection method for the leveling effect of ultra-high strength steel plates for new energy vehicles according to claim 1, characterized in that: The step of obtaining the detection result of the straightening effect based on the characteristic indicators of each of the regions to be detected includes the following steps: Based on the feature indices of the feature types included in the subset of features to be tested corresponding to the region to be detected, a defect probability sequence of the region to be detected is calculated, wherein the defect probability sequence includes the defect probability of each feature type. Based on the defect probability sequence of the area to be detected, the corresponding demand weight sequence in the feature demand mapping table is retrieved to obtain the comprehensive defect index of the area to be detected. If there is a test area with a comprehensive defect index greater than or equal to the third preset threshold, the test result is unqualified. If the comprehensive defect index of all areas to be tested is less than the fourth preset threshold, the test result is qualified; the fourth preset threshold is less than the third preset threshold.

Images