Air conditioner plastic panel surface defect detection method and system

Abstract

The application relates to the technical field of detection, in particular to a detection method and system for surface defects of an air conditioner plastic panel, wherein the method comprises the following steps: acquiring each two-dimensional image of the air conditioner plastic panel to be detected collected under multiple visual angles and a three-dimensional design model of the air conditioner plastic panel to be detected; based on each two-dimensional image and the three-dimensional design model, respectively extracting geometric image features of each two-dimensional image and geometric model features of the three-dimensional design model in each two-dimensional image located at the same visual angle; comparing and analyzing each geometric image feature and the geometric model feature of each two-dimensional image located at the same visual angle to obtain each image feature difference value; comparing each image feature difference value with each preset image feature threshold value to obtain a detection result of surface defects of the air conditioner plastic panel. The application aims to solve the problem that the prior art cannot accurately detect structural defects in non-appearance functional areas.

Description

Technical Field

[0001] This invention relates to the field of detection technology, specifically to a method and system for detecting surface defects in air conditioner plastic panels. Background Technology

[0002] As a crucial component of air conditioning products, the plastic panel of an air conditioner includes non-appearance functional areas designed to securely connect with other air conditioner components, support internal components, or enhance its own structural strength. These non-appearance functional areas include snap-fit ​​arrays on the panel edges for quick engagement with other air conditioner housings, mounting positions on the inner side of the panel for fixing circuit boards, screw holes or positioning holes for mounting fan motors or air guides, and reinforcing ribs distributed to improve the overall deformation resistance of the panel. During production, non-appearance functional areas of the air conditioner plastic panel are prone to issues such as missing snap-fit ​​array materials, cavities in screw holes, or substandard dimensions.

[0003] Existing automated optical inspection systems rely primarily on acquiring single two-dimensional images for grayscale or color analysis when detecting surface defects in non-functional areas of air conditioner plastic panels. However, these non-functional areas have complex three-dimensional structures, and their locations, such as those inside the panel or deep cavities, are often concealed or poorly lit, obscuring the true surface condition. Furthermore, two-dimensional images acquired under normal lighting are prone to shadows, making it difficult to obtain effective image information. Consequently, the detection results for non-functional areas of air conditioner plastic panels are prone to misjudgment or missed detection, and the systems cannot accurately detect structural defects such as breakage, missing material, deformation, and dimensional deviations in these areas. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for detecting surface defects in air conditioner plastic panels, which solves the problem that existing technologies for detecting non-appearance functional areas of air conditioner plastic panels may result in misjudgments or missed detections, and cannot accurately detect structural defects in non-appearance functional areas.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for detecting surface defects on an air conditioner plastic panel, applied to non-functional areas of the air conditioner plastic panel surface, comprising the following steps:

[0006] Acquire each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and a three-dimensional design model of the plastic panel of the air conditioner under test;

[0007] Based on each of the two-dimensional images and the three-dimensional design model, the geometric image features of each of the two-dimensional images and the geometric model features of the three-dimensional design model located at the same viewpoint in each of the two-dimensional images are extracted respectively;

[0008] The geometric image features and the geometric model features of each two-dimensional image at the same viewpoint are compared and analyzed to obtain the feature difference of each image.

[0009] The difference of each image feature is compared with a preset threshold for each image feature to obtain the detection result of surface defects of the air conditioner plastic panel.

[0010] According to an embodiment of the present invention, the step of extracting geometric image features of each two-dimensional image and geometric model features of the three-dimensional design model located at the same viewpoint in each two-dimensional image, based on each two-dimensional image and the three-dimensional design model, includes:

[0011] Based on each of the two-dimensional images, the acquisition viewpoint features of each of the two-dimensional images are obtained from multiple viewpoints;

[0012] By comparing the feature of each acquisition viewpoint with the feature of the viewpoint model of the three-dimensional design model at the same viewpoint in each two-dimensional image, the difference data of each viewpoint is obtained.

[0013] Based on each of the viewpoint difference data, pose correction is performed on each of the two-dimensional images to obtain each corrected two-dimensional image;

[0014] Based on each corrected two-dimensional image and the three-dimensional design model, the geometric image features of each two-dimensional image and the geometric model features of the three-dimensional design model located at the same viewpoint in each two-dimensional image are extracted respectively.

[0015] According to one embodiment of the present invention, after the step of comparing each image feature difference with a preset image feature threshold to obtain the detection result of surface defects of the air conditioner plastic panel, the method further includes:

[0016] The results of testing each of the air conditioner plastic panels were counted sequentially to obtain the number of times the air conditioner plastic panels were found to have defects;

[0017] If the number of tests exceeds a preset statistical threshold, the three-dimensional model of the qualified product of the air conditioner plastic panel to be tested is collected as the three-dimensional design model of the air conditioner plastic panel to be tested.

[0018] According to one embodiment of the present invention, the step of extracting geometric image features of each two-dimensional image and geometric model features of the three-dimensional design model located at the same viewpoint in each two-dimensional image based on each corrected two-dimensional image and the three-dimensional design model includes:

[0019] Based on each corrected two-dimensional image, geometric image features of the contour point set, circle center coordinates, and line parameters in each two-dimensional image are extracted;

[0020] Based on the three-dimensional design model, the geometric features of the contour point set, circle center coordinates, and line parameters in the three-dimensional design model are extracted.

[0021] According to one embodiment of the present invention, the steps of acquiring each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and the three-dimensional design model of the plastic panel of the air conditioner under test include:

[0022] The imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple perspectives.

[0023] The two-dimensional acquired images are subjected to image correction processing to obtain each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives;

[0024] Collect the original 3D design data of the plastic panel of the air conditioner under test;

[0025] The non-appearance functional areas in the original three-dimensional design data are segmented to obtain the three-dimensional design model of the plastic panel of the air conditioner to be tested.

[0026] According to one embodiment of the present invention, the step of acquiring two-dimensional images of the plastic panel of the air conditioner under test from multiple viewing angles using an imaging device includes:

[0027] Using viewpoint selection rules, an imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple viewpoints; wherein, the viewpoint selection rules include:

[0028] For each predetermined defect pattern obtained, the amount of geometric feature change caused by the defect pattern under different perspectives is evaluated, and all observation perspectives with geometric feature change exceeding a first preset threshold are selected as candidate observation perspectives.

[0029] Based on the selected multiple candidate observation perspectives, the degree of overlap of information obtained between different candidate observation perspectives is calculated, and redundant observation perspectives with information overlap exceeding a second preset threshold are removed to obtain the second observation perspective.

[0030] Based on the preset maximum number of observation angles, select a combination of observation angles from the second observation angle as multiple perspectives.

[0031] According to one embodiment of the present invention, for each acquired predetermined defect pattern, evaluating the amount of geometric feature change caused by the defect pattern under different viewpoints, and selecting all observation viewpoints whose geometric feature change exceeds a first preset threshold as candidate observation viewpoints includes:

[0032] Obtain the geometric shape definition for each of the predetermined defect patterns, the geometric shape definition including multiple geometric shape variations of the predetermined defect patterns and combined defect patterns formed by the simultaneous occurrence of multiple predetermined defect patterns;

[0033] Based on the geometric morphology variant or the combined defect morphology, the amount of geometric feature change caused by the defect mode under different perspectives is evaluated, and all observation perspectives with geometric feature change exceeding a first preset threshold are selected as candidate observation perspectives.

[0034] According to one embodiment of the present invention, the step of comparing each image feature difference with a preset image feature threshold to obtain the detection result of surface defects of an air conditioner plastic panel includes:

[0035] Each image feature difference is compared with a preset image feature threshold to obtain the result of whether the surface defect of the air conditioner plastic panel is qualified or unqualified.

[0036] If the result is that the surface defect of the air conditioner plastic panel is unqualified, the defect type and defect location are confirmed based on the three-dimensional design model of the air conditioner plastic panel to be tested.

[0037] According to one embodiment of the present invention, the step of confirming the defect type and defect location based on the three-dimensional design model of the plastic panel of the air conditioner under test includes:

[0038] Based on the three-dimensional design model of the plastic panel of the air conditioner under test and the preset defect judgment rules, the defect type and defect location are confirmed.

[0039] Among them, the preset defect judgment rules are determined based on the non-appearance structure of the air conditioner plastic panel and its various defect modes.

[0040] To improve the solution, this application also proposes a detection system for surface defects in air conditioner plastic panels, the system comprising:

[0041] The acquisition module is used to acquire each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and the three-dimensional design model of the plastic panel of the air conditioner under test.

[0042] The feature extraction module is used to extract the geometric image features of each two-dimensional image and the geometric model features of the three-dimensional design model located at the same viewpoint in each two-dimensional image, based on each two-dimensional image and the three-dimensional design model.

[0043] The comparison and recognition module is used to compare and analyze each geometric image feature and the geometric model feature of each two-dimensional image at the same viewpoint to obtain the feature difference of each image.

[0044] The detection module is used to compare each image feature difference with a preset image feature threshold to obtain the detection result of surface defects of the air conditioner plastic panel.

[0045] Compared with the prior art, the method and system for detecting surface defects of air conditioner plastic panels of the present invention have the following advantages:

[0046] This invention combines multi-view acquired 2D images with a 3D design model of the panel under test. Multi-view acquisition captures details from different directions, effectively reducing the impact of shadows and occlusion. Using the 3D design model as a benchmark provides more accurate reference. Geometric image features of each 2D image and geometric model features of the 3D design model at the same viewpoint in each 2D image are extracted, converting both the acquired 2D images and the 3D design model into quantifiable feature data. By comparing and analyzing each geometric image feature with the geometric model features at the same viewpoint in each 2D image, the geometric deviation between the actual panel represented by the 2D image and the ideal design model can be directly quantified. This effectively distinguishes between the complex contours of non-functional areas of the air conditioner plastic panel and geometric anomalies caused by defects, avoiding misjudging normal structural features as defects. By comparing the difference in each image feature with a preset threshold for each image feature, and setting a reasonable threshold, it is determined whether the difference in image feature exceeds an acceptable range, thereby determining whether a defect exists on the surface of the air conditioner plastic panel. This enables more accurate detection of non-functional areas of the air conditioner plastic panel. Attached Figure Description

[0047] To more clearly illustrate the specific embodiments of the present invention, the accompanying drawings used in the specific embodiments will be briefly described below. In all the drawings, the elements or parts are not necessarily drawn to scale.

[0048] Figure 1 This is a flowchart of a method for detecting surface defects in an air conditioner plastic panel according to the present invention.

[0049] Figure 2 This is a structural block diagram of a system for detecting surface defects in an air conditioner plastic panel according to the present invention.

[0050] In the figure: Acquisition module 210, feature extraction module 220, comparison and recognition module 230, detection module 240.

[0051] The implementation and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0052] The following drawings disclose several embodiments of the present invention. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the invention. That is, in some embodiments of the invention, these practical details are not essential. Furthermore, for the sake of simplicity, some conventional structures and components will be shown in the drawings in a simple schematic manner.

[0053] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0054] Furthermore, in this invention, the use of terms such as "first" and "second" is for descriptive purposes only and does not specifically refer to any order or sequence, nor is it intended to limit the invention. They are merely used to distinguish components or operations described using the same technical terms, and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but only if they are feasible for those skilled in the art. If a combination of technical solutions is contradictory or impossible to implement, such a combination should be considered nonexistent and not within the scope of protection claimed by this invention.

[0055] To further understand the content, features, and effects of this invention, the following embodiments are provided, and detailed descriptions are given below in conjunction with the accompanying drawings:

[0056] Please see Figure 1 This invention provides a method for detecting surface defects on air conditioner plastic panels, applicable to non-functional areas of the air conditioner plastic panel surface, comprising the following steps:

[0057] S100. Acquire each 2D image of the plastic panel of the air conditioner under test from multiple perspectives and a 3D design model of the plastic panel. Each 2D image can be acquired by setting multiple fixed positions on the plastic panel of the air conditioner under test. By acquiring each 2D image of the plastic panel of the air conditioner under test in non-functional areas from multiple perspectives, the loss of defect information caused by occlusion or uneven lighting in a single perspective is overcome, thus enabling a more comprehensive acquisition of surface information in non-functional areas. At the same time, the 3D design model can be obtained from standard CAD files obtained from the product development department, providing a precise comparison benchmark for subsequent geometric feature comparison.

[0058] S200. Based on each of the two-dimensional images and the three-dimensional design model, extract the geometric image features of each of the two-dimensional images and the geometric model features of the three-dimensional design model at the same viewpoint in each of the two-dimensional images. Extracting geometric image features can capture information such as the shape and contour of defects reflected in the two-dimensional images. Extracting the geometric model features of the three-dimensional design model at the corresponding viewpoint obtains the geometric shape that non-appearance functional areas should have under ideal conditions at that viewpoint. Specifically, existing image processing algorithms are used to extract quantifiable geometric image features such as contour lines, circles, and straight lines from each two-dimensional image and to extract quantifiable geometric model features such as contour lines, circles, and straight lines from the three-dimensional design model.

[0059] S300: Compare and analyze each geometric image feature with the geometric model feature of each two-dimensional image at the same viewpoint to obtain the feature difference for each image. Specifically, both the geometric image features and the geometric model features include a set of contour line points, center coordinates, and line parameters. For example, for a screw hole, the image feature difference is obtained by calculating the average distance between the corresponding contour line point sets, the offset between the corresponding center coordinates, and the difference between the corresponding line parameters between the geometric image features and the geometric model features. This effectively identifies the geometric deviation of the screw hole caused by defects.

[0060] S400. Compare each image feature difference with a preset image feature threshold to obtain the detection result of surface defects on the air conditioner plastic panel. For example, the average distance of the contour point set is used as the image feature threshold of 1 mm, the offset between the center coordinates of the circles is used as the image feature threshold of 0.2 mm, and the straight line parameter is used as the offset angle of the image feature threshold of 1 degree. If each image feature difference meets the preset image feature threshold, the detection result that there are no defects on the surface of the air conditioner plastic panel is obtained. By setting reasonable thresholds, it is determined whether the image feature difference exceeds the acceptable range, thereby determining whether a defect exists. This allows for rapid detection results and improves detection accuracy.

[0061] This invention acquires two-dimensional images of the air conditioner plastic panel under test from multiple perspectives, along with a three-dimensional design model of the panel. Multi-view acquisition captures details from different directions, effectively reducing the impact of shadows and occlusions. Using the three-dimensional design model as a benchmark provides more accurate comparison. Geometric features of each two-dimensional image and geometric model features of the three-dimensional design model at the same perspective in each two-dimensional image are extracted, transforming both the acquired two-dimensional images and the three-dimensional design model into quantifiable feature data. By comparing and analyzing each geometric image feature with the geometric model features at the same perspective in each two-dimensional image, the geometric deviation between the actual panel represented by the two-dimensional image and the ideal design model can be directly quantified. This effectively distinguishes between the complex contours of the air conditioner plastic panel's non-functional structural areas and geometric anomalies caused by defects, avoiding misjudging normal structural features as defects. By comparing the feature difference of each image with a preset feature threshold, a reasonable image feature threshold is set to determine whether the image feature difference exceeds an acceptable range, thereby determining whether a defect exists on the surface of the air conditioner plastic panel. This achieves more accurate detection of non-functional areas of the air conditioner plastic panel.

[0062] In some of the above embodiments, due to the influence of the installation position of the acquisition device, the acquired two-dimensional images have pose deviations, which will lead to errors in direct feature comparison and affect the detection accuracy. To address this, this application further proposes a step of extracting geometric image features of each two-dimensional image and geometric model features of the three-dimensional design model from the same viewpoint in each two-dimensional image, based on each two-dimensional image and the three-dimensional design model, including:

[0063] Based on each of the two-dimensional images, the acquisition viewpoint features of each of the two-dimensional images are obtained from multiple viewpoints.

[0064] By comparing the features of each acquired viewpoint with the viewpoint model features of the 3D design model at the same viewpoint in each 2D image, the viewpoint difference data for each image is obtained. This viewpoint difference data reflects the degree of deviation in pose between the 2D image and the 3D design model.

[0065] Based on each viewpoint difference data, pose correction is performed on each two-dimensional image to obtain a corrected two-dimensional image. In this embodiment, an existing pose correction algorithm is used to perform geometric transformation on the two-dimensional image based on the viewpoint difference data, so that its pose is consistent with the pose of the three-dimensional design model under the corresponding viewpoint. Through pose correction, the pose of the two-dimensional image is more consistent with the three-dimensional design model, thereby reducing the error of subsequent feature comparison.

[0066] Based on each corrected 2D image and the 3D design model, geometric image features of each 2D image and geometric model features of the 3D design model at the same viewpoint in each 2D image are extracted. Since the 2D images have undergone pose correction, the extracted geometric image features can more accurately reflect the true condition of the panel surface, thereby improving the accuracy of defect detection.

[0067] Specifically, based on each 2D image, obtaining the acquisition viewpoint features of each 2D image from multiple perspectives involves pre-setting or detecting several identifiable marker points on each 2D image. These marker points also have corresponding known 3D coordinates on the 3D design model. The acquisition viewpoint features can be specifically the pixel coordinates of these marker points on the 2D image. Comparing each acquisition viewpoint feature with the viewpoint model features of the 3D design model at the same perspective in each 2D image can be achieved by projecting the 3D coordinates of the corresponding marker points on the 3D design model onto a virtual camera plane consistent with the acquisition viewpoint of the 2D image, obtaining its 2D projected coordinates as the viewpoint model features, and then calculating the deviation between the marker point coordinates and the projected coordinates on the 2D image as viewpoint difference data. Alternatively, a transformation matrix from the 3D model coordinate system to the 2D image coordinate system can be calculated based on these corresponding points as viewpoint difference data. Based on each viewpoint difference data, attitude correction for each 2D image can be performed by using the calculated transformation matrix to perform a geometric transformation on the entire 2D image, remapping the pixel positions in the image to obtain the attitude-corrected 2D image. Based on each corrected 2D image and 3D design model, geometric image features of each 2D image and geometric model features of the 3D design model at the same viewpoint in each 2D image are extracted. On the corrected 2D image, algorithms such as edge detection, contour extraction, circle detection, or line detection are used to extract geometric image features such as contour point sets, circle center coordinates, or line parameters. Simultaneously, geometric model features such as contour point sets, circle center coordinates, or line parameters for the corresponding region are extracted from the 3D design model. Therefore, by performing pose correction on the 2D image before extracting geometric image features, errors caused by image pose deviations are eliminated or reduced, improving the accuracy of subsequent geometric feature comparison and further enhancing the precision of surface defect detection for air conditioner plastic panels.

[0068] In some of the above embodiments, when the geometry of certain non-appearance functional areas of the actually produced panel deviates from the computer-aided design model benchmark currently used by the detection system due to design changes or stable systematic factors in the manufacturing process, it may lead to qualified structures being misjudged as defects. This application further proposes that after comparing each of the image feature differences with a preset image feature threshold to obtain the detection result of surface defects on the air conditioner plastic panel, it also includes:

[0069] The results of each air conditioner plastic panel inspection are counted sequentially to determine the number of times defects were detected in the air conditioner plastic panel. The number of inspections is at least four.

[0070] If the number of inspections exceeds a preset statistical threshold, a 3D model of a qualified air conditioner plastic panel is collected as the 3D design model of the air conditioner plastic panel to be tested. The preset statistical threshold is three times. Specifically, a structured light 3D scanner is used to perform a high-precision scan on a 3D air conditioner plastic panel of the same model that has been manually inspected and confirmed to be defect-free, obtaining its complete 3D point cloud data and generating a triangular mesh model. This generated triangular mesh model is used as the 3D design model for inspecting similar air conditioner plastic panels. This application dynamically adjusts the inspection benchmark by comparing the number of inspections indicating defects in the air conditioner plastic panel with a preset statistical threshold, thereby reducing misjudgments caused by inherent differences between the model and the actual qualified product, improving the accuracy of defect detection and adaptability to changes in the production process.

[0071] In some of the above embodiments, this application further proposes the step of extracting geometric image features of each of the two-dimensional images and geometric model features of the three-dimensional design model located at the same viewpoint in each of the two-dimensional images based on each corrected two-dimensional image and the three-dimensional design model, including:

[0072] Based on each corrected two-dimensional image, geometric image features, including contour point sets, center coordinates, and line parameters, are extracted from each two-dimensional image. Contour points describe the outline of objects in the image, center coordinates can locate circular or arc-shaped structures, and line parameters describe the straight edges in the image, thereby reducing errors in subsequent feature comparisons. In this embodiment, the extraction of contour point sets can be achieved by first preprocessing the image (e.g., grayscale conversion, noise reduction), then applying an edge detection algorithm to obtain the edges of the image, and finally using a contour tracking algorithm to extract the complete contour from the edge pixels, resulting in a series of ordered contour point sets. For the extraction of center coordinates, the edge detection results are used to detect circular or arc-shaped structures in the image, thereby obtaining the center coordinates and radii of these structures.

[0073] Based on the three-dimensional design model, the geometric features of the contour point set, circle center coordinates, and line parameters in the three-dimensional design model are extracted.

[0074] This application achieves a precise correspondence between extracted image features and model features, thereby enabling more accurate detection of geometric defects such as panel deformation, missing material, hole misalignment, or edge bending, thus improving the accuracy of defect detection.

[0075] In some of the above embodiments, this application further proposes a step for obtaining each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and a three-dimensional design model of the plastic panel of the air conditioner under test, including:

[0076] The imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple perspectives.

[0077] Image correction processing is performed on the acquired two-dimensional images to obtain each two-dimensional image of the plastic panel of the air conditioner under test from multiple viewing angles. This image correction processing eliminates or reduces various errors that occur during image acquisition, such as using existing geometric correction algorithms for distortion correction and noise filtering algorithms for noise removal; thereby improving image quality and providing a more accurate data foundation for subsequent feature extraction and defect detection. By correcting the acquired two-dimensional images, it ensures that the two-dimensional images used in subsequent analysis more realistically reflect the actual condition of the panel, reducing misjudgments caused by image quality issues.

[0078] Collect the original 3D design data of the plastic panel of the air conditioner to be tested.

[0079] The non-appearance functional areas in the original 3D design data are segmented to obtain the 3D design model of the air conditioner plastic panel to be tested. Specifically, segmentation is achieved using segmentation algorithms based on geometric features or semantic information. Segmentation removes data irrelevant to defect detection, such as data from appearance areas, thereby reducing computational load, improving detection efficiency, and preventing these irrelevant data from interfering with defect detection results. Focusing only on non-appearance functional areas allows for a more focused detection process, improving the accuracy and reliability of the detection.

[0080] This application eliminates image distortion and noise by performing image correction processing on the original two-dimensional images acquired by the imaging device, resulting in high-quality two-dimensional images. Simultaneously, by segmenting the acquired original three-dimensional design data and removing data irrelevant to non-functional areas, a simplified three-dimensional design model focused on key regions is obtained. The high-quality two-dimensional images and simplified three-dimensional design model enable more accurate and efficient subsequent geometric feature extraction and comparative analysis based on these data, improving the accuracy and efficiency of defect detection on air conditioner plastic panels. This is particularly beneficial for detecting defects in complex and detailed non-functional areas, where high-quality input data significantly reduces false positives and false negatives.

[0081] In some of the above embodiments, this application further proposes a step of acquiring two-dimensional images of the plastic panel of the air conditioner under test from multiple viewing angles using an imaging device, including:

[0082] Using viewpoint selection rules, an imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple viewpoints; wherein, the viewpoint selection rules include:

[0083] For each acquired predetermined defect pattern, the amount of geometric feature change caused by that defect pattern under different viewing angles is evaluated. Viewing angles where the amount of geometric feature change exceeds a first preset threshold are selected as candidate viewing angles. Here, the amount of geometric feature change refers to the degree of change in the geometric shape of the predetermined defect pattern in a two-dimensional image relative to a defect-free state or a standard shape under different viewing angles. It can be evaluated using image processing techniques by comparing the differences between the geometric attributes of the defect region projected onto the image plane under different viewing angles and standard or reference values. By evaluating the amount of geometric feature change, the observability of defects under different viewing angles is quantified, and those viewing angles most conducive to defect detection are selected. Selecting viewing angles where the amount of geometric feature change exceeds the first preset threshold ensures that the selected viewing angles have sufficient sensitivity to the defect pattern and can effectively capture the features of the defect.

[0084] Based on the selected candidate observation perspectives, the degree of overlap of information acquired between different candidate observation perspectives is calculated. Redundant observation perspectives with an information overlap exceeding a second preset threshold are removed to obtain the second observation perspective. Here, the degree of information overlap refers to the redundancy of information about the surface features of the plastic panel of the air conditioner under test acquired by different candidate observation perspectives.

[0085] Based on the preset maximum number of observation angles, select a combination of observation angles from the second observation angle as multiple perspectives.

[0086] Specifically, for example, regarding a snap-fit ​​breakage, images of the panel are captured from different lateral angles, and the projected length of the breakage area in the image is calculated. If the change in the projected length of the breakage area at a certain viewpoint exceeds 3 cm (a first preset threshold of 1 cm), then that viewpoint is considered a candidate. Next, based on the selected candidate viewpoints, the degree of overlap in the information acquired between them is calculated. For example, if viewpoints A and B can both clearly see the breakage of the same snap-fit, and the position and shape of the breakage area captured by them in the image are very similar, then viewpoints A and B are considered to have a high degree of information overlap, and one of the viewpoints is removed. Finally, assuming that after the above steps, there are 5 remaining second observation viewpoints, while the preset maximum number of observation viewpoints is 3, 3 can be selected from these 5 viewpoints. For example, the viewpoint combination that covers the most different defect types can be selected, or the 3 viewpoint combinations with the largest sum of geometric feature changes can be selected. These 3 finalized viewpoints are the multiple viewpoints used for actual acquisition. This scheme can selectively choose the most advantageous viewpoints for defect capture based on predetermined defect pattern characteristics, avoiding blind or empirical viewpoint selection. By eliminating redundant viewing angles, unnecessary image acquisition and processing burdens are reduced, improving detection efficiency. Limiting the total number of observation angles makes the solution more adaptable to the cycle time requirements of actual production lines. This optimized viewing angle acquisition method, combined with subsequent detection methods based on geometric feature comparison, significantly improves the accuracy and reliability of detecting complex defects in non-functional areas of air conditioner plastic panels, effectively reducing false alarm and false negative rates and ensuring product quality.

[0087] In some of the above embodiments, this application further proposes a step of evaluating the amount of geometric feature change caused by each predetermined defect pattern from different perspectives, and selecting observation perspectives whose geometric feature change exceeds a first preset threshold as candidate observation perspectives, including:

[0088] Obtain the geometric morphology definition for each of the predetermined defect patterns. The geometric morphology definition includes multiple geometric morphology variants of the predetermined defect pattern and combined defect patterns formed by the simultaneous occurrence of multiple predetermined defect patterns. The geometric morphology definition describes the geometric attributes of the predetermined defect pattern in three-dimensional space, such as shape, size, and location. Multiple geometric morphology variants of the predetermined defect pattern refer to the various specific geometric forms exhibited by the same type of defect under different locations, degrees of influence, or environmental factors. These can be defined using CAD model parameter variations, actual defect sample scan data, or deformation models based on physical simulation.

[0089] Based on the geometric morphology variant or the combined defect morphology, the amount of geometric feature change caused by the defect mode under different perspectives is evaluated, and all observation perspectives with geometric feature change exceeding a first preset threshold are selected as candidate observation perspectives.

[0090] By acquiring the geometric morphology definition of each predetermined defect pattern, which includes multiple geometric morphological variations of the predetermined defect pattern as well as combined defect patterns formed by the simultaneous occurrence of multiple predetermined defect patterns, the amount of geometric feature change caused by the defect pattern under different viewpoints is evaluated based on a more comprehensive geometric morphological description. The evaluation process considers various complex forms of defects, not just idealized single forms. This enables more accurate prediction of how various actual possible defect forms will affect the representation of their geometric features in 2D images under different observation viewpoints. The evaluation results can more realistically reflect the visibility of defects under different viewpoints. This allows the selected candidate observation viewpoints to more effectively capture various complex defects that occur in panels in actual production, avoiding missed detections or misjudgments caused by considering only a single form. More accurate viewpoint selection provides more optimized input for subsequent image acquisition and feature extraction, thereby improving the accuracy and reliability of the entire detection method.

[0091] In some of the above embodiments, this application further proposes a step of comparing each of the image feature differences with a preset image feature threshold to obtain the detection result of surface defects of the air conditioner plastic panel, including:

[0092] The difference in each image feature is compared with a preset threshold for each image feature to determine whether the surface defect of the air conditioner plastic panel is acceptable or unacceptable. The specific judgment of whether the detection result is acceptable or unacceptable allows for the rapid screening of panels with defects.

[0093] If the surface defects of the air conditioner's plastic panel are found to be unqualified, the defect type and location are confirmed based on the three-dimensional design model of the panel. The three-dimensional design model contains complete geometric shape, size, and structural information, providing a precise reference for subsequent defect analysis. Confirming the defect type and location involves identifying the specific defect attributes of the unqualified area, such as scratches, dents, breaks, or dimensional deviations, and determining the specific coordinates or structural region of the defect in the panel's three-dimensional space. This aims to provide detailed defect information, facilitating analysis of the cause of the defect and quality traceability.

[0094] Specifically, when a test result indicates that a plastic panel of an air conditioner is defective, there is a significant discrepancy between the image features and the model features of a certain abnormal area. The location information of this abnormal area in the 2D image, combined with the viewing angle information at the time of acquisition, is mapped to the corresponding area on the 3D design model. Then, the system analyzes the geometric features of this 3D area, such as measuring its size, shape, or surface flatness, and comparing it with the precise geometric data of the 3D design model at that location. If the actual size of a certain clip structure is significantly smaller than the design size, the defect type is determined to be "size deviation." Simultaneously, the specific location of the defect is determined based on the 3D design model, for example, "the first clip at the lower right corner of the panel." Detailed defect type and location information facilitates subsequent quality analysis, root cause tracing, and production process improvement.

[0095] In some of the above embodiments, this application further proposes a step for confirming the defect type and defect location based on the three-dimensional design model of the air conditioner plastic panel under test, including:

[0096] Based on the three-dimensional design model of the plastic panel of the air conditioner under test and the preset defect judgment rules, the defect type and defect location are confirmed.

[0097] The preset defect judgment rules are determined based on the non-appearance structure of the air conditioner plastic panel and its various defect patterns. In this embodiment, the preset defect judgment rules refer to the standards or logical set used to determine which known defect type a detected geometric deviation corresponds to. These rules can be implemented using a structured database, rule engine, or lookup table, with the aim of providing a basis for accurate defect classification. The determination based on the non-appearance structure of the air conditioner plastic panel and its various defect patterns means that the process of establishing the preset defect judgment rules is based on an in-depth analysis of the structural characteristics of non-appearance functional areas (such as clips, screw holes, and reinforcing ribs) on the air conditioner plastic panel, and the collection and study of various typical defect forms that may occur in these areas during actual production (such as missing material, deformation, breakage, bubbles, and dimensional deviations). This allows for the extraction of judgment criteria for each defect pattern, ensuring the relevance and effectiveness of the judgment rules and enabling them to accurately identify common defects in specific areas of specific products.

[0098] This application's solution combines a 3D design model of the air conditioner's plastic panel under test with preset defect judgment rules. This allows for more accurate identification of defect types and locations after detecting non-conforming results. Specifically, the 3D design model provides the ideal geometric shape and spatial location information of the air conditioner's plastic panel under test, used to initially locate the area where the defect occurred and the corresponding ideal structural features. Based on this, the preset defect judgment rules include detailed judgment criteria for non-appearance structures of the air conditioner's plastic panel and their various defect patterns. When a geometric deviation is detected in a certain area, the 3D design model is first used to determine which type of non-appearance structure the area belongs to (e.g., a clip, a screw hole, or a reinforcing rib). Then, according to the preset defect judgment rules for this type of structure and the detected deviation characteristics (e.g., the shape, size, or positional relationship of the deviation), a matching defect pattern is found. This combined approach allows the system to not only know "where" the problem occurred (located through the 3D model) but also "what type" of problem occurred (classified through the judgment rules). For example, if an incomplete edge is detected in a certain snap-fit ​​area, the 3D design model confirms that this location is the tip of the snap-fit. A pre-defined defect judgment rule includes a clause stating: "If an edge is missing at the snap-fit ​​tip and the missing length exceeds X, it is judged as a missing material defect." By simultaneously referencing the structural context provided by the 3D model and the defect knowledge provided by the judgment rules, the system can eliminate interference from the complexity of the structure itself, accurately distinguish normal structural features from actual defects, and achieve accurate confirmation of defect type and location. This solves the problem that relying solely on an ideal 3D model cannot accurately determine the actual defect pattern.

[0099] For a more complete solution, please refer to [link / reference]. Figure 2 This application also proposes a detection system for surface defects of air conditioner plastic panels, the system comprising an acquisition module 210, a feature extraction module 220, a comparison and identification module 230, and a detection module 240.

[0100] The acquisition module 210 is used to acquire each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and the three-dimensional design model of the plastic panel of the air conditioner under test.

[0101] The feature extraction module 220 is used to extract the geometric image features of each of the two-dimensional images and the geometric model features of the three-dimensional design model located at the same viewpoint in each of the two-dimensional images, based on each of the two-dimensional images and the three-dimensional design model.

[0102] The comparison and recognition module 230 is used to compare and analyze each geometric image feature and the geometric model features of each two-dimensional image located at the same viewpoint to obtain the feature difference of each image.

[0103] The detection module 240 is used to compare each image feature difference with a preset image feature threshold to obtain the detection result of surface defects of the air conditioner plastic panel.

[0104] This application, through the configuration of an acquisition module 210, a feature extraction module 220, a comparison and recognition module 230, and a detection module 240, uses the acquisition module 210 to acquire each two-dimensional image of the air conditioner plastic panel under test from multiple perspectives and a three-dimensional design model of the air conditioner plastic panel under test. The feature extraction module 220 extracts geometric image features from each two-dimensional image and geometric model features of the three-dimensional design model from the same perspective in each two-dimensional image, based on each two-dimensional image and the three-dimensional design model. The comparison and recognition module 230 compares and analyzes each geometric image feature and the geometric model features from the same perspective in each two-dimensional image to obtain a feature difference for each image. The detection module 240 compares each image feature difference with a preset image feature threshold to obtain the detection result of surface defects on the air conditioner plastic panel. By comparing each image feature difference with a preset image feature threshold, a reasonable image feature threshold is set, and it is determined whether the image feature difference exceeds an acceptable range, thereby determining whether a defect exists on the surface of the air conditioner plastic panel. This enables more accurate detection of non-functional areas of the air conditioner plastic panel.

[0105] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for detecting surface defects in an air conditioner plastic panel, characterized in that, For non-functional areas on the surface of an air conditioner's plastic panel, the following steps are included: Acquire each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and a three-dimensional design model of the plastic panel of the air conditioner under test; Based on each of the two-dimensional images and the three-dimensional design model, the geometric image features of each of the two-dimensional images and the geometric model features of the three-dimensional design model located at the same viewpoint in each of the two-dimensional images are extracted respectively; The geometric image features and the geometric model features of each two-dimensional image at the same viewpoint are compared and analyzed to obtain the feature difference of each image. The difference of each image feature is compared with a preset threshold for each image feature to obtain the detection result of surface defects of the air conditioner plastic panel; The steps for acquiring each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and the three-dimensional design model of the plastic panel include: The imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple perspectives. The two-dimensional acquired images are subjected to image correction processing to obtain each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives; Collect the original 3D design data of the plastic panel of the air conditioner under test; The non-appearance functional areas in the original three-dimensional design data are segmented to obtain the three-dimensional design model of the plastic panel of the air conditioner to be tested. The steps of acquiring two-dimensional images of the plastic panel of the air conditioner under test from multiple perspectives using an imaging device include: Using viewpoint selection rules, an imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple viewpoints; wherein, the viewpoint selection rules include: For each predetermined defect pattern obtained, the amount of geometric feature change caused by the defect pattern under different perspectives is evaluated, and the observation perspective where the amount of geometric feature change exceeds a first preset threshold is selected as a candidate observation perspective. Based on the selected multiple candidate observation perspectives, the degree of overlap of information obtained between different candidate observation perspectives is calculated, and redundant observation perspectives with information overlap exceeding a second preset threshold are removed to obtain the second observation perspective. Based on the preset maximum number of observation angles, select a combination of observation angles from the second observation angle as multiple perspectives.

2. The method for detecting surface defects in an air conditioner plastic panel according to claim 1, characterized in that, The steps of extracting geometric image features of each two-dimensional image and geometric model features of the three-dimensional design model from the same viewpoint in each two-dimensional image, based on each two-dimensional image and the three-dimensional design model, include: Based on each of the two-dimensional images, the acquisition viewpoint features of each of the two-dimensional images are obtained from multiple viewpoints; By comparing the feature of each acquisition viewpoint with the feature of the viewpoint model of the three-dimensional design model at the same viewpoint in each two-dimensional image, the difference data of each viewpoint is obtained. Based on each of the viewpoint difference data, pose correction is performed on each of the two-dimensional images to obtain each corrected two-dimensional image; Based on each corrected two-dimensional image and the three-dimensional design model, the geometric image features of each two-dimensional image and the geometric model features of the three-dimensional design model located at the same viewpoint in each two-dimensional image are extracted respectively.

3. The method for detecting surface defects in an air conditioner plastic panel according to claim 1, characterized in that, After comparing each image feature difference with a preset image feature threshold to obtain the detection result of surface defects on the air conditioner plastic panel, the method further includes: The results of testing each of the air conditioner plastic panels were counted sequentially to obtain the number of times the air conditioner plastic panels were found to have defects; If the number of tests exceeds a preset statistical threshold, the three-dimensional model of the qualified product of the air conditioner plastic panel to be tested is collected as the three-dimensional design model of the air conditioner plastic panel to be tested.

4. The method for detecting surface defects in an air conditioner plastic panel according to claim 2, characterized in that, The step of extracting geometric image features of each two-dimensional image and geometric model features of the three-dimensional design model from the same viewpoint in each two-dimensional image based on the corrected two-dimensional image and the three-dimensional design model includes: Based on each of the corrected two-dimensional images, geometric image features are extracted from each of the two-dimensional images, including a set of contour points, center coordinates of circles, and line parameters; Based on the three-dimensional design model, geometric model features are extracted from the three-dimensional design model. The geometric model features include contour point set, circle center coordinates, and line parameters.

5. The method for detecting surface defects in an air conditioner plastic panel according to claim 1, characterized in that, For each acquired predetermined defect pattern, the step of evaluating the amount of geometric feature change caused by the defect pattern under different viewpoints, and selecting the observation viewpoint where the amount of geometric feature change exceeds a first preset threshold as a candidate observation viewpoint includes: Obtain the geometric shape definition for each of the predetermined defect patterns, the geometric shape definition including multiple geometric shape variations of the predetermined defect patterns and combined defect patterns formed by the simultaneous occurrence of multiple predetermined defect patterns; Based on the geometric morphology variant or the combined defect morphology, the amount of geometric feature change caused by the defect pattern under different perspectives is evaluated, and the observation perspective with the amount of geometric feature change exceeding a first preset threshold is selected as the candidate observation perspective.

6. The method for detecting surface defects in an air conditioner plastic panel according to claim 1, characterized in that, The steps for comparing each image feature difference with a preset image feature threshold to obtain the detection result of surface defects on the air conditioner plastic panel include: Each image feature difference is compared with a preset image feature threshold to obtain the result of whether the surface defect of the air conditioner plastic panel is qualified or unqualified. If the result is that the surface defect of the air conditioner plastic panel is unqualified, the defect type and defect location are confirmed based on the three-dimensional design model of the air conditioner plastic panel to be tested.

7. The method for detecting surface defects in an air conditioner plastic panel according to claim 6, characterized in that, Based on the three-dimensional design model of the plastic panel of the air conditioner under test, the steps to confirm the defect type and defect location include: Based on the three-dimensional design model of the plastic panel of the air conditioner under test and the preset defect judgment rules, the defect type and defect location are confirmed. Among them, the preset defect judgment rules are determined based on the non-appearance structure of the air conditioner plastic panel and its various defect modes.

8. A system for detecting surface defects in an air conditioner plastic panel, characterized in that, The system includes: The acquisition module is used to acquire each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and the three-dimensional design model of the plastic panel of the air conditioner under test. The feature extraction module is used to extract the geometric image features of each two-dimensional image and the geometric model features of the three-dimensional design model located at the same viewpoint in each two-dimensional image, based on each two-dimensional image and the three-dimensional design model. The comparison and recognition module is used to compare and analyze each geometric image feature and the geometric model feature of each two-dimensional image at the same viewpoint to obtain the feature difference of each image. The detection module is used to compare each image feature difference with a preset image feature threshold to obtain the detection result of surface defects of the air conditioner plastic panel; The acquisition of each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives and the three-dimensional design model of the plastic panel of the air conditioner under test include: The imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple perspectives. The two-dimensional acquired images are subjected to image correction processing to obtain each two-dimensional image of the plastic panel of the air conditioner under test from multiple perspectives; Collect the original 3D design data of the plastic panel of the air conditioner under test; The non-appearance functional areas in the original three-dimensional design data are segmented to obtain the three-dimensional design model of the plastic panel of the air conditioner to be tested. The imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple perspectives, including: Using viewpoint selection rules, an imaging device acquires two-dimensional images of the plastic panel of the air conditioner under test from multiple viewpoints; wherein, the viewpoint selection rules include: For each predetermined defect pattern obtained, the amount of geometric feature change caused by the defect pattern under different perspectives is evaluated, and the observation perspective where the amount of geometric feature change exceeds a first preset threshold is selected as a candidate observation perspective. Based on the selected multiple candidate observation perspectives, the degree of overlap of information obtained between different candidate observation perspectives is calculated, and redundant observation perspectives with information overlap exceeding a second preset threshold are removed to obtain the second observation perspective. Based on the preset maximum number of observation angles, select a combination of observation angles from the second observation angle as multiple perspectives.

Images