A method and system for screen surface bubble detection
By combining dual-mode imaging with frequency domain processing and overlap ratio determination, the problem of false detection of bubbles on the screen surface was solved, achieving bubble detection with high accuracy and low false detection rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGSU FULAT AUTOMATION EQUIP CO LTD
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-24
AI Technical Summary
In existing technologies, bubbles under the protective film on the screen surface are easily misjudged as screen defects during inspection, resulting in a high false detection rate for AVI devices and making it difficult to accurately distinguish bubbles from other pseudo-defects.
A dual-mode imaging method was adopted, acquiring images under both screen-on and screen-off states. Frequency domain processing was combined with fast Fourier transform and Gaussian high-pass filtering. By performing morphological dilation and intersection operations, the overlap ratio was used to determine bubble defects, and a differential area threshold was set to filter bubble regions.
It effectively distinguishes between bubbles and false defects, improves the accuracy of detection, reduces the false detection rate, and achieves efficient and accurate detection of bubbles.
Smart Images
Figure CN122453740A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image analysis technology, and in particular relates to a method and system for detecting bubbles on the surface of a screen. Background Technology
[0002] In the display manufacturing process, AVI (Auto Visual Inspection) equipment is widely used to detect various defects on the screen surface. In actual production, the screen surface is usually covered with a protective film. If there are air bubbles between the protective film and the screen itself, these air bubbles will interfere with the optical imaging results when the AVI equipment performs defect detection, leading to false detections. That is, the air bubbles under the protective film will be mistaken for defects in the screen itself, thus affecting the accurate judgment of the screen quality.
[0003] Therefore, there is an urgent need for a method and system that can accurately and quickly detect bubbles under the protective film on the screen surface to distinguish them from actual defects in the screen itself and reduce the false detection rate of AVI devices. Summary of the Invention
[0004] To address the problems existing in the prior art, this invention proposes a method and system for detecting bubbles on the screen surface.
[0005] The technical solution of the present invention is as follows:
[0006] A method for detecting bubbles on a screen surface, comprising:
[0007] For the effective display areas of the first image captured when the screen is on and the second image captured when the screen is off and illuminated by an external light source, frequency domain processing based on fast Fourier transform and Gaussian high-pass filtering is performed to enhance the high-frequency components of the bubble edge, resulting in the first spatial domain image and the second spatial domain image.
[0008] Multi-peak automatic segmentation is performed on the first spatial domain image and the second spatial domain image respectively to obtain bubble candidate regions. Morphological dilation and intersection operations are performed on the bubble candidate regions to connect neighboring regions while keeping the original area of each bubble candidate region unchanged, so as to form a complete bubble region.
[0009] A first candidate bubble region set is selected from the complete bubble region corresponding to the first spatial domain image based on a first preset area threshold, and a second candidate bubble region set is selected from the complete bubble region corresponding to the second spatial domain image based on a second preset area threshold;
[0010] When the overlap ratio of any overlapping region in the first candidate bubble region set and the second candidate bubble region set is greater than a preset ratio threshold, the overlapping region is determined to be a bubble defect. The overlap ratio is the ratio of the intersection area to the union area of the overlapping regions.
[0011] Furthermore, the first preset area threshold is greater than the second preset area threshold.
[0012] Furthermore, the first preset area threshold is set to be greater than 5 pixels, and the second preset area threshold is set to be greater than 3 pixels.
[0013] Furthermore, the morphological dilation and intersection operations performed on the bubble candidate regions, such that neighboring regions are connected while maintaining the original area of each bubble candidate region, specifically include:
[0014] A morphological dilation operation is performed on the candidate bubble region to obtain the dilated region. Connectivity component analysis is then performed on the dilated region to split it into multiple independent sub-regions. The intersection region between each sub-region and the candidate bubble region before dilation is then calculated. The intersection region is the complete bubble region.
[0015] Furthermore, the time interval between acquiring the first image and the second image is a time-division strobe interval, and the time-division strobe interval is no greater than 100 microseconds.
[0016] Furthermore, the frequency domain processing based on Fast Fourier Transform and Gaussian high-pass filtering includes:
[0017] The original image is centered, and then a fast Fourier transform is performed to obtain the frequency domain image.
[0018] The frequency domain image is multiplied by the transfer function of a Gaussian high-pass filter to suppress low-frequency components and preserve high-frequency components.
[0019] The filtered frequency domain image is subjected to an inverse fast Fourier transform, then centered and the real part is taken to obtain the spatial domain image.
[0020] Furthermore, the transfer function of the Gaussian high-pass filter is:
[0021] ;
[0022] In the formula, It is a frequency domain point Distance to the center of the frequency rectangle; and These are the frequencies in the horizontal and vertical directions, respectively. It is the cutoff frequency.
[0023] Furthermore, the automatic multi-peak segmentation includes:
[0024] The grayscale histogram of the spatial domain image is obtained, and the grayscale histogram is smoothed to eliminate noise interference. The local minimum value in the smoothed histogram is searched as the segmentation threshold, and the image is segmented into multiple bubble candidate regions using the segmentation threshold.
[0025] Furthermore, the preset ratio threshold is 0.4.
[0026] A screen surface bubble detection system includes a light source control module, an imaging module, and an image processing module;
[0027] The light source control module is used to output a screen lighting control signal to the screen under test and to output a strobe control signal to an external light source;
[0028] The imaging module is time-synchronized with the light source control module and is used to acquire a first image when the screen is lit and a second image when the screen is off and the external light source is illuminating, in response to the flicker control signal.
[0029] The image processing module is communicatively connected to the imaging module and is used to receive the first image and the second image, and to execute the screen surface bubble detection method as described in any of the preceding claims to determine bubble defects.
[0030] Compared with the prior art, the present invention has the following beneficial effects:
[0031] This invention provides a method and system for detecting bubbles on the surface of a screen. By employing a technical solution of dual-mode imaging and frequency domain feature fusion judgment, this method and system effectively solves the problem of difficulty in distinguishing bubble features from false defects in a single imaging mode, and achieves high accuracy in bubble detection and low false detection rate control.
[0032] The method of this invention acquires transmitted illumination images under screen illumination and external reflected illumination images under screen de-illumination. Since bubbles exhibit different imaging characteristics under the two optical conditions—showing concentric interference fringes and appearing larger due to halo effects under transmitted illumination, and appearing as bright spots with an image size close to their true size under reflected illumination—by setting differentiated area filtering thresholds (greater than 5 pixels in screen mode and greater than 3 pixels in illumination mode), bubbles can be effectively extracted under their respective imaging modes. This solves the problem of bubble features being easily confused with false defects such as dust and scratches under a single light source, improving the targeting of candidate region extraction.
[0033] This invention performs frequency domain processing on both acquired dual-channel images based on Fast Fourier Transform and Gaussian high-pass filtering. This processing, by converting the image to the frequency domain and suppressing low-frequency components representing smooth backgrounds while enhancing high-frequency components representing edges and textures, significantly reduces background interference caused by uniform reflections from screen content or surfaces, highlighting detailed features such as halos, stripes, or bright spots at the bubble edges. Furthermore, by combining multi-peak automatic segmentation with morphological dilation and intersection operations, adjacent regions broken due to optical effects are connected into complete bubbles while maintaining the original area of each bubble region. This effectively solves the problem of missed detections or underestimation of bubble area caused by imaging fragmentation.
[0034] This invention introduces a dual-path candidate region overlap ratio determination mechanism. By calculating the ratio of the intersection area to the union area of the candidate region sets in the dot-screen mode and the candidate region sets in the lighting mode, and setting an overlap ratio threshold (greater than 0.4), only regions that stably exhibit bubble characteristics in both imaging modes are ultimately determined as bubble defects. This mechanism fundamentally filters out noise or false defects that occasionally appear in a single mode, solving the problem of high false detection rates in single-mode detection. Attached Figure Description
[0035] Figure 1 This is a schematic diagram of the process for detecting bubbles on the screen surface;
[0036] Figure 2 A schematic diagram of the screen structure with a protective film;
[0037] Figure 3 This is a schematic diagram of bubble imaging in white field dot-screen mode;
[0038] Figure 4 This is a schematic diagram showing the extraction results of the effective display area (AA area) of the screen;
[0039] Figure 5 This is a schematic diagram of the frequency domain processing results of an image in dot-screen mode;
[0040] Figure 6 This is a schematic diagram of the bubble candidate region detection results in dot-screen mode;
[0041] Figure 7 This is a schematic diagram of the bubble imaging under external lighting mode;
[0042] Figure 8 This is a schematic diagram of the frequency domain processing results of an image under lighting conditions;
[0043] Figure 9 This is a schematic diagram of the detection results of the candidate bubble region under illumination mode. Detailed Implementation
[0044] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.
[0045] Example 1:
[0046] The present invention provides a method for detecting bubbles on the surface of a screen, such as... Figure 1 As shown, it includes:
[0047] For the first image captured when the screen is on and the second image captured when the screen is off but illuminated by an external light source, the same effective display area (i.e., the effective physical display area) that corresponds completely in spatial location is subjected to frequency domain processing based on Fast Fourier Transform and Gaussian high-pass filtering to enhance the high-frequency components of the bubble edges, resulting in a first spatial domain image and a second spatial domain image. In other words, under two different lighting conditions, the distribution of the screen's identical physical location is processed to ensure effective subsequent overlap and comparison. The extraction of the effective display area employs a multi-peak automatic segmentation method based on grayscale histograms. By segmenting the image when the screen is on, the AA region of the screen is accurately captured (see...). Figure 4 ).
[0048] Multi-peak automatic segmentation is performed on the first spatial domain image and the second spatial domain image respectively to obtain bubble candidate regions. Morphological dilation is performed on the bubble candidate regions and intersection operation is performed with the independent connected components obtained by connected component analysis, so that neighboring regions are connected to form complete bubble regions while keeping the original area of each bubble candidate region unchanged.
[0049] A first candidate bubble region set is selected from the complete bubble region corresponding to the first spatial domain image based on a first preset area threshold, and a second candidate bubble region set is selected from the complete bubble region corresponding to the second spatial domain image based on a second preset area threshold;
[0050] When the overlap ratio of any overlapping region between the first and second candidate bubble regions exceeds a preset threshold, the overlapping region is determined to be a bubble defect. If an overlapping region is determined to be a bubble defect and contains multiple non-connected candidate regions, only the one with the largest area is retained as the final detection result output. The overlap ratio is the ratio of the intersection area to the union area of the overlapping regions. To avoid outputting a fragmented, noisy defect mask, only the largest connected component in the overlapping region is retained as the true outline of the bubble in the output.
[0051] Figure 2 This is a schematic diagram of the structure of a screen with a protective film. Figure 2 The blue (lower layer) is the screen body, and the green (upper layer) is the protective film. In the implementation of this invention, the first and second images are acquired using the same imaging device, and are acquired sequentially under time-division strobe control. During the acquisition of the first image, the screen is illuminated to a white field, and the backlight transmits through the screen and the protective film. Due to Fresnel reflection and thin-film interference, concentric interference fringes are formed in the bubble area. At the same time, due to the halo effect, the imaging area is larger than the actual physical size, and the bubble appears as a bright spot with concentric circular fringes, and the area is magnified (larger than the actual bubble). Figure 3 As shown. During the second image acquisition, the screen is off, and the external low-angle ring light source is lit. The curved surface of the bubble wall produces local specular reflection, which appears as a single bright reflective spot without concentric circle features, and the image size is closer to the real size, as shown. Figure 7 As shown, by comparing and verifying these two modes, the magnified size of the first image is corrected by using the true size of the second image, thus reducing false detections. Fusion judgment can significantly improve detection accuracy.
[0052] Example 2:
[0053] This embodiment, based on Embodiment 1, is further designed in that the first preset area threshold is greater than the second preset area threshold to accommodate the difference in optical characteristics between the bubble's larger image size due to halo effect when the screen is lit and its closer to its true size when illuminated by an external light source. When the screen is lit, due to the bubble's optical characteristics, light is reflected at the top and bottom surfaces of the bubble, and the reflected light interferes with each other. This reflection and scattering causes light energy to diffuse in space, resulting in an image area larger than the actual physical size of the bubble. When the screen is off, the bubble wall is curved, reflecting external light sources. There is no backlight transmission, no thin-film interference, and no halo scattering effect, making the size close to the true physical size of the bubble.
[0054] Example 3:
[0055] This embodiment is further designed based on embodiment two, in which the first preset area threshold is set to be greater than 5 pixels and the second preset area threshold is set to be greater than 3 pixels, in order to adapt to the difference in optical characteristics of the bubble, which appears larger due to the halo effect when the screen is lit and appears closer to its real size when illuminated by an external light source.
[0056] Example 4:
[0057] This embodiment, based on Embodiment 1, further designs the following: morphological dilation and intersection operations are performed on the bubble candidate regions to connect adjacent regions while maintaining the original area of each bubble candidate region. Specifically, this includes:
[0058] Morphological dilation is performed on the candidate bubble region to obtain the dilated region. Connectivity component analysis is then performed on the dilated region to split it into multiple independent sub-regions. The intersection region between each sub-region and the candidate bubble region before dilation is then calculated. The intersection region is the complete bubble region.
[0059] The following example illustrates the method in detail:
[0060] 1) Mathematical representation of a binary region:
[0061] Let the image be a discrete two-dimensional grid E={(x,y)|0≤x,0≤y}⊂Z², representing the set of all integer coordinate points (x,y), such as (0,0), (1,-1), (5,100), etc. The bubble candidate regions obtained through multi-peak automatic segmentation are the binary foreground point set R⊆E. If a pixel (x,y)∈R, then that point is foreground; otherwise, it is background.
[0062] 2) Definition of neighborhood relations and connectivity:
[0063] In connected component analysis, the adjacency relationship between two points can be defined using 4-connectivity or 8-connectivity:
[0064] When two points p=(x1,y1) and q=(x2,y2) are 4-adjacent, it is necessary to satisfy ||pq||1=|x1−x2|+|y1−y2|=1;
[0065] When two points p=(x1,y1) and q=(x2,y2) are 8-adjacent, they must satisfy max(|x1−x2|,|y1−y2|)=1.
[0066] 3) Connected paths and connected components:
[0067] For a region R, if there exists a sequence of points p1, p2, ..., p n If any two points ∈ R such that their adjacent points satisfy a given adjacency relation (4-adjacent or 8-adjacent), then this sequence of points forms a connected path. A subset C⊆R of region R is called a connected component if and only if:
[0068] (i) There exists a connected path between any two points in C that lies entirely within C;
[0069] (ii) C is a maximal subset, meaning that no other point in R can be added while still satisfying condition (i).
[0070] Therefore, region R can be decomposed into a finite number of disjoint connected components {C1,C2,...,C...} k},satisfy or ;
[0071] This decomposition process can be formally represented as a connected component decomposition operator: Connect(R,η)={C1,C2,...,C k}, where η∈{4,8} is the selected connectivity criterion.
[0072] 4) Decomposition and intersection operations of connected components in the expanded region:
[0073] Let the original bubble candidate region be S1⊆E.
[0074] Morphological dilation is performed on the bubble candidate region S1 (the dilation structuring element is a circular kernel with a radius of 1 pixel), resulting in the dilated region S2. The above connected component decomposition is then performed on S2, using the 8-connectivity criterion, to obtain a set of independent sub-regions Connect(S2,8)={C1,C2,...,C...}. m For each sub-region Cᵢ, calculate its intersection with the pre-expansion bubble candidate region S1, Cᵢ'=Cᵢ∩S1. The final complete bubble region is: ;
[0075] 5) Mathematical definition of intersection operation:
[0076] In a digital image, let regions A and B be sets of binary foreground pixels, and their intersection be defined as C = A ∩ B = {p ∈ R² | p ∈ A and p ∈ B}. In the discrete pixel domain, the intersection operation is equivalent to a pixel-wise logical AND operation: f C (x,y)=f A (x,y)∧f B (x,y), where C=A∩B, and f is the corresponding binary image function. For solving the intersection of arbitrary polygonal regions, it can be achieved by calculating the intersection points and constructing the intersection polygon. However, this method directly uses pixel-by-pixel logical operations in the discrete pixel domain, which is more efficient and meets engineering requirements.
[0077] Example 5:
[0078] This embodiment, based on Embodiment 1, is further designed in that the acquisition time interval between the first and second images is a time-division strobe interval, which is no more than 100 microseconds. At this time scale, the relative displacement between the screen under test and the imaging device caused by the mechanical vibration of the production line is less than 0.1 pixels, which is visually equivalent to complete stillness. This ensures that the spatial position of the same bubble area in the two images is accurately aligned, and the overlap ratio can be calculated directly without additional image registration steps. The sub-pixel level displacement does not affect the subsequent overlap ratio calculation because the overlapping area is a pixel-level binary mask.
[0079] Example 6:
[0080] This embodiment, based on Embodiment 1, further incorporates frequency domain processing based on Fast Fourier Transform and Gaussian high-pass filtering, including:
[0081] The original image is centered, and then a fast Fourier transform is performed to obtain the frequency domain image.
[0082] Multiply the frequency domain image by the transfer function of a Gaussian high-pass filter to suppress low-frequency components and preserve high-frequency components;
[0083] Perform an inverse fast Fourier transform on the filtered frequency domain image, then perform centering and take the real part to obtain the spatial domain image.
[0084] The complete algorithm flow for frequency domain processing based on Fast Fourier Transform and Gaussian high-pass filtering is as follows:
[0085] 1) Centering: Let the original image be f(x,y) with size M×N. First, perform centering processing and calculate f. c (x,y)=f(x,y)·(-1)^(x+y), which shifts the low-frequency components to the center of the frequency domain.
[0086] 2) Fourier Transform: For f c Perform a two-dimensional fast Fourier transform on (x,y) to obtain the frequency domain image F(u,v).
[0087] 3) Frequency domain filtering: Multiply F(u,v) point by point with the transfer function H(u,v) of the Gaussian high-pass filter to obtain the filtered frequency domain image G(u,v)=H(u,v)·F(u,v).
[0088] 4) Inverse Fourier Transform: Perform an inverse fast Fourier transform on G(u,v) to obtain g'(x,y). At this time, g'(x,y) should be a real matrix, but a small imaginary part may be introduced in the calculation.
[0089] 5) Decentralization and taking the real part: Multiply g′(x,y) by (-1)^(x+y) again to restore the original frequency domain position, and then take the real part of the result to obtain the final spatial domain sharpened image g(x,y): g(x,y)=Real[g′(x,y)·(-1)^(x+y)];
[0090] The sharpening intensity can be controlled by adjusting the cutoff frequency D0, such as... Figure 5 and Figure 8 As shown, the smaller D0 is, the stronger the high-pass filtering and the more prominent the bubble edges; the larger D0 is, the closer the result is to the original image. In this embodiment, the value of D0 ranges from 10 to 30.
[0091] The aforementioned cutoff frequency D0 range (10–30) was experimentally determined based on image size (e.g., M=N=1024) and the pixel width of the bubble target (typically 5–20 pixels). The D0 value determines the strength of the high-pass filter: a smaller D0 (e.g., close to 10) suppresses larger low-frequency regions in the frequency domain, significantly weakens the smooth background in the image, and extremely sharpens high-frequency details such as bubble edges, making it suitable for enhancing weak small bubble edges. A larger D0 (e.g., close to 30) only suppresses extremely low-frequency DC components, preserving more of the original grayscale information of the image, making it suitable for enhancing large bubbles with obvious edges and avoiding excessive background loss.
[0092] Example 7:
[0093] This embodiment, based on Embodiment Six, is further designed in that the transfer function of the Gaussian high-pass filter in this example is: ;in, It is the midpoint of the frequency domain Distance to the center of the frequency rectangle M and N represent the width and height of the image, respectively. and These are the frequencies in the horizontal and vertical directions, respectively. It is the cutoff frequency.
[0094] Example 8:
[0095] This embodiment, based on Embodiment 1, is further designed in that the automatic multi-peak segmentation in this example includes:
[0096] The grayscale histogram of the spatial domain image is obtained, and the grayscale histogram is smoothed to eliminate noise interference. The local minimum value in the smoothed histogram is searched as the segmentation threshold, and the image is segmented into multiple bubble candidate regions using the segmentation threshold.
[0097] The grayscale histogram Hσ(g) was smoothed using Gaussian smoothing, and the smoothed histogram is as follows: Where g is the gray level, ranging from 0 to 255; σ is the standard deviation of the Gaussian kernel, controlling the smoothness. In this embodiment, σ is set to a value between 1.0 and 2.0. When σ is small (e.g., 1.0), more details of the histogram are preserved, suitable for images with complex gray-level distributions and weak differences between bubbles and the background; when σ is large (e.g., 2.0), the image is smoother, retaining only the main peaks and troughs of the histogram, suitable for images with high noise or uniform regions. A threshold is used to segment the spatial domain image into multiple binary regions, serving as candidate bubble regions. See the example of bubble candidate region detection results in dot-screen mode. Figure 6 For an example of bubble candidate region detection results in the lighting mode, please refer to [link / reference]. Figure 9 .
[0098] Example 9:
[0099] This embodiment is a further design based on Embodiment 1, in which the preset ratio threshold is 0.4.
[0100] Example 10:
[0101] The present invention provides a screen surface bubble detection system, comprising a light source control module, an imaging module, and an image processing module;
[0102] The light source control module is used to output a screen lighting control signal to the screen under test and to output a strobe control signal to an external light source.
[0103] The imaging module is synchronized with the light source control module in time and is used to respond to the flicker control signal to acquire the first image when the screen is lit and the second image when the screen is off and the external light source is illuminating the screen.
[0104] The image processing module, which is communicatively connected to the imaging module, is used to receive the first image and the second image, and to execute the screen surface bubble detection method as described in any of the above embodiments to determine bubble defects.
[0105] Example 11:
[0106] An electronic device includes a memory and a processor. The memory stores a computer program, and the processor is used to invoke and run the computer program stored in the memory to perform the methods of any of the above embodiments.
[0107] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above embodiments.
[0108] Application Examples:
[0109] This example uses the method and system of this invention to verify the performance of real screen samples in a mass production scenario. The total number of test samples was 2000 pieces, including 960 screens with actual bubble defects and 1040 normal good screens. After precise manual labeling, the total number of bubble defects was 1423. The bubble types covered common production defect scenarios: 612 micro bubbles with a diameter of 50-100μm (approximately 3-5 pixels in imaging), accounting for 43.0%; 647 medium bubbles with a diameter of 100-200μm (approximately 5-10 pixels in imaging), accounting for 45.5%; and 164 large bubbles with a diameter of 200-350μm (approximately 10 pixels in imaging), accounting for 11.5%. The distribution of bubble count per screen was as follows: 721 screens with a single bubble, 208 screens with 2 to 3 bubbles, and 31 screens with 4 or more dense bubbles. In terms of location distribution, bubbles accounted for 62% in the central display area, 28% in the edge areas prone to missed detection, and 10% in the four corner blind areas, comprehensively covering areas difficult to detect.
[0110] The test hardware configuration is as follows: The camera uses a 12-megapixel global shutter CMOS sensor with a 2 / 3-inch sensor surface and a resolution of 4096×3000; the lens is a 16mm fixed-focus industrial lens; the light source uses a combination of time-division strobe backlight and a ring low-angle light source; the main control platform is an Intel i7-10700 (8-core) processor with 8GB RAM, and the algorithm is optimized for CPU parallelism, without a dedicated GPU.
[0111] Some test results are as follows Figures 3-9 As shown, the test results indicate that the method of this invention detected all 1423 manually labeled bubbles, achieving a detection rate of 100%, with a false negative number (FN) of 0. Based on testing with 1040 good screens, the false positive number (FP) was 17, with a false positive rate (FPR) of approximately 1.63% (17 / 1040). After secondary filtering based on area and shape features, the false positive rate can be further reduced to below 0.3%. The average processing time per screen is 46.8 ms, meeting the mass production testing requirement of ≤50 ms. ROC curve analysis of the dual-mode fusion shows that the system AUC reaches 0.998, and the false positive rate is only 0.31% at the operating point maintaining a 100% detection rate, significantly better than the single white dot screen mode (AUC=0.996, FPR=1.12%) and the single external lighting mode (AUC=0.987, FPR=0.87%).
[0112] In summary, the screen surface bubble detection method and system provided by this invention achieves high-precision, low-false-detection, and fast-cycle detection of bubbles under the screen surface protective film through dual-mode time-division stroboscopic imaging, parallel frequency domain high-pass filtering, area-preserving connected domain merging, and dual-path overlap ratio fusion judgment. The test data is authentic and valid, meeting the actual mass production testing requirements.
[0113] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for detecting bubbles on a screen surface, characterized in that, include: For the effective display areas of the first image captured when the screen is on and the second image captured when the screen is off and illuminated by an external light source, frequency domain processing based on fast Fourier transform and Gaussian high-pass filtering is performed to enhance the high-frequency components of the bubble edge, resulting in the first spatial domain image and the second spatial domain image. Multi-peak automatic segmentation is performed on the first spatial domain image and the second spatial domain image respectively to obtain bubble candidate regions. Morphological dilation and intersection operations are performed on the bubble candidate regions to connect neighboring regions while keeping the original area of each bubble candidate region unchanged, so as to form a complete bubble region. A first candidate bubble region set is selected from the complete bubble region corresponding to the first spatial domain image based on a first preset area threshold, and a second candidate bubble region set is selected from the complete bubble region corresponding to the second spatial domain image based on a second preset area threshold; When the overlap ratio of any overlapping region in the first candidate bubble region set and the second candidate bubble region set is greater than a preset ratio threshold, the overlapping region is determined to be a bubble defect. The overlap ratio is the ratio of the intersection area to the union area of the overlapping regions.
2. The method for detecting bubbles on the screen surface according to claim 1, characterized in that, The first preset area threshold is greater than the second preset area threshold.
3. The method for detecting bubbles on the screen surface according to claim 2, characterized in that, The first preset area threshold is set to be greater than 5 pixels, and the second preset area threshold is set to be greater than 3 pixels.
4. The method for detecting bubbles on the screen surface according to claim 1, characterized in that, The step of performing morphological dilation and intersection operations on the bubble candidate regions to connect neighboring regions while maintaining the original area of each bubble candidate region, specifically includes: A morphological dilation operation is performed on the candidate bubble region to obtain the dilated region. Connectivity component analysis is then performed on the dilated region to split it into multiple independent sub-regions. The intersection region between each sub-region and the candidate bubble region before dilation is then calculated. The intersection region is the complete bubble region.
5. The method for detecting bubbles on the screen surface according to claim 1, characterized in that, The time interval between acquiring the first image and the second image is a time-division strobe interval, which is no more than 100 microseconds.
6. The method for detecting bubbles on the screen surface according to claim 1, characterized in that, The frequency domain processing based on Fast Fourier Transform and Gaussian high-pass filtering includes: The original image is centered, and then a fast Fourier transform is performed to obtain the frequency domain image. The frequency domain image is multiplied by the transfer function of a Gaussian high-pass filter to suppress low-frequency components and preserve high-frequency components. The filtered frequency domain image is subjected to an inverse fast Fourier transform, then centered and the real part is taken to obtain the spatial domain image.
7. The method for detecting bubbles on the screen surface according to claim 6, characterized in that, The transfer function of the Gaussian high-pass filter is: ； In the formula, It is a frequency domain point Distance to the center of the frequency rectangle; and These are the frequencies in the horizontal and vertical directions, respectively. It is the cutoff frequency.
8. The method for detecting bubbles on the screen surface according to claim 1, characterized in that, The automatic multi-peak segmentation includes: The grayscale histogram of the spatial domain image is obtained, and the grayscale histogram is smoothed to eliminate noise interference. The local minimum value in the smoothed histogram is searched as the segmentation threshold, and the image is segmented into multiple bubble candidate regions using the segmentation threshold.
9. The method for detecting bubbles on the screen surface according to claim 1, characterized in that, The preset ratio threshold is 0.
4.
10. A screen surface bubble detection system, characterized in that, It includes a light source control module, an imaging module, and an image processing module; The light source control module is used to output a screen lighting control signal to the screen under test and to output a strobe control signal to an external light source; The imaging module is time-synchronized with the light source control module and is used to acquire a first image when the screen is lit and a second image when the screen is off and the external light source is illuminating, in response to the flicker control signal. The image processing module is communicatively connected to the imaging module and is used to receive the first image and the second image, and to execute the screen surface bubble detection method as described in any one of claims 1 to 9 to determine bubble defects.