Notch detection method and system for circular domestic ceramic device under complex working conditions

Abstract

The invention discloses a notch detection method and system for a circular domestic ceramic device under a complex working condition. The method comprises the following steps: acquiring original image data of the circular domestic ceramic device; preprocessing the original image data; extracting the outer boundary of the circular daily ceramic device from the preprocessed image data; extracting a plurality of complementary boundary morphological features, wherein the features comprise a local concavity coefficient, a discrete curvature, an angle gradient and a radial deviation; enhancing the extracted boundary morphological features through a feature enhancement strategy combining local contrast enhancement and nonlinear transformation; carrying out hierarchical fusion on the enhanced boundary morphological features; and gap detection is carried out based on the enhanced fusion features. According to the method, the four-dimensional morphological feature space containing the local concavity coefficient, the discrete curvature gradient, the angle change rate and the radial deviation is constructed, and the nonlinear feature enhancement and hierarchical feature fusion strategy is combined, so that efficient detection of circular and non-circular ceramic notches can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for detecting notches of circular daily-use ceramic components under complex working conditions. Background Art

[0002] Product quality control is a critical link in the field of modern ceramic precision manufacturing. Ceramic edge integrity testing, as a core quality control point, directly determines the functional reliability and safety of the product. According to the GB / T 3303-2018 standard, "Terminology of Defects in Daily-Use Ceramics," combined with industry statistics, there are more than ten types of defects that may occur in daily-use ceramic products during the production process, primarily including spots, cracks, slag, breakage, glaze defects, and deformation. These defects not only seriously affect the product's appearance quality, but also significantly reduce its mechanical strength and service life, and even pose a potential threat to user safety. Traditional manual inspection methods have limitations such as strong subjectivity, poor consistency, and low efficiency, making it difficult to meet the high-precision requirements of modern ceramic manufacturing.

[0003] In recent years, with the rapid development of computer vision and artificial intelligence technologies, intelligent defect detection methods based on computer vision and deep learning have become a research hotspot and a key area of technological innovation in the ceramic manufacturing industry due to their efficiency, accuracy, and repeatability. These methods offer new solutions for intelligent and automated quality control of ceramic products. This AI-based approach uses convolutional neural networks to learn features and train models for ceramic defect images, effectively enabling automated defect recognition and detection in industrial scenarios. Specifically, "Sun P, Hua C, Ding W, et al. Ceramic tableware surface defect detection based on deep learning [J]. Engineering Applications of Artificial Intelligence, 2025, 141: 109723" systematically improves the YOLOv5s architecture by introducing an attention mechanism module to enhance feature selection capabilities, adding a small-scale detection layer to improve fine-grained feature capture, and employing depthwise separable convolution to construct a lightweight tile detection system. This significantly improves the problem of insufficient feature information for small object defects. "Sun P, Hua C, Ding W, et al. Ceramic tableware surface defect detection based on deep learning [J]. Engineering Applications of Artificial Intelligence, 2025, 141: 109723." developed a ceramic tableware surface defect detection system based on the YOLOv8 framework. By optimizing the feature pyramid structure and loss function design, it achieved a high defect detection accuracy. "MIN B, TIN H, NASRIDINOV A, et al. Abnormal detection and classification in i-ceramic images [C] / / 2020 IEEE international conference on big data and smart computing (BigComp). IEEE, Busan, 2020: 17–18." innovatively adopted a transfer learning strategy and used a pre-trained convolutional neural network to build a classification model, achieving good results in ceramic crack detection tasks."Cao T, Song K, Xu L, et al. Balanced multi-scale target score network for ceramic tile surface defect detection [J]. Measurement, 2024, 224: 113914." In order to solve the problem of heterogeneity in the distribution of ceramic defects, a content-aware feature reorganization method and a dynamic attention mechanism are proposed. By establishing a feature importance weight evaluation system, the feature information loss of the YOLOv5s algorithm in complex defect scenarios is effectively reduced. The improved model achieves a significant improvement in the mAP index compared with the baseline version. In response to various defects on the surface of ceramic tiles, "Wang Jianguo, Sun Fuzhong, Yuan Zilong, et al. Ceramic tile surface defect detection algorithm based on improved Faster RCNN [J]. Journal of Nanjing University of Technology (Natural Science Edition), 2025, 47 (01): 49-55." An improved fast convolutional neural network algorithm based on machine vision is proposed, and the detection accuracy of the model is further improved. "Zhu Yonghong, Wu Songtao. Ceramic tile defect detection and classification method based on improved YOLOv8n[J]. Journal of Ceramics, 2024, 45(06):1255-1264." A ceramic tile defect detection and classification method based on YOLOv8n with dynamic snake convolution and attention mechanism is proposed. The addition of the attention mechanism module in this model effectively enhances the model's feature extraction ability for small objects. Although ceramic defect detection methods based on deep learning are efficient and accurate, they still face core challenges such as strong data dependence, weak model generalization ability, and high deployment barriers.

[0004] The main idea of computer vision-based methods is to design feature representations for different ceramic defects, thereby achieving customized ceramic defect detection. "Wang Junxiang, Peng Huacang, Hu Honghao, et al. Research on a rapid ceramic roundness detection system based on computer vision [J]. Journal of Ceramics, 2015, 36(5): 530-535." proposed a research scheme for an online detection system for daily-use ceramics based on computer vision technology. This scheme can give corresponding roundness judgments through processes such as extracting the outer boundary of the ceramic, initially locating the center of the circle, and analyzing the stability of the center of the circle. On this basis, "Peng Huacang. Research on appearance defect detection technology of circular daily-use ceramics based on digital image processing technology [D]. Jingdezhen: Jingdezhen Ceramic Institute, 2015." et al. designed a typical notch detection system for circular daily-use ceramics, which has certain application scenarios in the industrial field. "Wang J, Liu Y, Zhang D, et al. A new computer vision based multi-indentation inspection system for ceramics [J]. Multimedia Tools and Applications, 2017, 76: 2495-2513." This paper further studies the characterization of the internal characteristics of daily ceramic gaps. When the roundness of the ceramics meets the national standard, the gap position is analyzed and located by the radius residual and radius residual index of the circular ceramic outer boundary. "Zhao Yi, Li Yuxin, Rao Zhi, et al. Design of a machine vision ceramic inspection system based on cloud computing [J]. Automation Applications, 2020 (7): 72–74." In response to the problem of low detection efficiency of the method in the literature "Wang J, Liu Y, Zhang D, et al. A new computer vision based multi-indentation inspection system for ceramics [J]. Multimedia Tools and Applications, 2017, 76: 2495-2513.", a ceramic gap detection system based on cloud computing is designed by innovatively introducing MySQL database, which greatly reduces the system maintenance cost. This type of detection algorithm can effectively detect circular ceramic gaps, but in real scenes, factors such as lens distortion, improper placement, uneven lighting, and motion blur can cause the roundness of circular ceramics to be distorted, which in turn makes the detection of non-circular ceramic gaps ineffective. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method for detecting notches in circular daily-use ceramic components under complex working conditions, which can achieve high-precision detection of circular and non-circular notches on the edges of circular daily-use ceramic components.

[0006] To achieve the above objectives, the technical solutions provided by the present invention are:

[0007] A method for detecting notches in circular daily-use ceramic components under complex working conditions comprises:

[0008] Collect image data of circular daily-use ceramic components;

[0009] Extracting the outer boundary of circular daily-use ceramic components from image data;

[0010] Extracting multiple boundary morphological features, including local concavity coefficient, discrete curvature, angular gradient and radial deviation;

[0011] The extracted boundary morphological features are enhanced through a feature enhancement strategy combining local contrast enhancement and nonlinear transformation;

[0012] Perform hierarchical fusion on the enhanced boundary morphological features;

[0013] Gap detection is performed based on the enhanced fusion features.

[0014] This technical solution can achieve high-precision detection of circular and non-circular notches on the edges of circular daily-use ceramic components by extracting multiple complementary boundary morphological features and combining nonlinear feature enhancement with hierarchical feature fusion strategies.

[0015] Furthermore, the local concavity coefficient describes the degree of concavity or convexity of the boundary point relative to the local line segment and is defined as follows:

[0016]

[0017] In formula (9), p i represents the boundary point, L i-w,i+w Represents point p i-w and p i+w The line segments formed, Represents point p i To line segment L i-w,i+w distance, For line segment L i-w,i+w The unit normal vector of From the boundary point p i The vector pointing to the center point c of the ceramic edge; negative values indicate inward concavity, positive values indicate outward convexity;

[0018] The discrete curvature describes the curvature of the boundary and is calculated as follows:

[0019]

[0020] In formula (10), R i is the boundary point pi The local circle fitting radius obtained by least squares fitting within the neighborhood;

[0021] The angular gradient describes the degree of boundary turning, which is defined as the rate of change of the angle between adjacent boundary vectors:

[0022]

[0023] In formula (17), Represents the boundary vector; further calculate the local rate of change of the angle:

[0024]

[0025] The radial deviation describes the degree of deviation from the boundary point to the standard circle and directly reflects the roundness anomaly:

[0026]

[0027] In formula (19), is the distance from the boundary point to the center of the circle, is the average radius; in order to enhance the ability to distinguish radial deviations, the standardized radial deviation is introduced:

[0028]

[0029] In formula (20), is the standard deviation of the radial distance; N is the number of boundary sampling points.

[0030] Furthermore, the local circle fitting radius R i The calculation process includes:

[0031] Let the local coordinate set Among them, u and v are the coordinates of the boundary points, u represents the horizontal coordinate, and v represents the vertical coordinate; w is the window size, which indicates how many points are taken around the current point i for fitting; j is the summation variable, and the mean value of the points within the range from iw to i+w is Then the boundary coordinate fluctuation is expressed as: Calculate the matrix elements required for fitting according to equations (11), (12), and (13):

[0032]

[0033] Construct the system of equations and solve it:

[0034]

[0035] The matrix is invertible, so:

[0036]

[0037] The final local circle fitting radius is:

[0038]

[0039] In equations (15) and (16), (a, b) represents the coordinates of the center of the fitting circle.

[0040] Furthermore, the extracted boundary morphological features are enhanced through a feature enhancement strategy combining local contrast enhancement and nonlinear transformation, including:

[0041] Local contrast enhancement: Adaptive background estimation method is used to enhance the contrast between features and background, as shown in formula (21):

[0042]

[0043] In formula (21), is the maximum value of the feature in the local window, and are the mean and standard deviation of the features in the large window respectively; ∈ is a small positive number to prevent the denominator from being zero; w l are the window radius respectively;

[0044] Nonlinear transformation: Apply the power function to enhance significant features and suppress subtle changes, as shown in formula (22):

[0045] F nonlinear (i) = sgn(F enhanced (i))·|F enhanced (i)| α (twenty two)

[0046] In formula (22), sgn(·) is the sign function, and α>1 is the nonlinear coefficient used to enhance the salient features.

[0047] Furthermore, the enhanced boundary morphological features are hierarchically fused, including two levels: base layer fusion and spatial consistency constraint;

[0048] Among them, the basic layer fusion is to combine the features in a weighted manner, and the weight is determined based on the importance of the features, as shown in formula (23):

[0049]

[0050] In formula (23), w j is the weight coefficient of each feature, satisfying

[0051] Spatial consistency constraint: Apply weighted median filtering to enhance the spatial consistency of features, as shown in formula (24):

[0052] F filtered(i) = median{w(|ji|)·F fusion (j)|j∈[iw m ,i+w m ]}(twenty four)

[0053] In formula (24), w(d) is the distance weight function, w m is the filter window radius.

[0054] Furthermore, based on the enhanced fusion features, adaptive threshold and spatial clustering analysis are used for gap detection, including:

[0055] Adaptive threshold: Determine the optimal threshold based on the feature distribution, as shown in formula (25):

[0056]

[0057] In formula (25), and are the mean and standard deviation of the enhanced fusion features, respectively, and η is the adjustment coefficient;

[0058] Spatial clustering: Perform spatial connectivity analysis on points exceeding the threshold, as shown in formula (26):

[0059]

[0060] In formula (26), represents the kth cluster, satisfying:

[0061]

[0062] In formula (27), F filtered (i) is the enhanced fusion feature;

[0063] Gap assessment and location: Evaluate the significance of each cluster, as shown in formula (28):

[0064]

[0065] In formula (28), |C k | is cluster C k The size of σ({r i |i∈C k}) represents the standard deviation of the radius variation within the cluster; then, the cluster with the highest significance is selected as the gap position, as shown in formula (29):

[0066]

[0067] Furthermore, performing gap detection also includes:

[0068] Calculate the geometric features of the detected gap based on the gap position:

[0069]

[0070] In formulas (30) and (31), L defect With W defect Respectively represent the length and depth of the notch.

[0071] Furthermore, the collected image data of the circular daily-use ceramic device is original image data, and the original image data is pre-processed before the outer boundary of the circular daily-use ceramic device is extracted;

[0072] Preprocess the raw image data, including:

[0073] The weighted average method is used to convert the RGB image into a grayscale image, and then the Otsu algorithm is used to adaptively calculate the optimal threshold T opt ; The Otsu method is based on the principle of maximizing the between-class variance, and its objective function is:

[0074]

[0075] In formula (1), is the inter-class variance, ω0(T) and ω1(T) are the proportions of the two types of pixels in the entire image after the threshold T segmentation, μ0(T) and μ1(T) are the average grayscale values of the two types of pixels. After obtaining the optimal threshold, binarization is performed according to formula (2):

[0076]

[0077] In order to eliminate noise and maintain boundary integrity, median filtering and morphological opening and closing operations are used, as shown in Equations (3), (4), and (5):

[0078]

[0079] Where w represents the filter window size, S1 and S2 represent the structural elements of morphological opening and closing operations respectively. and represent the erosion and dilation operations respectively.

[0080] Furthermore, a boundary tracking algorithm based on connected component labeling is used to extract the outer boundary of the circular daily-use ceramic component. The connected component labeling algorithm is based on depth-first search and is defined as:

[0081]

[0082] In formula (6), L(x,y) is the label value of pixel (x,y), Ω lis the l∈{1,2,...,N}th connected region; the boundary is extracted according to the connected region, as shown in formula (7):

[0083]

[0084] In formula (7), B l is the boundary of the lth connected region, N8(x,y) represents the 8-neighborhood of pixel (x,y); the longest boundary is selected as the ceramic edge contour:

[0085] B bowl =argmax l |B l |(8)

[0086] In formula (8), |B l | indicates the boundary B l The number of points included.

[0087] Furthermore, to achieve the above-mentioned purpose, the present invention further provides a notch detection system for circular daily-use ceramic components under complex working conditions, which is used to implement the notch detection method for circular daily-use ceramic components under complex working conditions, and includes an image acquisition module, an image preprocessing module, a ceramic boundary extraction module, a feature extraction module, a feature enhancement module, a feature fusion module, and a notch detection module;

[0088] in,

[0089] The image acquisition module is used to collect original image data of the circular daily-use ceramic device;

[0090] The image preprocessing module is used to preprocess the original image data;

[0091] The ceramic boundary extraction module is used to extract the outer boundary of the circular daily-use ceramic component from the pre-processed image data;

[0092] The feature extraction module is used to extract multiple complementary boundary morphological features;

[0093] The feature enhancement module is used to enhance the extracted boundary morphological features;

[0094] The feature fusion module is used to perform hierarchical fusion on the enhanced boundary morphological features;

[0095] The gap detection module detects gaps based on the enhanced fusion features.

[0096] Compared with the existing technology, the principles and advantages of this technical solution are as follows:

[0097] 1. This technical solution constructs a four-dimensional morphological feature space encompassing the local concavity coefficient, discrete curvature gradient, angular rate of change, and radial deviation. Combining nonlinear feature enhancement with a hierarchical feature fusion strategy, it enables efficient detection and precise location of circular and non-circular ceramic gaps. Experimental results demonstrate that this technical solution achieves a detection accuracy of over 98% under complex backgrounds. The system also processes approximately two frames of image per second in real time, significantly improving the precision and efficiency of ceramic product quality control. This provides a reliable theoretical basis and technical support for automated online inspection in ceramic manufacturing, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the services required for use in the embodiments or the prior art descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0099] Figure 1 This is a connection block diagram of a notch detection system for circular daily-use ceramic components under complex working conditions according to an embodiment of the present invention;

[0100] Figure 2 This is a principle flow chart of a method for detecting notches in circular daily-use ceramic components under complex working conditions according to an embodiment of the present invention;

[0101] Figure 3 This is a schematic diagram of the outer boundary extraction and gap location of a circular daily-use ceramic device image under normal conditions;

[0102] Figure 4 Schematic diagram of outer boundary extraction and gap location of circular daily-use ceramic component images under complex working conditions. DETAILED DESCRIPTION

[0103] The present invention will be further described below in conjunction with specific embodiments:

[0104] like Figure 1 As shown, the present embodiment of the present invention provides a notch detection system for circular daily-use ceramic components under complex working conditions, comprising an image acquisition module, an image preprocessing module, a ceramic boundary extraction module, a feature extraction module, a feature enhancement module, a feature fusion module, and a notch detection module.

[0105] in,

[0106] The image acquisition module is used to collect original image data of the circular daily-use ceramic device;

[0107] The image preprocessing module is used to preprocess the original image data;

[0108] The ceramic boundary extraction module is used to extract the outer boundary of the circular daily-use ceramic component from the pre-processed image data;

[0109] The feature extraction module is used to extract multiple complementary boundary morphological features;

[0110] The feature enhancement module is used to enhance the extracted boundary morphological features;

[0111] The feature fusion module is used to perform hierarchical fusion on the enhanced boundary morphological features;

[0112] The gap detection module detects gaps based on the enhanced fusion features.

[0113] like Figure 2 As shown in FIG, the working principle of the notch detection system for circular daily-use ceramic components under complex working conditions includes the following steps:

[0114] S1, using an image acquisition module to collect original image data of a circular daily-use ceramic device;

[0115] S2, preprocessing the original image data using an image preprocessing module;

[0116] The quality of the circular daily-use ceramic device image is the basis for subsequent gap feature extraction. This embodiment adopts an adaptive processing strategy, including grayscale conversion, adaptive threshold segmentation and morphological filtering.

[0117] First, the RGB image is converted into a grayscale image using the weighted average method. Then, the Otsu algorithm is applied to adaptively calculate the optimal threshold T opt ; The Otsu method is based on the principle of maximizing the between-class variance, and its objective function is:

[0118]

[0119] In formula (1), is the inter-class variance, ω0(T) and ω1(T) are the proportions of the two types of pixels in the entire image after the threshold T segmentation, μ0(T) and μ1(T) are the average grayscale values of the two types of pixels. After obtaining the optimal threshold, binarization is performed according to formula (2):

[0120]

[0121] In order to eliminate noise and maintain boundary integrity, median filtering and morphological opening and closing operations are used, as shown in Equations (3), (4), and (5):

[0122]

[0123] Where w represents the filter window size, S1 and S2 represent the structural elements of morphological opening and closing operations respectively. and represent the erosion and dilation operations respectively.

[0124] S3. Using the ceramic boundary extraction module, a boundary tracking algorithm based on connected region labeling is used to extract the outer boundary of the circular daily-use ceramic component from the preprocessed image data. The connected region labeling algorithm is based on a depth-first search and is defined as:

[0125]

[0126] In formula (6), L(x,y) is the label value of pixel (x,y), Ω l is the l∈{1,2,...,N}th connected region; the boundary is extracted according to the connected region, as shown in formula (7):

[0127]

[0128] In formula (7), B l is the boundary of the lth connected region, N8(x,y) represents the 8-neighborhood of pixel (x,y); the longest boundary is selected as the ceramic edge contour:

[0129] B bowl =argmax l |B l |(8)

[0130] In formula (8), |B l | indicates the boundary B l The number of points included.

[0131] S4. extracting a plurality of complementary boundary morphological features using a feature extraction module, wherein the boundary morphological features include a local concavity coefficient, a discrete curvature, an angular gradient, and a radial deviation;

[0132] The local concavity coefficient describes the degree of concavity or convexity of the boundary point relative to the local line segment. This is the key feature for detecting gaps and is defined as follows:

[0133]

[0134] In formula (9), p i represents the boundary point, L i-w,i+w Represents point p i-w and p i+w The line segments formed, Represents point p i To line segment L i-w,i+w distance, For line segment Li-w,i+w The unit normal vector of From the boundary point p i The vector pointing to the center point c of the ceramic edge; negative values indicate inward concavity, positive values indicate outward convexity;

[0135] The discrete curvature describes the curvature of the boundary and is calculated as follows:

[0136]

[0137] In formula (10), R i is the boundary point p i The local circle fitting radius obtained by least squares fitting within the neighborhood;

[0138] Local circle fitting radius R i The calculation process includes:

[0139] Let the local coordinate set Among them, u and v are the coordinates of the boundary points, u represents the horizontal coordinate, and v represents the vertical coordinate; w is the window size, which indicates how many points are taken around the current point i for fitting; j is the summation variable, and the mean value of the points within the range from iw to i+w is Then the boundary coordinate fluctuation is expressed as: Calculate the matrix elements required for fitting according to equations (11), (12), and (13):

[0140]

[0141] Construct the system of equations and solve it:

[0142]

[0143] The matrix is invertible, so:

[0144]

[0145] The final local circle fitting radius is:

[0146]

[0147] In equations (15) and (16), (a, b) represents the coordinates of the center of the fitting circle.

[0148] The angular gradient describes the degree of boundary turning, which is defined as the rate of change of the angle between adjacent boundary vectors:

[0149]

[0150] In formula (17), Represents the boundary vector; further calculate the local rate of change of the angle:

[0151]

[0152] The radial deviation describes the degree of deviation from the boundary point to the standard circle and directly reflects the roundness anomaly:

[0153]

[0154] In formula (19), is the distance from the boundary point to the center of the circle, is the average radius; in order to enhance the ability to distinguish radial deviations, the standardized radial deviation is introduced:

[0155]

[0156] In formula (20), is the standard deviation of the radial distance; N is the number of boundary sampling points.

[0157] S5. To improve the discriminative ability of features, the feature enhancement module enhances the extracted boundary morphological features through a feature enhancement strategy that combines local contrast enhancement and nonlinear transformation. Specifically, the following steps are performed:

[0158] Local contrast enhancement: Adaptive background estimation method is used to enhance the contrast between features and background, as shown in formula (21):

[0159]

[0160] In formula (21), is the maximum value of the feature in the local window, and are the mean and standard deviation of the features in the large window respectively; ∈ is a small positive number to prevent the denominator from being zero; w l are the window radius respectively;

[0161] Nonlinear transformation: Apply the power function to enhance significant features and suppress subtle changes, as shown in formula (22):

[0162] F nonlinear (i) = sgn(F enhanced (i))·|F enhanced (i)| α (twenty two)

[0163] In formula (22), sgn(·) is the sign function and α>1 is the nonlinear coefficient used to enhance the salient features.

[0164] S6. In order to integrate the advantages of each feature, the feature fusion module is used to perform hierarchical fusion of the enhanced boundary morphological features, including two levels: base layer fusion and spatial consistency constraint;

[0165] Among them, the basic layer fusion is to combine the features in a weighted manner, and the weight is determined based on the importance of the features, as shown in formula (23):

[0166]

[0167] In formula (23), w j is the weight coefficient of each feature, satisfying

[0168] Spatial consistency constraint: Apply weighted median filtering to enhance the spatial consistency of features, as shown in formula (24):

[0169] F filtered (i) = median{w(|ji|)·F fusion (j)|j∈[iw m ,i+w m ]}(twenty four)

[0170] In formula (24), w(d) is the distance weight function, w m is the filter window radius.

[0171] S7, the gap detection module uses adaptive threshold and spatial clustering analysis to perform gap detection based on the enhanced fusion features, including:

[0172] Adaptive threshold: Determine the optimal threshold based on the feature distribution, as shown in formula (25):

[0173]

[0174] In formula (25), and are the mean and standard deviation of the enhanced fusion features, respectively, and η is the adjustment coefficient;

[0175] Spatial clustering: Perform spatial connectivity analysis on points exceeding the threshold, as shown in formula (26):

[0176]

[0177] In formula (26), represents the kth cluster, satisfying:

[0178]

[0179] In formula (27), F filtered (i) is the enhanced fusion feature;

[0180] Gap assessment and location: Evaluate the significance of each cluster, as shown in formula (28):

[0181]

[0182] In formula (28), |C k | is cluster C k The size of σ({r i |i∈C k}) represents the standard deviation of the radius variation within the cluster; then, the cluster with the highest significance is selected as the gap position, as shown in formula (29):

[0183]

[0184] Calculate the geometric features of the detected gap based on the gap position:

[0185]

[0186] In formulas (30) and (31), L defect With W defect Respectively represent the length and depth of the notch.

[0187] To verify the effectiveness of the notch detection method proposed in this paper under complex working conditions such as lens distortion, misalignment, uneven lighting, and motion blur, an industrial camera was used to capture 100 images of circular household ceramic components. All of these images were used for notch detection. The experimental platform used Matlab 2016 as the software platform. This section provides two case studies under normal conditions and complex background conditions. Figure 3-4 shown.

[0188] based on Figure 3 and Figure 4 The experimental comparison shows that the method of the present invention shows the advantage of robustness under both normal and complex working conditions. Figure 3 ), four-dimensional morphological features (local concavity coefficient, discrete curvature gradient, angle gradient, radial deviation) are used to accurately locate the gap through hierarchical fusion; in scenes with lens distortion and light interference ( Figure 4 ), the local concavity coefficient and discrete curvature gradient effectively decouple deformation and notches. Spatial consistency constraints are combined to suppress noise, and adaptive threshold clustering enables stable notch detection in complex backgrounds. Compared to traditional notch detection methods that focus solely on circular ceramics, this method overcomes global geometric constraints through a multi-level feature complementation mechanism, providing a new paradigm for high-precision notch detection in complex industrial scenarios.

[0189] To further test the efficiency of the proposed inspection method, this experiment employed a similar experimental environment to that used in the paper "Wang J, Liu Y, Zhang D, et al. A new computer vision-based multi-indentation inspection system for ceramics [J]. Multimedia Tools and Applications, 2017, 76:2495-2513." The experimental data included 50 ceramic images taken under normal conditions and 50 images taken under complex conditions such as lens distortion, misalignment, uneven lighting, and motion blur. Inspection metrics included accuracy and average processing time per ceramic image. The results are shown in Table 1. The method of the present invention achieves detection accuracy of 100% and 98% under normal and complex backgrounds, respectively, which is a significant improvement over the literature "Wang J, Liu Y, Zhang D, et al. A new computer vision based multi-indentation inspection system for ceramics[J]. Multimedia Tools and Applications, 2017, 76: 2495-2513." with only a slight increase in processing time, verifying the strong robustness and real-time performance of the proposed method in complex scenes.

[0190] Table 1 Statistical results of the notch detection experiment

[0191]

[0192] The embodiments described above are only preferred embodiments of the present invention and are not intended to limit the scope of implementation of the present invention. Therefore, any changes made based on the shape and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting notches in circular daily-use ceramic components under complex working conditions, characterized in that: include: Collect image data of circular daily-use ceramic components; Extracting the outer boundary of circular daily-use ceramic components from image data; Extracting multiple boundary morphological features, including local concavity coefficient, discrete curvature, angular gradient and radial deviation; The extracted boundary morphological features are enhanced through a feature enhancement strategy combining local contrast enhancement and nonlinear transformation; Perform hierarchical fusion on the enhanced boundary morphological features; Gap detection is performed based on the enhanced fusion features.

2. The method for detecting notches of circular daily-use ceramic components under complex working conditions according to claim 1, characterized in that: The local concavity coefficient describes the degree of concavity or convexity of a boundary point relative to the local line segment and is defined as follows: In formula (9), p i represents the boundary point, L i-w,i+w Represents point p i-w and p i+w The line segments formed, Represents point p i To line segment L i-w,i+w distance, For line segment L i-w,i+w The unit normal vector of From the boundary point p i The vector pointing to the center point c of the ceramic edge; negative values indicate inward concavity, positive values indicate outward convexity; The discrete curvature describes the curvature of the boundary and is calculated as follows: In formula (10), R i is the boundary point p i The local circle fitting radius obtained by least squares fitting within the neighborhood; The angular gradient describes the degree of boundary turning, which is defined as the rate of change of the angle between adjacent boundary vectors: In formula (17), Represents the boundary vector; further calculate the local rate of change of the angle: The radial deviation describes the degree of deviation from the boundary point to the standard circle and directly reflects the roundness anomaly: In formula (19), is the distance from the boundary point to the center of the circle, is the average radius; in order to enhance the ability to distinguish radial deviations, the standardized radial deviation is introduced: In formula (20), is the standard deviation of the radial distance; N is the number of boundary sampling points.

3. The method for detecting notches of circular daily-use ceramic components under complex working conditions according to claim 2, characterized in that: Local circle fitting radius R i The calculation process includes: Let the local coordinate set Among them, u and v are the coordinates of the boundary points, u represents the horizontal coordinate, and v represents the vertical coordinate; w is the window size, which indicates how many points are taken around the current point i for fitting; j is the summation variable, and the mean value of the points within the range from iw to i+w is Then the boundary coordinate fluctuation is expressed as: Calculate the matrix elements required for fitting according to equations (11), (12), and (13): Construct the system of equations and solve it: The matrix is invertible, so: The final local circle fitting radius is: In equations (15) and (16), (a, b) represents the coordinates of the center of the fitting circle.

4. The method for detecting notches of circular daily-use ceramic components under complex working conditions according to claim 1, characterized in that: The extracted boundary morphological features are enhanced through a feature enhancement strategy that combines local contrast enhancement and nonlinear transformation, including: Local contrast enhancement: Adaptive background estimation method is used to enhance the contrast between features and background, as shown in formula (21): In formula (21), is the maximum value of the feature in the local window, and are the mean and standard deviation of the features in the large window respectively; ∈ is a small positive number to prevent the denominator from being zero; w l are the window radius respectively; Nonlinear transformation: Apply the power function to enhance significant features and suppress subtle changes, as shown in formula (22): F nonlinear (i)=sgn(F enhanced (i))·|F enhanced (i)| α (22) In formula (22), sgn(·) is the sign function, and α>1 is the nonlinear coefficient used to enhance the salient features.

5. The method for detecting notches of circular daily-use ceramic components under complex working conditions according to claim 1, characterized in that: The enhanced boundary morphological features are hierarchically fused, including two levels: base layer fusion and spatial consistency constraint; Among them, the basic layer fusion is to combine the features in a weighted manner, and the weight is determined based on the importance of the features, as shown in formula (23): In formula (23), w j is the weight coefficient of each feature, satisfying Spatial consistency constraint: Apply weighted median filtering to enhance the spatial consistency of features, as shown in formula (24): F filtered (i)=median{w(|j-i|)·F fusion (j)|j∈[i-w m ,i+w m ]}(24) In formula (24), w(d) is the distance weight function, w m is the filter window radius.

6. The method for detecting notches of circular daily-use ceramic components under complex working conditions according to claim 1, characterized in that: Based on the enhanced fusion features, adaptive threshold and spatial clustering analysis are used for gap detection, including: Adaptive threshold: Determine the optimal threshold based on the feature distribution, as shown in formula (25): In formula (25), and are the mean and standard deviation of the enhanced fusion features, respectively, and η is the adjustment coefficient; Spatial clustering: Perform spatial connectivity analysis on points exceeding the threshold, as shown in formula (26): C={C1,C2,...,C m } (26) In formula (26), represents the kth cluster, satisfying: In formula (27), F filtered (i) is the enhanced fusion feature; Gap assessment and location: Evaluate the significance of each cluster, as shown in formula (28): In formula (28), |C k | is cluster C k The size of σ({r i |i∈C k }) represents the standard deviation of the radius variation within the cluster; then, the cluster with the highest significance is selected as the gap position, as shown in formula (29):

7. The method for detecting notches of circular daily-use ceramic components under complex working conditions according to claim 6, characterized in that: Performing gap detection also includes: Calculate the geometric features of the detected gap based on the gap position: In formulas (30) and (31), L defect With W defect Respectively represent the length and depth of the notch.

8. The method for detecting notches of circular daily-use ceramic components under complex working conditions according to claim 1, characterized in that: The collected image data of the circular daily-use ceramic device is the original image data, and the original image data is pre-processed before the outer boundary of the circular daily-use ceramic device is extracted; Preprocess the raw image data, including: The weighted average method is used to convert the RGB image into a grayscale image, and then the Otsu algorithm is used to adaptively calculate the optimal threshold T opt ; The Otsu method is based on the principle of maximizing the between-class variance, and its objective function is: In formula (1), is the inter-class variance, ω0(T) and ω1(T) are the proportions of the two types of pixels in the entire image after the threshold T segmentation, μ0(T) and μ1(T) are the average grayscale values of the two types of pixels. After obtaining the optimal threshold, binarization is performed according to formula (2): In order to eliminate noise and maintain boundary integrity, median filtering and morphological opening and closing operations are used, as shown in Equations (3), (4), and (5): Where w represents the filter window size, S1 and S2 represent the structural elements of morphological opening and closing operations respectively. and represent the erosion and dilation operations respectively.

9. The method for detecting notches of circular daily-use ceramic components under complex working conditions according to claim 1, characterized in that: The outer boundary of circular daily-use ceramic components is extracted using a boundary tracking algorithm based on connected component labeling. The connected component labeling algorithm is based on depth-first search and is defined as: In formula (6), L(x,y) is the label value of pixel (x,y), Ω l is the l∈{1,2,...,N}th connected region; the boundary is extracted according to the connected region, as shown in formula (7): In formula (7), B l is the boundary of the lth connected region, N8(x,y) represents the 8-neighborhood of pixel (x,y); the longest boundary is selected as the ceramic edge contour: B bowl =argmax l |B l | (8) In formula (8), |B l | indicates the boundary B l The number of points included.

10. A notch detection system for circular daily-use ceramic components under complex working conditions, characterized in that: A method for detecting notches of circular daily-use ceramic components under complex working conditions according to any one of claims 1 to 9, comprising an image acquisition module, an image preprocessing module, a ceramic boundary extraction module, a feature extraction module, a feature enhancement module, a feature fusion module, and a notch detection module; in, The image acquisition module is used to collect original image data of the circular daily-use ceramic device; The image preprocessing module is used to preprocess the original image data; The ceramic boundary extraction module is used to extract the outer boundary of the circular daily-use ceramic component from the pre-processed image data; The feature extraction module is used to extract multiple complementary boundary morphological features; The feature enhancement module is used to enhance the extracted boundary morphological features; The feature fusion module is used to perform hierarchical fusion on the enhanced boundary morphological features; The gap detection module detects gaps based on the enhanced fusion features.

Images