Image feature detection method based on flexible connection

Abstract

The invention provides an image feature detection method based on flexible connection, relates to the technical field of image analysis and processing, and aims to solve the defects of an existing image feature detection method in the aspects of accuracy, processing speed, detection sensitivity, false detection rate and the like. The method comprises the following steps: firstly, acquiring an input image, converting the input image into a grayscale image, carrying out Gaussian filtering processing, then calculating a gradient magnitude and direction, and carrying out non-maximum suppression and dual-threshold processing to obtain a preliminary edge image and a final edge image; and then connecting the weak edge pixels, extracting image features through a connected region analysis algorithm, calculating parameters such as a geometric center, a perimeter and an area, judging whether the features meet requirements or not according to a preset rule, and finally outputting the image features meeting the requirements. According to the method, parameters can be flexibly adjusted, different images can be adapted, the method is excellent in the fields of object recognition, medical image analysis, industrial detection and the like, the detection accuracy and efficiency are improved, the false detection rate is reduced, and the method has wide application prospects and practical value.

Description

Technical Field

[0001] The present invention belongs to the technical field of image analysis and processing, and more specifically, particularly relates to an image feature detection method based on soft connection. Background Art

[0002] In the fields of electrical and electronic, soft connections are often used to achieve flexible connection and current transmission between different components. However, the quality and reliability of soft connections are crucial, especially the integrity and stability of welding parts. In the production and inspection process of soft connections, the application of image feature detection technology faces many challenges.

[0003] Traditional image feature detection methods often have problems such as insufficient accuracy, slow processing speed, low detection sensitivity, and high false detection rate when detecting the welding quality of soft connections. It is difficult to accurately detect defects such as tiny cracks, cold welds, and pores that may appear in the welding parts of soft connections. The slow processing speed cannot meet the needs of efficient detection in large-scale production. The limitation of detection sensitivity may lead to missed detection of minor defects, affecting product quality. The high false detection rate will increase unnecessary costs and waste time.

[0004] In addition, traditional methods have poor adaptability to soft connection images and are difficult to effectively detect soft connections of different materials, shapes and sizes. Due to the diverse application scenarios of soft connections and the complex and changeable image features, there is an urgent need for a new image feature detection method specifically for soft connections that can overcome the above defects and achieve accurate, fast, sensitive and adaptable detection to ensure the quality and reliability of soft connections and meet the growing needs of related fields.

[0005] It is under such a background that the present invention is dedicated to providing an image feature detection method based on soft connection to solve the difficulties in the prior art. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides an image feature detection method based on soft connection to solve the above problems.

[0007] A method for detecting image features based on soft connection comprises the following steps: Step S1: Get the input image and convert it into a grayscale image. The image size is M×N pixels, where M ranges from [512, 1024] and N ranges from [512, 1024]. When performing grayscale conversion, the weighted average method is used, and the calculation formula is Gray=0.299×R+0.587×G+0.114×B, where R, G, and B are the red, green, and blue channel values ​​of the color image respectively; Step S2: Perform Gaussian filtering on the grayscale image. The Gaussian kernel size is K×K, K is [3,7], the standard deviation is σ, and the σ value range is [0.5,2.0]. The calculation formula of Gaussian filtering is: G(x,y)=1 / (2πσ^2)*exp(-((x-μx)^2+(y-μy)^2) / (2σ^2)), where (x,y) is the pixel coordinate, and (μx,μy) is the center coordinate of the Gaussian kernel; Step S3: Calculate the gradient magnitude and direction of the image. The gradient is calculated using the Sobel operator. The gradient magnitude calculation formula is G=√(Gx^2+Gy^2), where Gx=(I(x+1,y)-I(x-1,y)) / 2, Gy=(I(x,y+1)-I(x,y-1)) / 2, and I(x,y) is the grayscale value of the pixel (x,y); Step S4: Based on the gradient direction, non-maximum suppression is performed on the gradient amplitude to obtain a preliminary edge image. The non-maximum suppression uses a 3×3 neighborhood window to compare the central pixel gradient amplitude with the two adjacent pixel gradient amplitudes along the gradient direction. If the central pixel gradient amplitude is smaller than the adjacent pixel gradient amplitude, the central pixel grayscale value is set to 0. Step S5: perform threshold processing on the preliminary edge image using double thresholds T1 and T2, where the value range of T1 is [10, 50], and the value range of T2 is [50, 150]. Pixels with values ​​greater than T2 are marked as strong edge pixels, pixels with values ​​less than T1 are marked as non-edge pixels, and pixels with values ​​between T1 and T2 are marked as weak edge pixels. Step S6: Connect the weak edge pixels in the soft-connected image. If there are strong edge pixels in the eight neighborhoods of the weak edge pixels, mark them as strong edge pixels, thereby obtaining the final edge image. Step S7: on the final edge image, extract the soft connection image features by using the connected region analysis algorithm, the area threshold of the connected region is A, and the value range of A is [100,500] pixels; Step S8: Calculate the geometric center, perimeter and area parameters of the extracted soft connection image features. The calculation formula of the geometric center is (Xc, Yc) = (∑ (xi*ai) / A, ∑ (yi*ai) / A), where (xi, yi) is the pixel coordinate, ai is the pixel value, and A is the area of ​​the connected area; Step S9: judging whether the extracted soft connection image features meet the requirements according to the preset soft connection feature judgment rule, wherein the feature judgment rule is set based on the numerical range or proportional relationship of the geometric center, perimeter and area parameters of the soft connection image features; Step S10: outputting the soft link image features that meet the requirements, and the output soft link image features are in a specific data format.

[0008] Preferably, in step S2, the Gaussian kernel is generated by using a method of separating the Gaussian kernels, first calculating a one-dimensional Gaussian kernel Gx in the horizontal direction, then calculating a one-dimensional Gaussian kernel Gy in the vertical direction, and then obtaining a two-dimensional Gaussian kernel G through convolution.

[0009] Preferably, in step S3, the horizontal template of the Sobel operator is [-1, 0, 1; -2, 0, 2; -1, 0, 1], and the vertical template is [1, 2, 1; 0, 0, 0; -1, -2, -1], and when calculating the gradient, the edge pixels of the soft-connected image are mirror-filled.

[0010] Preferably, in step S4, the gradient direction is calculated using an inverse tangent function, the calculation formula is θ=arctan(Gy / Gx), and the direction value range is [0,360) degrees.

[0011] Preferably, in step S5, the values ​​of the dual thresholds T1 and T2 are adaptively determined according to the grayscale histogram distribution of the soft-connected image, or are manually set by the user based on experience.

[0012] Preferably, in step S6, for the connection processing of weak edge pixels in the soft-connected image, in addition to considering eight neighborhoods, it can also be expanded to sixteen neighborhoods according to the complexity and feature requirements of the image.

[0013] Preferably, in step S7, when analyzing the connected regions of the soft-connected image, the circumscribed rectangle and the minimum enclosing circle of the connected regions are recorded simultaneously.

[0014] Preferably, in step S8, in addition to calculating the geometric center, perimeter and area of ​​the soft-connected image, the eccentricity, rectangularity and circularity of the connected region are also calculated.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. Improve the accuracy of soft connection welding quality detection: It can more accurately detect the characteristics of soft connection welding parts, reduce false detection and missed detection, and provide a more reliable basis for ensuring the reliability of soft connections.

[0016] 2. Shorten the soft connection image processing time: Compared with traditional methods, it significantly reduces the soft connection image processing time, improves detection efficiency, and helps to quickly complete detection in production.

[0017] 3. Enhanced sensitivity of soft connection welding defect detection: It has higher detection sensitivity for tiny cracks, pores and other subtle defects in soft connection welding, which helps to detect potential problems in time.

[0018] 4. Reduce the false detection rate of soft connection detection: effectively reduce the false detection in soft connection quality detection and improve the reliability of soft connection product quality control.

[0019] 5. Adapt to different soft connection scenarios: Through parameter adjustment, it can adapt to soft connection images of different materials, shapes and sizes, and has wide applicability.

[0020] 6. Improve the efficiency of soft link image processing: Optimized algorithms and processes make the entire soft link image feature detection process more efficient and save computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0022] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0023] In the electrical field, the quality of soft connections is crucial to the stable operation of electrical equipment. The image feature detection method based on soft connections proposed in the present invention aims to provide a more accurate and efficient solution for the quality detection of soft connections. The application and effect of this method in soft connection detection will be demonstrated in detail through specific embodiments.

[0024] Image feature detection plays a vital role in many fields, such as computer vision, pattern recognition, medical imaging, etc. The image feature detection method based on soft connection proposed in the present invention aims to provide a more accurate and efficient image feature detection solution. The implementation process and effect of the method will be described in detail through multiple specific embodiments.

[0025] The image feature detection method based on soft connection of the present invention comprises the steps of acquiring input image, grayscale conversion, Gaussian filtering, gradient calculation, non-maximum suppression, threshold processing, weak edge pixel connection, connected region analysis, feature parameter calculation, feature judgment and output.

[0026] Example 1: Image feature detection in object recognition: Experimental setup: Input images: A set of color images containing different objects are selected with a size of 800×600 pixels.

[0027] Gaussian filter parameters: Gaussian kernel size K=5, standard deviation σ=1.2.

[0028] Thresholds: T1=20, T2=80.

[0029] The connected region area threshold A = 200 pixels.

[0030] Experimental process: The input image is processed sequentially according to the steps of the method.

[0031] Record the processing time and intermediate results of each step.

[0032] Experimental results: The edges and features of the object are successfully detected.

[0033] Characteristic parameter statistics: Compared with existing methods, the detection accuracy of this method is improved by 10% and the processing time is shortened by 20%.

[0034] Example 2: Lesion feature detection in medical images: Experimental setup: Input images: A set of medical image grayscale images with a size of 1024×1024 pixels.

[0035] Gaussian filter parameters: Gaussian kernel size K=7, standard deviation σ=1.8.

[0036] Thresholds: T1=15, T2=100.

[0037] The connected region area threshold A = 150 pixels.

[0038] Experimental process: Execute the image feature detection process.

[0039] Analyze the impact of different parameters on the test results.

[0040] Experimental results: Accurately detect the edges and features of lesions.

[0041] Characteristic parameter statistics: Compared with traditional methods, this method has a 15% higher sensitivity in detecting tiny lesions.

[0042] Example 3: Surface defect detection of industrial products: Experimental setup: Input image: A color image of the surface of an industrial product, with a size of 640 × 480 pixels.

[0043] Gaussian filter parameters: Gaussian kernel size K=3, standard deviation σ=0.8.

[0044] Thresholds: T1=30, T2=120.

[0045] The connected region area threshold A = 100 pixels.

[0046] Experimental process: The method of the present invention is applied to detect defect characteristics.

[0047] Adjust parameters to optimize detection performance.

[0048] Experimental results: Effectively detect defects on the product surface.

[0049] Characteristic parameter statistics: Compared with commonly used detection methods, the false positive rate of this method is reduced by 12%.

[0050] The processing time of the method of the present invention and the traditional method in each embodiment is statistically analyzed, and the results are shown in the following table: The detection accuracy of the method of the present invention and the existing method in each embodiment is compared, and the data are as follows: For the medical imaging example, the detection sensitivity of the method of the present invention and the traditional method for small lesions is statistically analyzed as shown in the following table: In the industrial product surface defect detection example, the false detection rates of the method of the present invention and the commonly used methods are calculated as shown in the following table: Algorithm formula and parameter description: 1. Grayscale conversion: The weighted average method formula is: Gray=0.299×R+0.587×G+0.114×B; in the embodiment, the specific color pixel value is substituted into the calculation to obtain the corresponding gray value.

[0051] 2. Gaussian filtering: Formula: G(x,y)=1 / (2πσ^2)*exp(-((x-μx)^2+(y-μy)^2) / (2σ^2)); For example, in Example 1, μx and μy are the coordinates of the center pixel of the image, σ=1.2, and the filtered value of each pixel is calculated.

[0052] 3. Gradient amplitude: formula: G = √(Gx^2+Gy^2); Gx=(I(x+1,y)-I(x-1,y)) / 2, Gy=(I(x,y+1)-I(x,y-1)) / 2; The gradient components Gx and Gy are obtained by calculating the difference between adjacent pixels, and then the gradient amplitude G is calculated.

[0053] 4. Geometric center: Formula: (Xc, Yc) = (∑(xi*ai) / A,∑(yi*ai) / A); When calculating the geometric center of the connected area, accumulation and averaging operations are performed based on the pixel coordinates and pixel values.

[0054] Through the above three embodiments and related data tables, we can clearly see the remarkable effects of the soft-link-based image feature detection method of the present invention in different fields and scenarios.

[0055] In terms of processing time, compared with traditional methods, the method of the present invention has achieved significant time reduction in each embodiment, improving the efficiency of image processing. The comparison of detection accuracy shows that the method of the present invention has been significantly improved in object recognition, medical image lesion detection and industrial product defect detection, and can more accurately capture the key features in the image. In terms of detection sensitivity, especially in the detection of tiny lesions in medical images, the present invention has demonstrated superior performance and provided a more reliable basis for the early diagnosis of the disease. In the false detection rate evaluation, the embodiment of industrial product surface defect detection shows that the present invention effectively reduces the situation of false detection and improves the reliability and accuracy of detection.

[0056] These examples and data fully demonstrate the effectiveness and superiority of the method of the present invention, and demonstrate its strong adaptability and outstanding performance in different application scenarios.

[0057] The soft-link-based image feature detection method proposed in the present invention provides an innovative and efficient solution for the field of image feature detection through a series of carefully designed steps and optimized algorithms. Whether it is accurately capturing the contours and features of an object in object recognition, or keenly discovering lesions in medical images, or accurately detecting product surface defects in industrial inspection, the present invention has demonstrated performance that exceeds traditional methods.

[0058] Its core advantage is that it can process images quickly and accurately, extract valuable feature information, and has good parameter adjustability, which can adapt to images of different types and qualities. This flexibility and high performance make the invention have broad application prospects and great practical value in many fields that rely on image feature detection, and is expected to bring significant technological progress and economic benefits to related industries.

[0059] It can be seen from the above embodiments for soft connections that the image feature detection method based on soft connections of the present invention performs well in different application scenarios, can effectively detect quality problems of soft connections, and provides strong support for quality control and assurance of soft connections in the electrical industry.

[0060] The embodiments of the present invention are given for the purpose of illustration and description, and are not intended to be exhaustive or to limit the invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

Claims

1. A method for detecting image features based on soft connection, characterized in that , including the following steps: Step S1: Get the input image and convert it into a grayscale image. The image size is M×N pixels, where M ranges from [512, 1024] and N ranges from [512, 1024]. When performing grayscale conversion, the weighted average method is used, and the calculation formula is Gray=0.299×R+0.587×G+0.114×B, where R, G, and B are the red, green, and blue channel values ​​of the color image respectively; Step S2: Perform Gaussian filtering on the grayscale image. The Gaussian kernel size is K×K, K is [3,7], the standard deviation is σ, and the σ value range is [0.5,2.0]. The calculation formula of Gaussian filtering is: G(x,y)=1 / (2πσ^2)*exp(-((x-μx)^2+(y-μy)^2) / (2σ^2)), where (x,y) is the pixel coordinate, and (μx,μy) is the center coordinate of the Gaussian kernel; Step S3: Calculate the gradient magnitude and direction of the image. The gradient is calculated using the Sobel operator. The gradient magnitude calculation formula is G=√(Gx^2+Gy^2), where Gx=(I(x+1,y)-I(x-1,y)) / 2, Gy=(I(x,y+1)-I(x,y-1)) / 2, and I(x,y) is the grayscale value of the pixel (x,y); Step S4: Based on the gradient direction, non-maximum suppression is performed on the gradient amplitude to obtain a preliminary edge image. The non-maximum suppression uses a 3×3 neighborhood window to compare the central pixel gradient amplitude with the two adjacent pixel gradient amplitudes along the gradient direction. If the central pixel gradient amplitude is smaller than the adjacent pixel gradient amplitude, the central pixel grayscale value is set to 0. Step S5: perform threshold processing on the preliminary edge image using double thresholds T1 and T2, where the value range of T1 is [10, 50], and the value range of T2 is [50, 150]. Pixels with values ​​greater than T2 are marked as strong edge pixels, pixels with values ​​less than T1 are marked as non-edge pixels, and pixels with values ​​between T1 and T2 are marked as weak edge pixels. Step S6: Connect the weak edge pixels in the soft-connected image. If there are strong edge pixels in the eight neighborhoods of the weak edge pixels, mark them as strong edge pixels, thereby obtaining the final edge image. Step S7: On the final edge image, the soft connection image features are extracted through the connected region analysis algorithm. The area threshold is A, and the value range of A is [100,500] pixels. ; Step S8: Calculate the geometric center, perimeter and area parameters of the extracted soft connection image features. The formula is (Xc,Yc)=(∑(xi*ai) / A,∑(yi*ai) / A), where (xi,yi) are pixel coordinates, ai is the pixel value, and A is the area of ​​the connected region; Step S9: judging whether the extracted soft connection image features meet the requirements according to the preset soft connection feature judgment rule, wherein the feature judgment rule is set based on the numerical range or proportional relationship of the geometric center, perimeter and area parameters of the soft connection image features; Step S10: outputting the soft link image features that meet the requirements, and the output soft link image features are in a specific data format.

2. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In step S2, the Gaussian kernel is generated by using a separation Gaussian kernel method. First, the one-dimensional Gaussian kernel Gx in the horizontal direction is calculated, and then the one-dimensional Gaussian kernel Gy in the vertical direction is calculated. Then, the two-dimensional Gaussian kernel G is obtained by convolution.

3. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In step S3, the horizontal template of the Sobel operator is [-1, 0, 1; -2, 0, 2; -1, 0, 1], and the vertical template is [1, 2, 1; 0, 0, 0; -1, -2, -1]. When calculating the gradient, the edge pixels of the soft-connected image are mirror-filled.

4. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In step S4, the gradient direction is calculated using the arc tangent function, the calculation formula is θ=arctan(Gy / Gx), and the direction value range is [0,360) degrees.

5. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In step S5, the values ​​of the dual thresholds T1 and T2 are adaptively determined according to the grayscale histogram distribution of the soft-connected image, or are manually set by the user based on experience.

6. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In step S6, for the connection processing of weak edge pixels in the soft-connected image, in addition to considering eight neighborhoods, it can also be expanded to sixteen neighborhoods according to the complexity and feature requirements of the image.

7. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In step S7, when the connected region analysis of the soft-connected image is performed, the circumscribed rectangle and the minimum enclosing circle of the connected region are recorded simultaneously.

8. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In step S8, in addition to calculating the geometric center, perimeter and area of ​​the soft-connected image, the eccentricity, rectangularity and circularity of the connected region are also calculated.

9. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In step S10 , the specific data format includes XML, JSON or binary format.

10. The image feature detection method based on soft connection as claimed in claim 1, characterized in that: In S8, the perimeter of the soft-connected image is obtained by calculating the sum of the Euclidean distances between the boundary pixels of the connected region, and the area is obtained by counting the number of pixels in the connected region.

Images