A machine vision-based method for detecting unidirectional prepreg texture angles

Abstract

This invention discloses a machine vision-based method for detecting the texture angle of unidirectional prepreg. The method includes the following steps: Step 1, acquiring an image of the unidirectional prepreg; Step 2, calculating a candidate set of texture angles; Step 3, calculating a candidate set of baseline angles; Step 4, filtering the candidate set of texture angles to obtain the texture angle of the unidirectional prepreg; Step 5, filtering the candidate set of baseline angles to obtain the baseline angle; Step 6, calculating the difference between the texture angle of the unidirectional prepreg and the baseline angle to determine whether the layup direction meets the process requirements. This invention utilizes an actuator equipped with an industrial camera to automatically measure the angle between the unidirectional prepreg texture and the baseline, thereby solving the problems of poor real-time performance, low accuracy, and poor reliability in traditional layup direction detection. This method is applicable to online texture angle detection of carbon fiber unidirectional prepreg layup and has advantages such as high efficiency, stability, and accuracy.

Description

Technical Field

[0001] This invention relates to the field of unidirectional prepreg layup direction detection, and more particularly to a machine vision-based method for detecting the texture angle of unidirectional prepreg. Background Technology

[0002] Unidirectional prepreg layup is a crucial step in the manufacturing of aircraft components. During layup, unidirectional prepreg, cut to a specific shape, is layered and assembled on a mold according to a predetermined direction and sequence. The mold containing multiple layers of unidirectional prepreg is then placed in a hot press furnace for pressure, heating, and curing to form the critical aircraft component. In this layup process, workshop workers lay the unidirectional prepreg according to a pre-set direction with the aid of a projection reference line. However, manual layup can easily result in an excessively large angle between the unidirectional prepreg texture and the projection reference line, thus affecting the overall performance of the component.

[0003] To address the issue of excessively large angles between the unidirectional prepreg texture and the laser projection reference line during the layup process, the production workshop has established strict process requirements (the angle between the unidirectional prepreg texture and the projection reference line should be less than ±1°) and a corresponding layup direction detection scheme (after each layer is laid up, workers need to visually inspect each area to see if the unidirectional prepreg layup direction is consistent with the laser projection reference line). However, this detection scheme is not only time-consuming and labor-intensive, but also prone to visual fatigue for the inspectors, making it unreliable to detect whether the unidirectional prepreg layup direction meets the process requirements. Compared with traditional manual inspection methods, machine vision-based inspection methods have advantages such as high real-time performance and high detection accuracy. Therefore, this invention proposes a machine vision-based method for detecting the unidirectional prepreg texture angle. Summary of the Invention

[0004] This invention proposes a machine vision-based method for detecting the texture angle of unidirectional prepreg, the main purpose of which is to overcome the problems of poor real-time performance, low accuracy and poor reliability of traditional layup direction detection.

[0005] The technical solution adopted is: a machine vision-based method for detecting unidirectional prepreg texture angles, comprising the following steps:

[0006] Step 1: Acquire unidirectional images of the prepreg using an industrial camera;

[0007] Step 2: Perform edge detection on the unidirectional prepreg image, obtain the texture line segment angle based on the extracted texture line segments, and obtain a texture angle candidate set;

[0008] Step 3: Perform color space conversion on the unidirectional prepreg image, obtain the baseline segment angles based on the extracted baseline segments, and obtain a candidate set of baseline angles;

[0009] Step 4: Process the candidate set of texture angles to obtain the texture angles of the unidirectional prepreg.

[0010] Step 5: Based on the candidate set of baseline angles for the unidirectional prepreg texture angle specification, obtain the candidate set of baseline angles for the initial screening, and vote on the candidate set of baseline angles for the initial screening to obtain the baseline angles;

[0011] Step 6: The unidirectional prepreg texture angle is detected by judging the difference between the baseline angle and the unidirectional prepreg texture angle.

[0012] Step 1 specifically involves: when the industrial camera reaches a pre-set position, it is triggered and completes unidirectional prepreg image acquisition.

[0013] Step 2 involves edge detection of the unidirectional prepreg image, obtaining the texture line segment angles based on the extracted texture line segments, and acquiring a candidate set of texture angles. The specific steps are as follows:

[0014] Step 2-1: Apply Gaussian blur to the unidirectional prepreg image g(x,y) to remove noise and obtain the Gaussian filtered image f. Gauss (x, y):

[0015] A Gaussian filtered image f is obtained by convolving a unidirectional prepreg image g(x, y) with a two-dimensional Gaussian function G(x, y, σ). Gauss (x, y), the expression is as follows:

[0016]

[0017]

[0018] Wherein, parameter σ is the width of the Gaussian filter;

[0019] Step 2-2: Apply Gaussian filtering to image f Gauss Perform Canny edge detection on (x, y) to obtain the edge point image f. Canny (x, y);

[0020] Steps 2-3: Process the edge point image f Canny Perform Hough transform on (x, y) to extract multiple texture line segments and form a texture line segment set;

[0021] Steps 2-4: Calculate the angles of the texture line segment set to obtain the texture angle candidate set {θ} = {θ1, θ2, θ3, ..., θ}. n}; where θ is the texture line angle and n is the number of texture lines.

[0022] Step 2-2, applying the Gaussian filtered image f GaussPerform Canny edge detection on (x, y), including the following steps:

[0023] Step 2.1: Calculate the Gaussian filtered image f Gauss The strength and direction of the (x, y) gradient:

[0024] Edge detection using the Sobel operator: The Sobel operators for the horizontal and vertical directions are represented as follows: Sobel x Sobel y ;

[0025]

[0026]

[0027] Sobel operator and f Gauss Perform convolution on (x, y) to obtain the gradient values ​​d in the horizontal and vertical directions. x d y :

[0028] d x = f(x, y) * Sobel x (x, y)(1.5)

[0029] d y = f(x, y) * Sobel y (x, y)(1.6)

[0030] Image gradient intensity is represented as:

[0031]

[0032] The gradient direction of an image is represented as:

[0033] θ M =arctan(d y / d x (1.8)

[0034] Step 2.2: Apply Gaussian filtering to the image f Gauss Non-maximum suppression is performed on (x, y) to obtain the non-maximum suppressed image f. NMS (x, y): Gaussian filtered image f Gauss Each pixel in (x, y) is determined based on the image gradient intensity M(x, y) and the image gradient direction θ. M The system makes a judgment to determine whether the pixel should be suppressed. The judgment rule is as follows: if the gradient strength of the pixel is greater than the gradient strength of two pixels along the positive and negative gradient directions, then the pixel is retained as an edge point; otherwise, the pixel is suppressed.

[0035] Step 2.3: Apply double threshold detection and edge connection: For non-maximum suppression image f NMS (x, y) is used to detect and connect edges using a double threshold method, resulting in an edge point image f. Canny (x, y).

[0036] Step 3: Perform color space conversion on the unidirectional prepreg image, obtain the baseline segment angles based on the extracted baseline segments, and obtain a candidate set of baseline angles. The specific steps are as follows:

[0037] Step 3.1: Convert the RGB color space in the unidirectional prepreg image to the HSI color space. By adjusting the S component in the three HSI components, the binarized image T(x, y) of the projection baseline is obtained.

[0038] Step 3.2: Perform Hough transform on T(x, y) to extract multiple baseline segments;

[0039] Step 3.3: Calculate the angles of the baseline segments to obtain the candidate set of baseline angles. in, denoted as θ, where θ is the angle of the baseline segment, and m is the number of baseline segments.

[0040] Step 4 involves processing the candidate set of texture angles to obtain the unidirectional prepreg texture angles. The specific steps are as follows:

[0041] Compression of the texture angle candidate set {θ}:

[0042]

[0043] Round the compressed {θ} to the nearest integer:

[0044]

[0045] A voting strategy is applied to the rounded {θ}, and the texture angle with the most votes is taken as the texture angle φ of the unidirectional prepreg.

[0046]

[0047] Where θ1, θ2......θ n N(θ) represents the angle of the texture line segment. i ( ) is an angle of θ i The number of texture segments is denoted by argmax, which represents the maximum number of texture segments N(θ) extracted. i The texture angle θ corresponding to ) i .

[0048] Step 5 involves obtaining a preliminary screening baseline angle candidate set based on the unidirectional prepreg texture angle specification baseline angle candidate set, and then voting on the preliminary screening baseline angle candidate set to obtain the baseline angle. The specific steps are as follows:

[0049] Step 7.1: Use the candidate set of unidirectional prepreg texture angle φ to standardize the baseline angle. The initial screening baseline angle candidate set {α} is obtained:

[0050]

[0051]

[0052] {α} = {α1, α2, ..., α} s}(4.3)

[0053] Where β is the difference between the baseline angle μ and the unidirectional prepreg texture angle φ, m is the number of baseline segments, and α j This indicates the extraction of a baseline angle that differs from the unidirectional prepreg texture angle by ±5°, and s represents the number of baseline angles extracted.

[0054] Step 7.2: Round {α}:

[0055]

[0056] Step 7.3: Vote on the rounded {α}, and take the angle with the most votes as the baseline angle μ:

[0057]

[0058] Where α1, α2......α s Let N(α) be the angle of the baseline segment. j ) represents the angle α j The number of baseline segments, where argmax represents the maximum number of baseline segments N(α) extracted. j The baseline angle α corresponding to ) j .

[0059] Step 6, which involves detecting the unidirectional prepreg texture angle by determining the difference between the baseline angle and the unidirectional prepreg texture angle, specifically involves:

[0060] Determine whether the difference β between the baseline angle μ and the unidirectional prepreg texture angle φ is greater than 1°:

[0061] β=|φ-μ|(5.1)

[0062] If β is greater than 1°, an alarm will be triggered.

[0063] Otherwise, a normal message will be displayed, indicating that the next layer of unidirectional prepreg will continue to be laid.

[0064] A machine vision-based unidirectional prepreg texture angle detection device, comprising:

[0065] The image acquisition module is used to acquire unidirectional images of prepreg using an industrial camera;

[0066] The image processing module is used to perform edge detection on the unidirectional prepreg image, obtain the texture line segment angle based on the extracted texture line segments, and obtain a texture angle candidate set; perform color space conversion on the unidirectional prepreg image, obtain the baseline line segment angle based on the extracted baseline line segments, and obtain a baseline angle candidate set; process the texture angle candidate set to obtain the unidirectional prepreg texture angle; obtain a preliminary baseline angle candidate set based on the unidirectional prepreg texture angle specification baseline angle candidate set, and vote on the preliminary baseline angle candidate set to obtain the baseline angle;

[0067] The texture angle detection module is used to detect the texture angle of unidirectional prepreg by judging the difference between the baseline angle and the texture angle of the unidirectional prepreg.

[0068] Compared with traditional ply orientation detection, this invention has the following advantages and beneficial effects:

[0069] (1) The present invention directly uses the pixels of non-contact images as the basis for calculation, which reduces the human error generated in the traditional contact measurement process.

[0070] (2) Compared with traditional methods, the present invention has high efficiency, high precision, automatic measurement, and can effectively improve the efficiency of unidirectional prepreg stacking. Attached Figure Description

[0071] Figure 1 This is a flowchart of a machine vision-based unidirectional prepreg texture angle detection method according to the present invention.

[0072] Figure 2 To match the design drawings of the actuator of this invention;

[0073] The components include: 1 base frame, 2 cantilever supports, 3 X-axis linear modules, and 4 X-axis bases.

[0074] Figure 3 This is a schematic diagram showing the angle between the unidirectional prepreg texture and the projection reference line described in this invention. Detailed Implementation

[0075] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0076] The detection method of this invention includes the following steps: Step 1, acquiring unidirectional prepreg images; Step 2, calculating a candidate set of texture angles; Step 3, calculating a candidate set of baseline angles; Step 4, filtering the candidate set of texture angles to obtain the texture angle of the unidirectional prepreg; Step 5, filtering the candidate set of baseline angles to obtain the baseline angle; Step 6, calculating the difference between the texture angle of the unidirectional prepreg and the baseline angle to determine whether the layup direction meets the process requirements. This invention utilizes an actuator equipped with an industrial camera to automatically measure the angle between the unidirectional prepreg texture and the baseline, thereby solving the problems of poor real-time performance, low accuracy, and poor reliability in traditional layup direction detection. This method is suitable for online texture angle detection of carbon fiber unidirectional prepreg layup and has advantages such as high efficiency, stability, and accuracy.

[0077] like Figure 1 As shown, a machine vision-based method for detecting the unidirectional texture angle of prepreg includes the following steps:

[0078] Step 1, Actuator Device (see...) Figure 2 The system moves the industrial camera, triggering it when it reaches a fixed location to complete the unidirectional acquisition of prepreg images.

[0079] The actuator, a prior art technology, includes a base frame, an X-axis linear module, an X-axis base, a cantilever bracket, and a Y-axis linear module. The X-axis linear module is fixed to the base frame, and its output end is connected to the X-axis base on the base frame. The X-axis linear module drives the X-axis base to reciprocate along the X-axis direction (i.e., the length direction of the base frame). One end of the cantilever bracket is fixedly connected to the X-axis base, and the Y-axis linear module is mounted on the cantilever bracket. The Y-axis linear module drives the industrial camera to reciprocate along the Y-axis (i.e., the length direction of the cantilever bracket). Therefore, the industrial camera can move along both the X and Y axes, achieving two degrees of freedom image acquisition.

[0080] Step 2: Perform Gaussian blur, Canny edge detection, Hough transform, extract texture segments, and calculate texture segment angles on the unidirectional prepreg image to obtain a candidate set of texture angles.

[0081] Step 2.1: Apply Gaussian blur to the unidirectional prepreg image to remove noise and obtain the Gaussian filtered image f. Gauss (x, y):

[0082] A Gaussian filtered image f is obtained by convolving a unidirectional prepreg image g(x, y) with a two-dimensional Gaussian function G(x, y, σ). Gauss (x, y), the expression is as follows:

[0083]

[0084]

[0085] Wherein, parameter σ is the width of the Gaussian filter.

[0086] Step 2.2: Apply Gaussian filtering to the image f Gauss Perform Canny edge detection on (x, y) to obtain the edge point image f. Canny (x, y):

[0087] Step 2.2.1: Calculate the Gaussian filtered image f Gauss The strength and direction of the (x, y) gradient: Edge detection is performed using the Sobel operator. The Sobel operators in the horizontal and vertical directions are expressed as follows: Sobel x Sobel y .

[0088]

[0089]

[0090] Sobel operator and f Gauss Perform convolution on (x, y) to obtain the gradient values ​​d in the horizontal and vertical directions. x d y :

[0091] d x = f(x, y) * Sobel x (x, y)#(1.5)

[0092] d y = f(x, y) * Sobel y (x, y)#(1.6)

[0093] Image gradient intensity is represented as:

[0094]

[0095] The gradient direction of an image is represented as:

[0096] θ M =arctan(d y / d x )#(1.8)

[0097] Step 2.2.2: Apply Gaussian filtering to the image f Gauss Non-maximum suppression is performed on (x, y) to obtain the non-maximum suppressed image f. NMS (x, y): Gaussian filtered image f Gauss Each pixel in (x, y) is determined based on the image gradient intensity M(x, y) and the image gradient direction θ. MThe system makes a judgment to determine whether the pixel should be suppressed. The judgment rule is as follows: if the gradient strength of the pixel is greater than the gradient strength of two pixels along the positive and negative gradient directions, then the pixel is retained as an edge point; otherwise, the pixel will be suppressed.

[0098] Step 2.2.3: Apply double threshold detection: For non-maximum suppression image f NMS (x, y) is used to detect and connect edges using a double threshold method, resulting in an edge point image f. Canny (x, y).

[0099] Step 2.3: Perform a Hough transform on f(x, y) to extract multiple texture line segments:

[0100] Transform the xy coordinate space into polar coordinate space, passing through the point (x... i y i A straight line is represented as:

[0101] ρ=x i cosθ+y i sinθ#(1.9)

[0102] Where ρ is the perpendicular distance from the line to the origin, and θ is the angle between the X-axis and the perpendicular line.

[0103] The parameters ρ and θ are discretized into a finite number of equally spaced discrete values, and the ρ-θ parameter space is quantized into a series of equally sized grid cells. A pixel in the xy space, after being transformed into the ρ-θ parameter space, corresponds to a sine curve, and the accumulator counter for the ρ-θ grid cell corresponding to this sine curve is incremented by 1. After all pixels in the image space have been transformed, the grid cells are statistically analyzed, and (ρ, θ) values ​​that meet the threshold condition are converted into straight lines and classified as texture segments.

[0104] Step 2.4: Calculate the angle of the texture line segments:

[0105] Let the starting coordinates of the i-th line segment be (x... i y i The endpoint coordinates are (x′). i y′ i If the angle between the texture line segment and the Y-axis of the camera coordinate system is:

[0106]

[0107] The candidate set of texture angles is represented as: {θ} = {θ1, θ2, θ3, ..., θ} n}. Where θ represents the texture line angle, n represents the number of texture lines, and argmax represents the maximum number of texture lines N(θ) extracted. i The texture angle θ corresponding to ) i.

[0108] Step 3: Perform color space conversion, Hough transform, extract baseline segments, and calculate baseline segment angles on the unidirectional prepreg image to obtain a candidate set of baseline angles.

[0109] Step 3.1: Convert the RGB color space in the unidirectional prepreg image to the HSI color space. By adjusting the S component in the three HSI components, the binarized image T(x, y) of the projection baseline is obtained.

[0110] RGB converted to saturation S component is:

[0111]

[0112] Step 3.2: Perform Hough transform on T(x, y) to extract multiple baseline segments.

[0113] Step 3.3: Calculate the angles of the baseline segments:

[0114] Let the starting coordinates of the i-th baseline segment be (x... i y i The endpoint coordinates are (x′). i y′ i If the baseline segment is such that the angle between it and the Y-axis of the camera coordinate system is expressed as:

[0115]

[0116] The candidate set of baseline angles is represented as: in, denoted as θ, where θ is the angle of the baseline segment, and m is the number of baseline segments.

[0117] Step 4: Compress, round, and perform majority voting on the candidate texture angle set {θ} to obtain the unidirectional prepreg texture angle φ. The specific steps are as follows:

[0118] Step 4.1: Compress the texture angle candidate set {θ}:

[0119]

[0120] Step 4.2: Round the compressed {θ}:

[0121]

[0122] Step 4.3: Apply a voting strategy to the rounded {θ}, and select the texture angle with the most votes as the unidirectional prepreg texture angle φ:

[0123]

[0124] Where θ1, θ2......θ n N(θ) represents the angle of the texture line segment. i ( ) is an angle of θ i The number of texture segments.

[0125] Step 5: Based on the candidate set of unidirectional prepreg texture angle φ, specify the baseline angle. The initial candidate set of baseline angles {α} is obtained, and then {α} is rounded down and subjected to majority voting to obtain the baseline angle μ. The specific steps are as follows:

[0126] Step 5.1: Use the candidate set of unidirectional prepreg texture angle φ to standardize the baseline angle. The initial screening baseline angle candidate set {α} is obtained:

[0127]

[0128]

[0129] {α} = {α1, α2...α} s}#(4.3)

[0130] Where β is the difference between the baseline angle μ and the unidirectional prepreg texture angle φ, m is the number of baseline segments, and α j This indicates the extraction of a baseline angle that differs from the unidirectional prepreg texture angle by ±5°, and s represents the number of baseline angles extracted.

[0131] Step 5.2: Round {α}:

[0132]

[0133] Step 5.3: Apply a voting strategy to the rounded {α}, and take the angle with the most votes as the baseline angle μ:

[0134]

[0135] Where α1, α2......α s Let N(α) be the angle of the baseline segment. j ) represents the angle α j The number of baseline segments, where argmax represents the maximum number of baseline segments N(α) extracted. j The baseline angle α corresponding to ) j .

[0136] Step 6: Determine whether the difference β between the baseline angle μ and the unidirectional prepreg texture angle φ is greater than 1° (see...). Figure 3 ):

[0137] β=|φ-μ|#(5.1)

[0138] If β is greater than 1°, the system will issue an alarm; otherwise, it will indicate that everything is normal and the workshop workers will continue to lay the next layer of unidirectional prepreg.

[0139] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A machine vision-based method for detecting the unidirectional texture angle of prepreg, characterized in that, Includes the following steps: Step 1: Acquire unidirectional images of the prepreg using an industrial camera; Step 2: Perform edge detection on the unidirectional prepreg image, obtain the texture line segment angle based on the extracted texture line segments, and obtain a texture angle candidate set; Step 3: Perform color space conversion on the unidirectional prepreg image, obtain the baseline segment angles based on the extracted baseline segments, and obtain a candidate set of baseline angles; Step 4: Process the candidate set of texture angles to obtain the texture angles of the unidirectional prepreg. The candidate texture angle set {θ} is compressed, rounded, and voted on. The texture angle with the most votes is selected as the texture angle φ of the unidirectional prepreg. Where θ1, θ2……θ n N(θ) represents the angle of the texture line segment. i ( ) is an angle of θ i The number of texture segments is denoted by argmax, which represents the maximum number of texture segments N(θ) extracted. i The texture angle θ corresponding to ) i ; Step 5: Based on the candidate set of baseline angles for unidirectional prepreg texture angle specification, extract the baseline angles that differ from the unidirectional prepreg texture angle by ±5° to obtain the initial screening baseline angle candidate set. Vote on the initial screening baseline angle candidate set to obtain the baseline angle. Step 6: Detect the unidirectional prepreg texture angle by judging the difference between the baseline angle and the unidirectional prepreg texture angle; Determine whether the difference β between the baseline angle μ and the unidirectional prepreg texture angle φ is greater than 1°: β=|φ-μ|(5.1) If β is greater than 1°, an alarm will be triggered. Otherwise, a normal message will be displayed, indicating that the next layer of unidirectional prepreg will continue to be laid.

2. The method for detecting unidirectional prepreg texture angle based on machine vision according to claim 1, characterized in that, Step 1 specifically involves: when the industrial camera reaches a pre-set position, it is triggered and completes unidirectional prepreg image acquisition.

3. The method for detecting unidirectional prepreg texture angle based on machine vision according to claim 1, characterized in that, Step 2 involves edge detection of the unidirectional prepreg image, obtaining the texture line segment angles based on the extracted texture line segments, and acquiring a candidate set of texture angles. The specific steps are as follows: Step 2-1: Apply Gaussian blur to the unidirectional prepreg image g(x,y) to remove noise and obtain the Gaussian filtered image f. Gauss (x,y): A Gaussian filtered image f is obtained by convolving a unidirectional prepreg image g(x,y) with a two-dimensional Gaussian function G(x,y,σ). Gauss (x, y), expressed as follows: Wherein, parameter σ is the width of the Gaussian filter; Step 2-2: Apply Gaussian filtering to image f Gauss Perform Canny edge detection on (x,y) to obtain the edge point image f. Canny (x,y); Steps 2-3: Process the edge point image f Cann Perform Hough transform on (x,y) to extract multiple texture line segments and form a texture line segment set; Steps 2-4: Calculate the angles of the texture line segment set to obtain the texture angle candidate set {θ} = {θ1, θ2, θ3, ..., θ}. n }; where θ is the texture line angle and n is the number of texture lines.

4. The method for detecting unidirectional prepreg texture angle based on machine vision according to claim 3, characterized in that, Step 2-2, applying the Gaussian filtered image f Gauss Perform Canny edge detection on (x,y), including the following steps: Step 2.1: Calculate the Gaussian filtered image f Gauss The strength and direction of the (x,y) gradient: Edge detection using the Sobel operator: The Sobel operators for the horizontal and vertical directions are represented as follows: Sobel x Sobel y ; Sobel operator and f Gauss Perform convolution on (x, y) to obtain the gradient values ​​d in the horizontal and vertical directions. x d y : d x =f Gauss (x,y)*Sobel x (x,y) (1.5) d y =f Gauss (x,y)*Sobel y (x,y) (1.6) Image gradient intensity is represented as: The gradient direction of an image is represented as: θ M =arctane(d y / d x ) (1.8) Step 2.2: Apply Gaussian filtering to the image f Gauss Perform nonmaximum suppression on (x,y) to obtain the nonmaximum suppressed image f. NMS (x,y): Gaussian filtered image f Gauss Each pixel in (x,y) is determined by the image gradient intensity M(x,y) and the image gradient direction θ. M The system makes a judgment to determine whether the pixel should be suppressed. The judgment rule is as follows: if the gradient strength of the pixel is greater than the gradient strength of two pixels along the positive and negative gradient directions, then the pixel is retained as an edge point; otherwise, the pixel is suppressed. Step 2.3: Apply double threshold detection and edge connection: For non-maximum suppression image f NMS (x, y) is used to detect and connect edges using a double threshold method, resulting in an edge point image f. Canny (x,y).

5. The method for detecting unidirectional prepreg texture angle based on machine vision according to claim 1, characterized in that, Step 3: Perform color space conversion on the unidirectional prepreg image, obtain the baseline segment angles based on the extracted baseline segments, and obtain a candidate set of baseline angles. The specific steps are as follows: Step 3.1: Convert the RGB color space in the unidirectional prepreg image to the HSI color space. By adjusting the S component in the three HSI components, the binarized image T(x,y) of the projection baseline is obtained. Step 3.2: Perform a Hough transform on T(x,y) to extract multiple baseline segments; Step 3.3: Calculate the angles of the baseline segments to obtain the candidate set of baseline angles. in, denoted as θ, where θ is the angle of the baseline segment, and m is the number of baseline segments.

6. The method for detecting unidirectional prepreg texture angle based on machine vision according to claim 1, characterized in that, Step 4 involves processing the candidate set of texture angles to obtain the unidirectional prepreg texture angles. The specific steps are as follows: Compression of the texture angle candidate set {θ}: Round the compressed {θ} to the nearest integer: A voting strategy is applied to the rounded {θ}, and the texture angle with the most votes is taken as the texture angle φ of the unidirectional prepreg.

7. The method for detecting unidirectional prepreg texture angle based on machine vision according to claim 1, characterized in that, Step 5 involves obtaining a preliminary screening baseline angle candidate set based on the unidirectional prepreg texture angle specification baseline angle candidate set, and then voting on the preliminary screening baseline angle candidate set to obtain the baseline angle. The specific steps are as follows: Step 7.1: Use the candidate set of unidirectional prepreg texture angle φ to standardize the baseline angle. The initial screening baseline angle candidate set {α} is obtained: {α}＝{α1,α2...α s}(4.3) Where β is the difference between the baseline angle μ and the unidirectional prepreg texture angle φ, m is the number of baseline segments, and α j This indicates the extraction of a baseline angle that differs from the unidirectional prepreg texture angle by ±5°, and s represents the number of baseline angles extracted. Step 7.2: Round {α}: Step 7.3: Vote on the rounded {α}, and take the angle with the most votes as the baseline angle μ: Where α1, α2……α s Let N(α) be the angle of the baseline segment. j ) represents the angle α j The number of baseline segments, where argmax represents the maximum number of baseline segments N(α) extracted. j The baseline angle α corresponding to ) j .

8. A machine vision-based unidirectional prepreg texture angle detection device, the device being used to implement the machine vision-based unidirectional prepreg texture angle detection method as described in any one of claims 1-7, characterized in that, include: The image acquisition module is used to acquire unidirectional images of prepreg using an industrial camera; The image processing module is used to perform edge detection on the unidirectional prepreg image, obtain the texture line segment angle based on the extracted texture line segments, and obtain a texture angle candidate set; perform color space conversion on the unidirectional prepreg image, obtain the baseline line segment angle based on the extracted baseline line segments, and obtain a baseline angle candidate set; process the texture angle candidate set to obtain the unidirectional prepreg texture angle; obtain a preliminary baseline angle candidate set based on the unidirectional prepreg texture angle specification baseline angle candidate set, and vote on the preliminary baseline angle candidate set to obtain the baseline angle; The texture angle detection module is used to detect the texture angle of unidirectional prepreg by judging the difference between the baseline angle and the texture angle of the unidirectional prepreg.

Images