Heat pipe orifice defect detection device and method based on machine vision

Abstract

The present invention relates to the technical field of heat pipe visual inspection, and provides a heat pipe orifice defect detection device and method based on machine vision, including a computer system, a workbench, a orifice visual inspection mechanism, and a heat pipe horizontal positioning mechanism. The computer system includes image preprocessing, feature extraction, and defect detection steps, and the orifice visual inspection mechanism includes a visual processing module: a camera, a lens, and a light source; the heat pipe horizontal positioning mechanism is used to make the heat pipe coaxial with the camera and fix it horizontally. The present invention is aimed at the automatic mesh threading and diameter reduction process of heat pipes, detects whether there are burrs or out-of-round defects at the incoming pipe orifice, improves the product processing yield, overcomes the subjective interference of manual visual inspection methods, quickly and accurately detects defects at the pipe orifice, and improves the process efficiency and detection accuracy of the product.

Description

Technical Field

[0001] The present invention relates to the technical field of copper pipe orifice defect detection, and in particular to a heat pipe orifice defect detection device and method based on machine vision. Background Art

[0002] With the continuous development of the semiconductor industry, heat pipes have been widely used in various electronic products as the main heat dissipation device for a large number of electronic power-consuming devices. In the manufacturing process of heat pipes, the copper material is relatively soft. After cutting the pipes in large quantities, it is easy to cause burr defects on the copper pipe cuts. At the same time, during the transportation of copper pipes, it is also very easy to cause the copper pipe mouth to lose roundness defects. In the processing of automated equipment, it is difficult to detect the failure of the mesh threading process and the diameter reduction process. In severe cases, the equipment will be unable to cut the mesh and the material will be broken and the die will be stuck, resulting in the scrapping of a batch of processed parts. Therefore, before the automatic mesh threading process and the automatic diameter reduction process, it is necessary to check whether the incoming copper pipe mouth has burrs or out-of-round defects to prevent the burr defects from hindering the wire mesh from entering the inner cavity of the copper pipe and avoid the out-of-round defects from causing forming defects in the copper pipe reduction process. Due to the large amount of detection and the small burr characteristics, manual inspection is often rough, and the scrap rate is high. With the increasing labor costs and production precision requirements, there is an urgent need for an efficient detection device that can realize the defects of the heat pipe mouth. Machine vision has been widely used to improve production efficiency and the degree of production automation. It has the characteristics of non-contact measurement. Applying machine vision to the incoming material process of mesh shrink tube automation equipment can realize efficient and accurate detection of heat pipe nozzle defects. Summary of the invention

[0003] In view of the technical problems existing in the prior art, the purpose of the present invention is to provide a heat pipe orifice defect detection device and method based on machine vision, so as to realize efficient, high-precision, non-contact quantitative defect detection of the heat pipe orifice, thereby improving the detection accuracy and efficiency of the product.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A heat pipe orifice defect detection device based on machine vision, including a computer system and a vision device;

[0006] The visual device is used to horizontally detect the pipe mouth defects at the front end of the copper pipe, and includes a workbench, a pipe mouth visual detection mechanism and a heat pipe horizontal positioning mechanism. The pipe mouth visual detection mechanism and the heat pipe horizontal positioning mechanism are both installed on the workbench, and the heat pipe horizontal positioning mechanism is arranged in front of the pipe mouth visual detection mechanism;

[0007] The heat pipe horizontal positioning mechanism is used to horizontally fix heat pipes of different outer diameters and lengths, and keep the heat pipe and the lens of the pipe mouth visual inspection mechanism in the same straight line. The pipe mouth visual inspection mechanism collects the pipe mouth image information and transmits it to the computer system;

[0008] The computer system is used to process the pipe nozzle image and identify defects, and includes three modules: image preprocessing, feature extraction and defect detection.

[0009] The pipe mouth visual inspection mechanism includes an image acquisition system and a visual instrument device; the image acquisition system is installed on the visual instrument device;

[0010] The image acquisition system includes a camera, a lens, a light source and a light source controller;

[0011] The visual instrument device includes a visual support plate, a fine-tuning slide mounting plate, a fine-tuning slide, a camera mounting plate, a light source fixing block, a light source fixing seat, and a light source sliding plate;

[0012] The lens is mounted on the camera, the camera is set on the camera mounting plate and connected to the computer system, the camera mounting plate is fixed on the fine-tuning slide, the fine-tuning slide is mounted on the vision support plate through the fine-tuning slide mounting plate, and the vision support plate is fixed on the workbench;

[0013] The light source is installed on the light source sliding block through the light source fixing block, the light source sliding block is slidably installed on the light source fixing seat, and the light source sliding block is installed on the inner side of the visual support plate, and the light source fixing seat and the bottom of the light source fixing block are arranged on the workbench.

[0014] The heat pipe horizontal positioning mechanism includes a positioning base, a heat pipe, a heat pipe fixer, a rear positioning block, and a positioning screw; the positioning base is fixed to the workbench by a flat key and screws, the heat pipe fixer and the rear positioning block are arranged in a slot opened above the positioning base, the heat pipe is placed on the V-groove on the heat pipe fixer, the rear positioning block positions the rear port of the heat pipe, and the heat pipe fixer and the rear positioning block are fixed to the base by positioning screws.

[0015] The camera, lens, light source and heat pipe are kept in the same straight line.

[0016] The heat pipe orifice defect detection method based on machine vision adopts the heat pipe orifice defect detection device based on machine vision, and comprises the following steps:

[0017] Step 1: The image preprocessing module includes image acquisition, image cropping and contour detection parts; the picture collected by the camera in the pipe mouth visual detection mechanism 2 is transmitted to the computer system, and the grayscale centroid method is used to crop the area of ​​interest containing the pipe mouth features, and the shape area and inner and outer contours of the pipe mouth end face are obtained through morphological processing.

[0018] Step 2: The feature extraction module includes target positioning, target flattening, and skeleton extraction. The least squares circle fitting method is used to calculate the fitting circle of the inner and outer contours, and the mean merging method is combined to complete the centroid position positioning and ring size detection of the pipe mouth. The target area is transformed into polar coordinates to obtain the flattened image of the pipe mouth. The central skeleton of the pipe mouth is extracted through the refinement algorithm and smoothing processing.

[0019] Step 3: The defect detection module includes two parts: burr defect detection and out-of-round defect detection. The skeleton gradient directional edge search algorithm is used to obtain the measured values ​​of the pipe mouth width at different positions, and the width tolerance value is set to distinguish burr defects. The skeleton line fluctuation characteristics are analyzed, and the position tolerance value is set to distinguish out-of-round defects. If any of the above defects is detected, the pipe mouth quality is judged to be unqualified.

[0020] In the step one, the lighting source is selected and arranged, a black and white camera is used as the camera, a telecentric lens is used as the lens to reduce image distortion, a ring-shaped white light source is used as the light source, and the pictures collected by the camera are transmitted to the computer system, and the image processing is performed by the software system.

[0021] In the step 1, a fast global threshold segmentation process is performed on the collected pipe mouth image, a circular area within a suitable circumscribed circle radius range is selected, and then the grayscale centroid of the area is calculated, and the circular ROI is set to be slightly larger than the circular area, the original image is reduced, and interference sources in the non-ROI area of ​​the image are removed. Copper pipes with different outer diameters can be automatically cropped, denoised and enlarged for display.

[0022] In the step 1, median filtering and mean filtering are used to preprocess the grayscale image, and the OTSU threshold segmentation method is selected to perform binary segmentation on the processed pipe mouth image. The Euler number feature is used to determine whether the extracted shape area is the pipe mouth end face. If the target area is extracted correctly, the inner and outer contour edges of the pipe mouth are obtained, and the roundness of the inner and outer contours is calculated.

[0023] In the step 2, the least squares circle fitting method based on the Tukey algorithm is used to fit the inner contour points and the outer contour points of the pipe mouth respectively, and the minimum circumscribed circle of the outer contour and the maximum inscribed circle of the inner contour are calculated at the same time. The center coordinates of the inner and outer fitting circles are merged and located using the mean merging method to determine the unique center of mass position of the pipe mouth and the radius of the ring.

[0024] In the step 2, the nozzle target area is converted from the Cartesian coordinate system to the polar coordinate system, and the inner and outer boundary ranges required for the nozzle polar coordinate transformation are calculated by obtaining the minimum circumscribed circle and the maximum inscribed circle, so as to automatically adapt to heat pipes of different diameters and flatten the circular ring features of the nozzle end face into straight line features. If the nozzle has a roundness defect, it will be a non-linear feature after flattening.

[0025] In the step 2, a Gaussian filter is used to smooth the edges of the target features after the expansion process, the central skeleton of the flattened nozzle area is extracted by the Rosenfeld thinning algorithm, the endpoint deletion pruning method is used to retain the unique skeleton feature line, and the bilateral mean filter is selected to smooth the skeleton line to obtain the point coordinates and gradient angles of the skeleton line.

[0026] In the step three, according to the obtained skeleton line and contour edge, a width detection algorithm based on gradient direction edge search is used to obtain the width measurement point pair corresponding to each position of the skeleton line, the width value of the point pair is calculated by Euclidean distance, and the width tolerance value is set to distinguish the burr defect. If the measurement value of a certain continuous length is greater than the width tolerance value, it is determined that there is a burr defect on the end face of the pipe mouth.

[0027] In the step three, according to the obtained skeleton line and the average width of the pipe mouth, the position tolerance value is set to determine the out-of-round defect, two parallel inner and outer position tolerance lines are established, and the degree of wrinkling of the skeleton fluctuation is analyzed. If the skeleton line intersects with the position tolerance line, it is determined that the end face of the pipe mouth has an out-of-round defect. If any of the above defects is detected, the quality of the pipe mouth is judged to be unqualified.

[0028] Compared with the prior art, the present invention has the following advantages:

[0029] (1) The method of the present invention is designed for industrial application environments, and corresponding image preprocessing steps are designed to automatically capture the ROI of the pipe mouth features of copper pipes with different pipe diameters, and remove the reflective interference of other parts that may exist in the camera field of view; for the copper powder remaining after the copper pipe is cut, a corresponding filtering processing algorithm is designed to remove the interference of copper powder bright spots; when there are burrs or out-of-round defects on the pipe mouth, the lighting effect is deteriorated, and a corresponding contour detection algorithm is designed to determine whether the contour feature extraction is successful or not.

[0030] (2) The method of the present invention adopts the Tukey loss self-learning algorithm to weight the contour points when considering the defects of the pipe mouth, and eliminates the abnormal deviation points of the contour that may be in the burr or out-of-round defect area, thereby improving the positioning accuracy of the pipe mouth; it can automatically adapt to copper pipes of different outer diameters for precise positioning, and automatically select a reasonable target area to flatten the image; and a central skeleton extraction algorithm for the copper pipe mouth is designed, which can reflect the real pipe mouth skeleton line.

[0031] (3) The method of the present invention designs corresponding defect detection methods according to the defect characteristics of different copper tube orifices, identifies burr defects by the thickness of the tube orifice, and identifies out-of-round defects by the skeleton position. A width detection algorithm and position detection method for the copper tube orifice are designed. This method can be quickly and effectively applied in industrial production.

[0032] (4) The device of the present invention is suitable for horizontal detection of small pipes and simulates real industrial application scenarios. The pipe mouth visual detection mechanism can conveniently adjust the position of the camera and the light source, and the heat pipe horizontal positioning mechanism can conveniently, quickly and accurately replace the parts to be tested, effectively improving the coaxiality of the camera and the heat pipe, improving the shooting accuracy, and accelerating the detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of a heat pipe orifice defect detection method of the present invention;

[0034] Figure 2 It is a schematic diagram of the overall assembly of the heat pipe orifice defect detection device of the present invention;

[0035] Figure 3 It is a schematic diagram of the structure of the pipe mouth visual detection mechanism of the present invention;

[0036] Figure 4 It is a schematic diagram of the structure of the heat pipe horizontal positioning mechanism of the present invention;

[0037] Among them: 1 is the bottom plate of the general assembly, 2 is the pipe mouth visual inspection device, 3 is the heat pipe horizontal positioning device; 2-1 is the visual support plate, 2-2 is the fine-tuning slide mounting plate, 2-3 is the fine-tuning slide, 2-4 is the camera mounting plate, 2-5 is the camera, 2-6 is the lens, 2-7 is the light source, 2-8 is the light source fixing block, 2-9 is the light source fixing seat, 2-10 is the light source sliding plate; 3-1 is the positioning base, 3-2 is the heat pipe, 3-3 is the heat pipe fixer, 3-4 is the rear positioning block, and 3-5 is the positioning screw. DETAILED DESCRIPTION

[0038] In order to better understand the present invention, the present invention is further described below in conjunction with the accompanying drawings, but the embodiments of the present invention are not limited thereto. In addition, based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] The heat pipe orifice defect detection device based on machine vision is composed of two parts: a vision device and a computer system.

[0040] like Figure 2 As shown, the visual device is used to horizontally detect the pipe mouth defects at the front end of the incoming heat pipe, including a workbench 1, as shown Figure 3 The pipe mouth visual detection mechanism 2 shown and Figure 4 The heat pipe horizontal positioning mechanism 3 shown, the pipe mouth visual inspection mechanism 2 and the heat pipe horizontal positioning mechanism 3 are both installed on the workbench 1, and the heat pipe horizontal positioning mechanism 3 is arranged in front of the pipe mouth visual inspection mechanism 2, wherein the camera 2-5, the lens 2-6, the light source 2-7 and the heat pipe 3-2 are kept in the same straight line.

[0041] The details are as follows:

[0042] like Figure 3 The nozzle visual inspection mechanism 2 shown in the figure includes an image acquisition system and a visual instrument device. The image acquisition system includes a camera 2-5, a lens 2-6, a light source 2-7 and a light source controller. The visual instrument device includes a visual support plate 2-1, a fine-tuning slide mounting plate 2-2, a fine-tuning slide 2-3, a camera mounting plate 2-4, a light source fixing block 2-8, a light source fixing seat 2-9, and a light source sliding plate 2-10. The image acquisition system is installed on the visual instrument device, the lens 2-6 is installed on the camera 2-5, and the camera 2-5 is set on the camera mounting plate 2 -4, and connected to the computer system, the camera mounting plate 2-4 is fixed on the fine-tuning slide 2-3, the fine-tuning slide 2-3 is connected to the fine-tuning slide mounting plate 2-2 and the visual support plate 2-1, and is fixed to the workbench 1 through the visual support plate 2-1; the light source 2-7 is arranged on the light source fixing block 2-8, the light source fixing block 2-8 is connected to the light source sliding block 2-10 and the light source fixing seat 2-9, the light source sliding block 2-10 is installed on the inner side of the visual support plate 2-1, and the light source fixing seat 2-9 and the bottom of the light source fixing block 2-8 are arranged on the workbench 1.

[0043] In this embodiment, the nozzle visual inspection mechanism 2 is used to detect nozzle defects at the front end of the incoming heat pipe; wherein the image acquisition system is installed on the visual instrument device, and the image acquisition system uses the visual instrument device to make the camera 2-5, the lens 2-6 and the light source 2-7 on the same horizontal straight line, and the camera 2-5 and the lens 2-6 are fixed on the three-axis fine-tuning slide 2-3, and the installation errors are offset by fine-tuning the horizontal and vertical directions to ensure that the heat pipe is in the center of the camera imaging to reduce the image imbalance distortion, and at the same time adjust the front and rear positions of the camera 2-5 to fine-tune the imaging focal length so that the camera 2-5 images clearly, the camera 2-5 uses a black and white camera, the lens 2-6 uses a telecentric lens to reduce image distortion, and the light source 2-7 uses a ring-shaped white light source; the light source fixing block 2-8 is provided with upper and lower waist holes, which are connected to the back thread of the light source 2-7 The holes are matched to adjust the upper and lower positions of the light source 2-7. The light source fixing block 2-8 is provided with left and right waist holes, which cooperate with the threaded holes on the front side of the light source sliding plate 2-10 to adjust the left and right positions of the light source 2-7. The light source sliding plate 2-10 is nested between the left and right visual support plates 2-1 and can slide back and forth. The fine-tuning slide mounting plate 2-2 is provided with front and rear waist holes, which cooperate with the threaded holes on the light source sliding plate 2-10 to adjust the front and rear positions of the light source 2-7; adjust the distance between the camera 2-5, lens 2-6, light source 2-7 and heat pipe 3-2, and adjust the brightness of the light source 2-7 through the light source controller so that the captured image shows that the end face area of ​​the pipe mouth is bright and the background area is dark. The grayscale difference between the two is large, and the image of the pipe mouth shape characteristics can be extracted. The camera 2-5 transmits the acquired image to the computer system for image processing and defect recognition.

[0044] like Figure 4 The heat pipe horizontal positioning mechanism 3 shown includes a positioning base 3-1, a heat pipe 3-2, a heat pipe fixer 3-3, a rear positioning block 3-4, and a positioning screw 3-5; the positioning base 3-1 is fixed to the workbench 1 by a flat key and screws, the heat pipe fixer 3-3 and the rear positioning block 3-4 are arranged in a slot opened above the positioning base 3-1, the heat pipe 3-2 is placed on the V-groove on the heat pipe fixer 3-3, the rear port of the heat pipe 3-2 is positioned on the rear positioning block 3-4, and the positioning screw 3-5 fixes the heat pipe fixer 3-3 and the rear positioning block 3-4 through the threaded hole on the side of the positioning base 3-1.

[0045] In this embodiment, the heat pipe horizontal positioning mechanism 3 is used to horizontally fix heat pipes of different outer diameters and lengths, keep the heat pipe 3-2 and the lens 2-6 in the same straight line, simulate the actual incoming material working condition, and set the front port of the heat pipe 3-2 within the appropriate focal length of the image acquisition system in the visual inspection mechanism 2; the positioning base 3-1 is coaxial with the image acquisition system in the visual inspection mechanism 2 as a whole through the flat key fixing mechanism, and the width of the heat pipe fixture 3-3 is slightly smaller than the width of the slot on the positioning base 3-1. For copper pipes of different lengths, the heat pipe fixture 3 -3 can slide forward and backward to adjust the position, the heat pipe holder 3-3 is positioned against the side without threaded holes, and the heat pipe holder 3-3 is fixed by the positioning screw 3-5. The implementation method of the rear positioning block 3-4 is the same as that of the heat pipe holder 3-3; heat pipes 3-2 of different diameters can be placed on the V-groove of the heat pipe holder 3-4, and the heat pipes of the same length batch can be quickly positioned through the rear end of the heat pipe 3-2 and the rear positioning block 3-4. Black light-absorbing cloth needs to be attached to the surface of the heat pipe holder 3-3 and the rear positioning block 3-4 to avoid interference from reflections of parts.

[0046] The computer system is used to process the nozzle image and identify defects, including three modules: image preprocessing, feature extraction and defect detection, such as Figure 1 As shown, the detection method comprises the following steps:

[0047] S1. Selection and layout of lighting sources. A black and white camera is used as the camera, a telecentric lens is used to reduce image distortion, and a ring-shaped white light source is used as the light source. The pictures collected by the camera are transmitted to the computer system, and the software system performs image processing.

[0048] S2. Perform fast threshold segmentation on the collected heat pipe nozzle image f(x, y) to obtain the nozzle binary image g(x, y), T gray To set the segmentation threshold:

[0049]

[0050] Secondly, the center of gravity coordinates are obtained according to the center of gravity calculation formula. The grayscale center of gravity (x0, y0) of the target area S in the grayscale image g(i, j) is

[0051]

[0052] Among them, u and j represent the x-coordinate and y-coordinate of the pixel point respectively, and N is the number of pixels in the area.

[0053] The circular ROI is set to be slightly larger than the annular area, the original image is reduced, and interference sources in the non-ROI area of ​​the image are removed. Copper tubes with different outer diameters can be automatically cropped, de-noised and enlarged for display.

[0054] S3. Use median filtering and mean filtering to preprocess the grayscale image and transform the original image g(x, y) into the filtered image f(x, y):

[0055]

[0056] Where: S xy is a window with a center point at (x, y) and a size of m×n.

[0057] The OTSU threshold segmentation method is selected to perform binary segmentation on the processed nozzle image. The Euler number feature is used to determine whether the extracted shape area is the nozzle end face. If the target area is extracted correctly, the inner and outer contour edges of the nozzle are obtained, and the roundness of the inner and outer contours is calculated.

[0058] S4. Using the least squares circle fitting method based on Tukey algorithm, the inner contour points and outer contour points of the nozzle are fitted with circles respectively. The corresponding parameter equation of the fitting circle curve is:

[0059] (x-x0) 2 +(y-y0) 2 =R 2

[0060] Where (x0, y0) is the coordinate of the center of the circle, and R is the radius of the circle. For circle fitting based on nonlinear least squares, the optimization objective function C is:

[0061]

[0062] Where: (x i ,y i ) is the coordinate of the feature point on the arc, and N is the number of feature points involved in the fitting calculation.

[0063] The minimum circumscribed circle of the outer contour and the maximum inscribed circle of the inner contour are calculated, and the center coordinates of the inner and outer fitting circles are merged and located using the mean merging method to determine the unique nozzle center of mass position and the ring radius.

[0064] S5. The nozzle target area is converted from the Cartesian coordinate system f(x, y) to the polar coordinate system f(ρ, θ), where ρ is the extreme value and θ is the polar angle. The transformation relationship is as follows:

[0065]

[0066] By obtaining the minimum circumscribed circle radius r outer and the maximum inscribed circle r inner , substitute into the following formula:

[0067] R OB =2r outer -R

[0068] R IB =2r inner -R

[0069] Calculate the outer circle boundary radius R required for the polar coordinate transformation of the nozzle OB and the inner circle radius R IB , automatically adapt to heat pipes of different diameters. Flatten the circular ring feature of the pipe end face into a straight line feature. If the pipe end has a round defect, it will be a non-linear feature after flattening.

[0070] S6. Use a Gaussian filter to smooth the edges of the target features after dilation, extract the central skeleton of the flattened nozzle area through the Rosenfeld refinement algorithm, use the endpoint deletion pruning method to retain the unique skeleton feature line, select the bilateral mean filter to smooth the skeleton line, and obtain the point coordinates and gradient angle of the skeleton line. The bilateral filter integrates the gradient angle information of each T point on both sides of the skeleton line, and uses the previously updated data during the window sliding process to obtain the optimal estimate D(x) of the gradient angle of the current point.

[0071] D(x)=∑D(xT:x+T) / (2T+1)

[0072] S7. According to the obtained skeleton line and contour edge, a width detection algorithm based on gradient direction edge search is used to obtain the width measurement point pair corresponding to each position of the skeleton line, and the width value of the point pair is calculated by the Euclidean distance D

[0073]

[0074] Set the width tolerance value T width To identify burr defects, if the continuous measurement value of a certain length is greater than the width tolerance value, it is determined that there is a burr defect on the end face of the pipe.

[0075] S8. Set the position tolerance value T according to the obtained skeleton line and the average width of the pipe mouth posTo identify the out-of-round defect, establish two parallel position tolerance lines inside and outside, and analyze the degree of wrinkles in the skeleton fluctuation. If the skeleton line intersects with the position tolerance line, it is determined that the end face of the pipe mouth has an out-of-round defect. If any of the above defects is detected, the quality of the pipe mouth is judged to be unqualified.

Claims

1. A heat pipe orifice defect detection device based on machine vision, characterized in that: including computer systems and visual devices; The visual device is used for horizontally detecting the pipe mouth defects at the front end of a copper pipe, and comprises a workbench (1), a pipe mouth visual detection mechanism (2) and a heat pipe horizontal positioning mechanism (3), wherein the pipe mouth visual detection mechanism (2) and the heat pipe horizontal positioning mechanism (3) are both installed on the workbench (1), and the heat pipe horizontal positioning mechanism (3) is arranged in front of the pipe mouth visual detection mechanism (2); The heat pipe horizontal positioning mechanism (3) is used to horizontally fix heat pipes of different outer diameters and lengths, and keep the heat pipe and the lens of the pipe mouth visual detection mechanism (2) in the same straight line. The pipe mouth visual detection mechanism (2) collects pipe mouth image information and transmits it to the computer system; The computer system is used to process the nozzle image and identify defects, including three modules: image preprocessing, feature extraction and defect detection; The pipe mouth visual inspection mechanism (2) comprises an image acquisition system and a visual instrument device; The image acquisition system is installed on the visual instrument device; The image acquisition system includes a camera (2-5), a lens (2-6), a light source (2-7) and a light source controller; The visual instrument device comprises a visual support plate (2-1), a fine-tuning slide mounting plate (2-2), a fine-tuning slide (2-3), a camera mounting plate (2-4), a light source fixing block (2-8), a light source fixing seat (2-9), and a light source sliding plate (2-10); The lens (2-6) is mounted on the camera (2-5), the camera (2-5) is set on the camera mounting plate (2-4) and connected to the computer system, the camera mounting plate (2-4) is fixed on the fine-tuning slide (2-3), the fine-tuning slide (2-3) is mounted on the vision support plate (2-1) through the fine-tuning slide mounting plate (2-2), and the vision support plate (2-1) is fixed on the workbench (1); The light source (2-7) is mounted on the light source sliding plate (2-10) via the light source fixing block (2-8); the light source sliding plate (2-10) is slidably mounted on the light source fixing seat (2-9); the light source sliding plate (2-10) is mounted on the inner side of the visual support plate (2-1); the bottoms of the light source fixing seat (2-9) and the light source fixing block (2-8) are arranged on the workbench (1); The heat pipe horizontal positioning mechanism (3) comprises a positioning base (3-1), a heat pipe (3-2), a heat pipe fixer (3-3), a rear positioning block (3-4), and a positioning screw (3-5); the positioning base (3-1) is fixed to the workbench (1) by means of a flat key and a screw, the heat pipe fixer (3-3) and the rear positioning block (3-4) are arranged in a slot hole opened above the positioning base (3-1), the heat pipe (3-2) is placed on the V-shaped groove on the heat pipe fixer (3-3), the rear positioning block (3-4) positions the rear port of the heat pipe (3-2), and the heat pipe fixer (3-3) and the rear positioning block (3-4) are fixed to the base (3-1) by means of the positioning screw (3-5); The camera (2-5), lens (2-6), light source (2-7) and heat pipe (3-2) are kept in the same straight line.

2. A heat pipe orifice defect detection method based on machine vision, characterized in that: The heat pipe orifice defect detection device based on machine vision according to claim 1 comprises the following steps: Step 1: The image preprocessing module includes image acquisition, image cropping and contour detection parts; the image captured by the camera (2-5) in the pipe mouth visual detection mechanism (2) is transmitted to the computer system, and the grayscale centroid method is used to crop the area of ​​interest containing the pipe mouth features, and the shape area and internal and external contours of the pipe mouth end face are obtained through morphological processing; The collected pipe mouth images are quickly segmented by global threshold, and the annular area with a suitable circumscribed circle radius is selected. Then, the grayscale centroid of the area is calculated, and the circular ROI is set slightly larger than the annular area. The original image is reduced and the interference source in the non-ROI area of ​​the image is removed. For copper pipes with different outer diameters, the cropping, denoising and enlarged display can be automatically performed. The grayscale image is preprocessed by median filtering and mean filtering, and the OTSU threshold segmentation method is selected to perform binary segmentation on the processed nozzle image. The Euler number feature is used to determine whether the extracted shape area is the nozzle end face. If the target area is extracted correctly, the inner and outer contour edges of the nozzle are obtained, and the roundness of the inner and outer contours is calculated. Step 2: The feature extraction module includes target positioning, target flattening, and skeleton extraction. The fitting circle of the inner and outer contours is calculated using the least squares circle fitting method, and the center of mass position of the nozzle and the ring size detection are completed by combining the mean merging method. The target area is transformed into polar coordinates to obtain the nozzle flattened image. The nozzle center skeleton is extracted through the thinning algorithm and smoothing processing. Step 3: The defect detection module includes two parts: burr defect detection and out-of-round defect detection. The skeleton gradient directional edge search algorithm is used to obtain the measured values ​​of the pipe mouth width at different positions, and the width tolerance value is set to distinguish burr defects. The skeleton line fluctuation characteristics are analyzed, and the position tolerance value is set to distinguish out-of-round defects. If any of the above defects is detected, the pipe mouth quality is judged to be unqualified.

3. The heat pipe orifice defect detection method based on machine vision according to claim 2 is characterized in that: In the step 2, the least squares circle fitting method based on the Tukey algorithm is used to fit the inner contour points and the outer contour points of the pipe mouth respectively, and the minimum circumscribed circle of the outer contour and the maximum inscribed circle of the inner contour are calculated at the same time. The center coordinates of the inner and outer fitting circles are merged and located using the mean merging method to determine the unique center of mass position of the pipe mouth and the radius of the ring.

4. The heat pipe orifice defect detection method based on machine vision according to claim 2 is characterized in that: In the step 2, the nozzle target area is converted from the Cartesian coordinate system to the polar coordinate system, and the inner and outer boundary ranges required for the nozzle polar coordinate transformation are calculated by obtaining the minimum circumscribed circle and the maximum inscribed circle, so as to automatically adapt to heat pipes of different diameters and flatten the circular ring features of the nozzle end face into straight line features. If the nozzle has a roundness defect, it will be a non-linear feature after flattening.

5. The heat pipe orifice defect detection method based on machine vision according to claim 2 is characterized in that: In the step 2, a Gaussian filter is used to smooth the edges of the target features after the expansion process, the central skeleton of the flattened nozzle area is extracted by the Rosenfeld thinning algorithm, the endpoint deletion pruning method is used to retain the unique skeleton feature line, and the bilateral mean filter is selected to smooth the skeleton line to obtain the point coordinates and gradient angles of the skeleton line.

6. The heat pipe orifice defect detection method based on machine vision according to claim 2 is characterized in that: In the step three, according to the obtained skeleton line and contour edge, a width detection algorithm based on gradient direction edge search is used to obtain the width measurement point pair corresponding to each position of the skeleton line, the width value of the point pair is calculated by Euclidean distance, and the width tolerance value is set to distinguish the burr defect. If the measurement value of a certain continuous length is greater than the width tolerance value, it is determined that there is a burr defect on the end face of the pipe mouth.

7. The heat pipe orifice defect detection method based on machine vision according to claim 2 is characterized in that: In the step three, according to the obtained skeleton line and the average width of the pipe mouth, the position tolerance value is set to determine the out-of-round defect, two parallel inner and outer position tolerance lines are established, and the degree of wrinkling of the skeleton fluctuation is analyzed. If the skeleton line intersects with the position tolerance line, it is determined that the end face of the pipe mouth has an out-of-round defect. If any of the above defects is detected, the quality of the pipe mouth is judged to be unqualified.

Images