Aluminum alloy frame weld defect detection device based on machine vision

Abstract

This invention discloses a machine vision-based aluminum alloy frame weld defect detection device, belonging to the field of weld defect detection technology, including a detection instrument. The invention effectively eliminates interference from image noise and uneven lighting on the detection results by filtering and analyzing images through a defect analysis module. It achieves automatic identification of slag inclusion defects by judging contour regularity through benchmark point positioning and extension line length fluctuation. Simultaneously, it combines the grayscale value range of various defects and the aspect ratio characteristics of cracks in historical data to achieve accurate classification of defect types through contour geometric shape comparison. Finally, it records and displays the quantity and location of various defects in real time, triggering a buzzer alarm when the total number of defects reaches a preset threshold. This defect analysis mechanism, through multi-level progressive analysis from pixel-level statistics to contour-level recognition, significantly improves the accuracy and automation of defect detection, effectively reducing the workload and subjective errors of manual interpretation.

Description

Technical Field

[0001] This invention relates to the field of weld defect detection technology, and more particularly to a machine vision-based weld defect detection device for aluminum alloy frames. Background Technology

[0002] Aluminum alloy frames, with their advantages of light weight, high strength, corrosion resistance, and good thermal conductivity, are widely used in aerospace, automotive manufacturing, construction engineering, electronic equipment, and many other fields. The quality of their welds directly determines the structural stability, load-bearing capacity, and service life of the aluminum alloy frame. During the welding process, defects such as cracks, porosity, slag inclusions, and incomplete penetration can easily occur due to factors such as welding technology, operating methods, and material properties. These defects can severely weaken the structural strength of the aluminum alloy frame and even cause safety hazards. Therefore, defect detection of aluminum alloy frame welds is necessary. Existing machine vision-based aluminum alloy frame weld defect detection devices require operators to hold the detector in one hand and the detection camera in the other, working together to perform machine recognition of the welds. When dealing with frames with many welds and high weld positions, prolonged hand-holding can easily lead to arm pain and fatigue, and the detector and camera are also prone to shaking, making it difficult for the camera to accurately align with the weld area, thus affecting the clarity of the detection image and the detection accuracy. Therefore, improvements are needed to address the aforementioned issues. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by proposing a machine vision-based aluminum alloy frame weld defect detection device.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a machine vision-based aluminum alloy frame weld defect detection device, including a detector, a data cable inserted into one side of the top of the detector, a detection camera fixedly connected to one end of the data cable, a display screen on one end of the detector, a control panel at the bottom of the display screen, and a fixing component for detachably fixing the detector to the aluminum alloy frame on the other end of the detector. The control panel includes a defect analysis module. The defect analysis module performs image fusion and anti-shake processing on N consecutively acquired weld seam images at the same time: using the first frame as a reference, it calculates the motion transformation matrix of each frame and aligns them in reverse; it evaluates the sharpness by calculating the gradient sum of each frame, and finally fuses all aligned images according to their weights to obtain a clear fused image; it performs grayscale processing and blocks on the image, performs statistical analysis on the grayscale values ​​within each block, removes outliers, and selects the image with the fewest outlier blocks as the analysis image; it draws the contours of consecutive outlier blocks on the analysis image, and determines the defect type through geometric feature analysis; finally, it records the number and location of defects and displays them in real time, triggering a buzzer alarm when a threshold is reached.

[0005] Preferably, the anti-shake analysis steps of the defect analysis module are as follows: M1: Continuously collected data at the same time. Motion alignment of the weld seam images is performed, using the first frame image as a reference, and the calculation of the... Motion transformation matrix of frame image relative to the first frame image And utilize its inverse transformation Perform spatial transformation on each frame of the image to obtain the aligned image. ; M2: Weighted fusion is performed on the aligned frames. First, the gradient sum Sᵢ of each frame is calculated as a sharpness evaluation index. The larger the gradient sum, the sharper the image. The calculation formula is as follows: , For the image in coordinates Gradient magnitude at; M3: Assign weights based on the sharpness of each frame. Higher resolution images receive greater weight, and the sum of all weights is 1. All aligned images are then weighted and fused according to their assigned weights to obtain a clear, merged image free from jitter and blur. .

[0006] Preferably, the image determination steps of the defect analysis module are as follows: Q1: Convert the fused image to grayscale and divide it into segments according to a preset size. Given several grayscale blocks of the same size, within each grayscale block... Statistical analysis was performed on the grayscale values ​​of each pixel to calculate the mean. and standard deviation Set the grayscale value fluctuation range to Mark grayscale values ​​outside the range as outliers and record the number of outliers. ; Q2: If If the grayscale block data is found to be abnormal, it will be re-detected. This is a preset proportional coefficient; if Then, after removing outliers, the mean of the remaining gray values ​​is recalculated. The grayscale value is used as the representative grayscale value of the grayscale block; if it is still determined to be abnormal after re-detection, the grayscale block is marked as an abnormal grayscale block; after all grayscale blocks have been determined, the number of abnormal grayscale blocks in each frame image is counted. Select The image with the smallest value is used as the analysis image.

[0007] Preferably, the defect determination steps of the defect analysis module are as follows: R1: Obtain the serial numbers of the abnormal grayscale blocks in the analysis image, draw the corresponding coordinate points in the coordinate system and connect the adjacent points to form a closed contour; record the length of the contour in the horizontal and vertical directions, and take the midpoint of the line connecting the two farthest points of the contour or the intersection of multiple lines as the reference point; draw extension lines with equal angle intervals outward from the reference point, mark the intersection points of the extension lines and the contour, and calculate the distance between the intersection points and the reference point. If the length difference between adjacent extension lines fluctuates within the preset range, the contour is determined to be a regular shape; otherwise, it is determined to be an irregular shape and marked as a slag inclusion defect. R2: Retrieves the grayscale range of various defects and the aspect ratio of cracks from historical data. If the outline meets the circular equation, it is determined to be a porosity defect. If the outline is long and the aspect ratio is within the crack feature range, it is determined to be a crack defect. Otherwise, it is determined to be an incomplete penetration defect. Records the number of various defects and the position of the reference point, displays them on the display screen in real time, and triggers a buzzer alarm when the total number of defects reaches the preset threshold.

[0008] Preferably, the fixing assembly includes a storage slot formed on the outer wall of one end of the detector. Three vertical rods are fixedly connected to the storage slot at equal intervals. Two lifting plates are symmetrically sleeved on the three vertical rods. A hinge plate is hinged to the outward-facing end face of the lifting plate. Two guide rods are symmetrically fixed to the end of the hinge plate away from the lifting plate. A horizontally placed L-shaped plate is slidably sleeved on the two guide rods. An installation hole is opened between the two guide rods on the hinge plate. A pull-back spring is provided in the installation hole. One end of the pull-back spring is fixedly connected to the outer wall of the L-shaped plate. Multiple first springs sleeved on the vertical rods are fixedly connected between the opposite faces of the two lifting plates. The middle section of the first spring is fixedly connected to the outer wall of the vertical rod.

[0009] Preferably, the lifting plate has a U-shaped cross-section when viewed from above. Two sets of spring blocks are symmetrically arranged on the inner walls of both sides of the lifting plate. Unlocking slots communicating with the spring blocks are symmetrically opened on both sides of the top surface of the lifting plate. The unlocking slots are connected to the spring blocks. An unlocking slider is slidably arranged in the unlocking slot. One end of the unlocking slider is fixedly connected to the locking block of the spring block. The outer walls of both sides of the hinge plate have slots that are adapted to the spring blocks.

[0010] Preferably, a positioning hole is provided in the middle of the outward-facing end of the L-shaped plate, and spring-locking steel balls are provided at both the upper and lower ends of the inner wall of the storage groove, with the spring-locking steel balls being adapted to the positioning hole.

[0011] Preferably, an elastic rope handle is fixed to one side of the outer wall of the detector, and an angled insertion hole for inserting a detection camera is provided on the other side of the outer wall of the detector.

[0012] Preferably, the detector has a data interface at the lower end of the angled insertion hole, and a charging port at the lower end of the data interface.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. By cooperating with the vertical rod and the lifting plate in the fixed assembly, the distance between the two lifting plates can be flexibly adjusted to accommodate the height of the horizontal bar in the aluminum alloy frame, improving the height adaptability of the device and enabling it to stably clamp horizontal bars of different heights. Furthermore, by cooperating with the L-shaped plate, guide rod, and return spring, the extension distance of the L-shaped plate can be flexibly adjusted to accommodate the width of the horizontal bar in the aluminum alloy frame, improving the width adaptability of the device and enabling it to fully adapt to horizontal bars of different sizes. Simultaneously, by cooperating with the spring clip and the hinge plate, the hinge plate can be fixed to the lifting plate to maintain a horizontal state, preventing the hinge plate from shaking or shifting during clamping, thus improving... This design ensures the stability of the fixed structure, enabling a secure connection between the hinge plate and the lifting plate, and guaranteeing reliable clamping. Finally, through the cooperation of the lifting plate, hinge plate, and L-shaped plate, the detector can be quickly and detachably fixed to the horizontal bar of the aluminum alloy frame, eliminating the need for operators to hold it with both hands. This frees up one hand, reduces physical exertion, and improves operational convenience. It also allows for stable fixing of the detector, enabling operators to focus on adjusting the camera to align with the weld seam. Ultimately, this solves the problem of arm pain caused by prolonged hand-holding of existing detectors, which affects image clarity and accuracy. This improves detection accuracy and efficiency while reducing operational safety hazards. 2. The defect analysis module performs motion alignment on N consecutively acquired weld seam images. By calculating the motion transformation matrix and its inverse transformation, precise spatial position correction is performed on each frame to eliminate image shift and rotation caused by jitter. Furthermore, the sum of gradients for each frame is calculated as a sharpness evaluation index; a larger gradient indicates richer image details and higher sharpness. Based on this, fusion weights are assigned, giving higher-sharp images a greater weight contribution during the fusion process. Finally, all aligned images are weighted and fused to obtain a high-quality, clear image free from jitter blur. This anti-shake processing mechanism fully integrates complementary information from multiple frames, effectively suppressing random jitter interference and significantly improving the image's signal-to-noise ratio and detail representation. This provides a clear and reliable image data foundation for subsequent high-precision defect identification, avoiding missed and false detections caused by image blur. 3. By filtering and analyzing images through the defect analysis module, interference from image noise and uneven lighting on the detection results is effectively eliminated. Furthermore, contour drawing and geometric feature analysis are performed on continuously distributed abnormal grayscale blocks. The regularity of the contour is judged by benchmark point positioning and fluctuations in the length of the extension line, achieving automatic identification of slag inclusion defects. Simultaneously, by combining the grayscale value range of various defects and the aspect ratio characteristics of cracks in historical data, accurate classification of defect types is achieved through geometric shape comparison of the contours. Finally, the number and location of various defects are recorded and displayed in real time, triggering a buzzer alarm when the total number of defects reaches a preset threshold. This defect analysis mechanism, through multi-level progressive analysis from pixel-level statistics to contour-level recognition, significantly improves the accuracy and automation of defect detection, effectively reducing the workload and subjective errors of manual interpretation, and providing reliable technical support for the quality assessment of aluminum alloy frame welds. Attached Figure Description

[0014] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a first-view schematic diagram of the overall structure proposed in this invention; Figure 2 This is a second-view schematic diagram of the overall structure proposed in this invention; Figure 3 This is an enlarged schematic diagram of the overall structure of the storage slot proposed in this invention; Figure 4 The present invention proposes Figure 2 Schematic diagram of the cross-sectional structure at point A in the middle; Figure 5 This is a flowchart of the system proposed in this invention.

[0015] The following items are numbered in the diagram: 1. Detector; 2. Data cable; 3. Detection camera; 4. Storage slot; 5. Vertical rod; 6. Lifting plate; 7. Hinge plate; 8. L-shaped plate; 9. Spring-locking steel ball; 10. Unlocking slider; 11. Elastic rope handle; 12. Spring latch; 13. First spring. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0017] Example 1: See Figures 1 to 4The machine vision-based aluminum alloy frame weld defect detection device of this invention includes a detector 1, a data cable 2 inserted into one side of the top of the detector 1, a detection camera 3 fixedly connected to one end of the data cable 2, a display screen on one end of the detector 1, a control panel at the bottom of the display screen, and a fixing component for detachably fixing the detector 1 to the aluminum alloy frame on the other end of the detector 1. The data cable 2 is an aviation connector cable, which has a fast transmission rate, strong anti-interference ability, and can stably transmit image data. The detection camera 3 is a CCD industrial camera, which has high resolution, clear imaging, and can accurately acquire weld images. The detector 1, as the core of the device, realizes the processing, analysis, and defect identification of weld images, and at the same time provides the installation foundation for various components. The data cable 2 connects the detection camera 3. Image data transmission between the detector 1 and the inspection device 3 is ensured to be smooth; the inspection camera 3 is responsible for acquiring images of the aluminum alloy frame welds, providing data support for defect detection; the display screen is used to display inspection data and defect identification results, facilitating real-time viewing by staff; the control panel is used for equipment parameter debugging and start / stop control, improving operational convenience; the fixing components enable the detector 1 to be detachably fixed to the aluminum alloy frame, freeing the operator from holding it with one hand and avoiding fatigue and equipment shaking caused by prolonged handheld operation; the above structures together form the basic frame of the device, providing stable structural support for the entire process of aluminum alloy frame weld defect detection, and initially solving the problems of inconvenient handheld operation and easily affected detection accuracy of existing devices; the fixing components include those installed on the detector... 1. A storage groove 4 is located on one end of the outer wall. Three vertical rods 5 are fixedly connected to the storage groove 4 at equal intervals. Two lifting plates 6 are symmetrically sleeved on the three vertical rods 5. A hinge plate 7 is hinged to the outward-facing end of the lifting plate 6. Two guide rods are symmetrically fixed to the end of the hinge plate 7 away from the lifting plate 6. A horizontally placed L-shaped plate 8 is slidably sleeved on the two guide rods. The hinge plate 7 has an installation hole between the two guide rods. A pull-back spring is installed in the installation hole. One end of the pull-back spring is fixedly connected to the outer wall of the L-shaped plate 8. Multiple first springs 13 are fixedly sleeved on the vertical rods 5 between the opposite faces of the two lifting plates 6. The middle section of the first spring 13 is fixedly connected to the outer wall of the vertical rod 5. The vertical rods 5 are made of No. 45 steel, which has high hardness and strong load-bearing capacity and can stably support the lifting plates 6. The lifting plates 6 are made of aluminum alloy, which is lightweight. The structure is convenient and robust, with a U-shaped cross-section when viewed from above, suitable for installation with vertical rod 5; the hinge plate 7 is made of stainless steel, which is corrosion-resistant, tough, and can rotate flexibly; the guide rod is made of stainless steel with a smooth surface, facilitating the sliding of the L-shaped plate 8; the L-shaped plate 8 is made of Q235 steel plate, bent and formed, with strong clamping stability; the return spring and the first spring 13 are cylindrical helical springs; the storage slot 4 can store the fixed components, avoiding damage to the components and space occupation when not in use; the vertical rod 5 provides a sliding guide for the lifting plate 6, ensuring smooth lifting; the height of the lifting plate 6 can be adjusted along the vertical rod 5 to match the height of the horizontal bar of the aluminum alloy frame; the hinge plate 7 can rotate around the lifting plate 6 to realize the unfolding and storage of the fixed components; the guide rod provides a sliding guide for the L-shaped plate 8, ensuring smooth extension and retraction;The pull-back spring drives the L-shaped plate 8 to reset and tighten, achieving a stable clamping of the horizontal bar; the first spring 13 can fix the position of the lifting plate 6, while buffering the impact force during clamping, improving the fixing stability, and thus realizing the function of the fixing component to adapt to horizontal bars of different heights and to clamp them stably; the lifting plate 6 has a U-shaped cross-section when viewed from above, and two sets of spring blocks 12 are symmetrically arranged on the inner walls of both sides of the lifting plate 6. The top surface of the lifting plate 6 has symmetrical unlocking slots on both sides that connect to the spring blocks 12. The unlocking slots are connected to the spring blocks 12, and an unlocking slider 10 is slidably arranged in the unlocking slot. One end of the unlocking slider 10 is fixedly connected to the block of the spring block. The outer walls of both sides of the hinge plate 7 have openings... A slot is adapted to the spring latch 12; the spring latch 12 is made of plastic, with good toughness and a firm engagement, which can fix the hinge plate 7; the unlocking slider 10 is made of ABS plastic with anti-slip texture on the surface for easy operation; the spring latch 12 is inserted into the slot of the hinge plate 7, which can fix the hinge plate 7 in a horizontal state, preventing the hinge plate 7 from shaking or shifting during clamping, and improving the stability of the fixing structure; the unlocking slot provides sliding space for the unlocking slider 10, which can drive the spring latch 12 to retract, making it easy to unlock and store the hinge plate 7, improving the convenience of operation, and thus realizing the functions of stable fixation and convenient unlocking of the hinge plate 7.

[0018] In this invention, a positioning hole is opened in the middle of the outward-facing end of the L-shaped plate 8, and spring-locking steel balls 9 are provided at both the upper and lower ends of the inner wall of the storage groove 4. The spring-locking steel balls 9 are adapted to the positioning hole. The spring-locking steel balls 9 are made of stainless steel balls with a smooth surface. They work with the built-in spring to achieve elastic extension and contraction, resulting in a good locking effect. When the hinge plate 7 is stored, the spring-locking steel balls 9 are inserted into the positioning hole of the L-shaped plate 8 to limit and fix the hinge plate 7 in the stored state, preventing the hinge plate 7 from shaking randomly after storage, protecting the fixed components, and ensuring a compact structure after storage, thus improving the portability of the device. An elastic rope handle 11 is fixed to one side of the outer wall of the detector 1, and an oblique insertion hole for inserting the detection camera 3 is opened on the other side of the outer wall of the detector 1. The elastic rope handle 11 is made of nylon elastic rope, which has good toughness and strong load-bearing capacity, making it easy to hold and move. The elastic rope handle 11 facilitates... The operator holds the mobile detector 1, which is especially suitable for high-altitude or mobile testing scenarios, improving operational convenience. The angled insertion port allows for the insertion of a fixed testing camera 3 when not in use, avoiding collisions and dust contamination caused by exposed placement of the camera 3, protecting it and extending its lifespan. A data interface is located at the lower end of the angled insertion port, and a charging port is located below the data interface. The data interface uses a USB interface, offering fast transmission speeds and facilitating the export of testing data. The charging port uses a Type-C interface, ensuring stable charging and quickly replenishing the power of the detector 1. The data interface facilitates the export of testing data from the detector 1 to computers and other devices, enabling subsequent data archiving, analysis, and traceability. The charging port allows for timely charging of the detector 1, preventing testing interruptions due to insufficient power, ensuring continuous testing work, and improving testing efficiency.

[0019] Working principle: When using this invention, if the detector 1 needs to be fixed on the horizontal bar of the aluminum alloy frame, the operator first unlocks the spring locking ball 9, causing it to disengage from the positioning hole of the L-shaped plate 8, releasing the storage limit on the hinge plate 7. Then, the hinge plate 7 is rotated 90° around the axis to release it and adjust it to a horizontal state. At this time, the spring clip 12 resets and engages with the slots on both sides of the hinge plate 7, fixing the hinge plate 7 in a horizontal state. Then, according to the height of the horizontal bar of the aluminum alloy frame, the positions of the two lifting plates 6 on the vertical bar 5 are adjusted to match the height of the horizontal bar. The elastic force of the first spring 13 is used to keep the two lifting plates 6 relatively stable. Then, according to the width of the horizontal bar, the L-shaped plate 8 is pulled to slide along the guide bar. The pullback spring is stretched, locking the L-shaped plate 8 on one side of the horizontal bar. After releasing the L-shaped plate 8, the pullback spring... The spring resets, causing the L-shaped plate 8 to tighten and firmly clamp the horizontal bar. After fixing, the operator holds the inspection camera 3, adjusts its angle to align with the weld area to be inspected, and starts the inspection instrument 1. The inspection camera 3 acquires images of the weld, which are transmitted to the inspection instrument 1 via the data cable 2. The inspection instrument 1 processes and analyzes the images to identify whether there are defects such as cracks, porosity, and slag inclusions in the weld. At the same time, the inspection data and results are displayed on the screen, allowing the operator to view the inspection status in real time and complete the comprehensive inspection of all welds. After the inspection is completed, the unlocking slider 10 is pushed, which in turn causes the spring block 12 to retract and disengage from the slot. Then, the hinge plate 7 is rotated 90° in the opposite direction around the axis for storage. At this time, the spring locking ball 9 is inserted into the positioning hole of the L-shaped plate 8, limiting and fixing the hinge plate 7 in the stored state. Thus, the device is used successfully.

[0020] Example 2: See Figure 5 The control panel has a defect analysis module. The defect analysis module performs fusion and image stabilization processing on N consecutively acquired weld seam images at the same time: using the first frame as a reference, it calculates the motion transformation matrix of each frame and aligns them in reverse; it evaluates the sharpness by calculating the gradient sum of each frame, and finally fuses all aligned images according to their weights to obtain a clear fused image; it performs grayscale processing and block segmentation on the image, performs statistical analysis on the grayscale values ​​within each block, removes outliers, and selects the image with the fewest outlier blocks as the analysis image; it draws the contours of consecutive outlier blocks on the analysis image, and determines the defect type through geometric feature analysis—irregular contours indicate slag inclusions, circular contours indicate porosity, elongated contours with a length-to-width ratio that meet the characteristics indicate cracks, otherwise it indicates incomplete penetration; finally, it records the number and location of defects and displays them in real time, triggering a buzzer alarm when a threshold is reached; Let the data obtained at the corresponding time be... Frame image, number The motion transformation of the frame image relative to the first frame image is Then the aligned first The frame image is: , for inverse transform, Statement No. Frame image in coordinates The color value at that location, For the first Frame of the original image; Assign weights to each frame The weights satisfy The clear image after fusion is: Weight , No. Frame image gradient summation , The gradient magnitude; A cleared fused image is obtained through the above multi-frame fusion algorithm. The system then uses this image as the basis for subsequent high-precision defect identification and moves on to the image analysis stage.

[0021] The real-time acquired image data is processed into grayscale, and the grayscale image is segmented according to the size of pixel blocks, resulting in... The image data consists of several identical grayscale blocks, numbered according to their row and column numbers on the grayscale image. The acquired image data is sorted by acquisition time, and the corresponding numbered grayscale blocks within a single image acquired at the same time are further analyzed. Average the gray values and standard deviation The calculation, and the mean obtained from the calculation. and standard deviation Range of grayscale data The setting compares the corresponding grayscale value data with the corresponding grayscale value fluctuation range, marks the corresponding grayscale value data that is outside the fluctuation range as an outlier, and records the number of outliers. ; like If the grayscale data is abnormal, the grayscale data will be detected again. This is a preset proportional coefficient; if If outliers are removed, the remaining grayscale data after outlier removal is averaged. The calculation, and the mean obtained from the calculation. This serves as the grayscale value data detected at the corresponding time. If, upon re-detection, the grayscale value data is still deemed abnormal, then the corresponding numbered grayscale block is determined to be abnormal. After determining the grayscale values ​​of all numbered grayscale blocks on the grayscale image, the number of abnormal grayscale blocks is calculated. Record and take the quantity The image with the smallest grayscale value is the image to be analyzed; The process involves identifying and determining the numbers of abnormal grayscale blocks in the analyzed image, plotting the corresponding coordinates of these blocks in a binary coordinate system, connecting adjacent consecutive coordinate points, and marking the closed contours formed by these connections. The lengths of the marked contours in the horizontal and vertical directions are recorded. A line is drawn connecting the two furthest points in each direction. If it's a single line, the midpoint is used as the reference point; if it's multiple lines, the point with the most intersections is used. Extension lines are drawn outwards from the reference point, with each adjacent extension line having the same angle. The intersections of the extension lines with the closed contour are marked, and the lengths between the extension lines and the reference point are calculated. If the length difference between adjacent extension lines is within a preset range, the contour is considered a regular shape; otherwise, it's considered an irregular shape, indicating slag inclusions at the weld location corresponding to the contour. Historical data is retrieved, and the grayscale values ​​at the locations where defects are identified are extracted to determine the range of grayscale value variation for the corresponding defects. The aspect ratio range of the crack is then determined based on the historical data. If the coordinates of the intersection point of the extended line and the closed contour satisfy the circular equation, it indicates that porosity exists at the weld location corresponding to the contour. If the corresponding contour is elongated and its aspect ratio is within the range of the crack's aspect ratio, it indicates that a crack exists at the weld location corresponding to the contour. Conversely, it indicates that incomplete penetration exists at the weld location corresponding to the contour. Record the quantity and baseline of each defect type and display this information on the control panel in real time; and trigger an alarm and issue a buzzer to notify staff when the total number of all defect types reaches the preset threshold.

[0022] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A machine vision-based aluminum alloy frame weld defect detection device, comprising a detector (1), a data cable (2) inserted into one side of the top of the detector (1), a detection camera (3) fixedly connected to one end of the data cable (2), a display screen provided on one end of the detector (1), and a control panel provided at the lower end of the display screen, characterized in that: The other end face of the detector (1) is provided with a fixing component for detachably fixing the detector (1) to the aluminum alloy frame; The control panel includes a defect analysis module. The defect analysis module performs fusion and anti-shake processing on N consecutively acquired weld seam images at the same time: taking the first frame as the reference, it calculates the motion transformation matrix of each frame image and aligns them in reverse. Sharpness is evaluated by calculating the sum of gradients for each frame. Finally, all aligned images are weighted and fused to obtain a sharp fused image. The image is then converted to grayscale and divided into blocks. The grayscale values ​​within each block are statistically analyzed, and outliers are removed. The image with the fewest outlier blocks is selected as the analysis image. The contours of consecutive outlier blocks are drawn on the analysis image, and the defect type is determined by geometric feature analysis. Finally, the number and location of defects are recorded and displayed in real time, and a buzzer alarm is triggered when a threshold is reached.

2. The machine vision-based aluminum alloy frame weld defect detection device according to claim 1, characterized in that: The anti-shake analysis steps of the defect analysis module are as follows: M1: Continuously collected data at the same time. Motion alignment of the weld seam images is performed, using the first frame image as a reference, and the calculation of the... Motion transformation matrix of frame image relative to the first frame image And utilize its inverse transformation Perform spatial transformation on each frame of the image to obtain the aligned image. ; M2: Weighted fusion is performed on the aligned frames. First, the gradient sum Sᵢ of each frame is calculated as a sharpness evaluation index. The larger the gradient sum, the sharper the image. The calculation formula is as follows: , For the image in coordinates Gradient magnitude at; M3: Assign weights based on the sharpness of each frame. The higher the resolution, the greater the weight, and the sum of all weights is 1; All aligned images are weighted and fused according to their assigned weights to obtain a clear, merged image that eliminates jitter and blur. .

3. The machine vision-based aluminum alloy frame weld defect detection device according to claim 2, characterized in that: The steps for determining the analysis image in the defect analysis module are as follows: Q1: Convert the fused image to grayscale and divide it into segments according to a preset size. Given several grayscale blocks of the same size, within each grayscale block... Statistical analysis was performed on the grayscale values ​​of each pixel to calculate the mean. and standard deviation Set the grayscale value fluctuation range to Mark grayscale values ​​outside the range as outliers and record the number of outliers. ; Q2: If If the grayscale block data is found to be abnormal, it will be re-detected. This is a preset proportional coefficient; if Then, after removing outliers, the mean of the remaining gray values ​​is recalculated. The gray value is used as the representative gray value of the gray block; if it is still determined to be abnormal after re-detection, the gray block is marked as an abnormal gray block. After identifying all grayscale blocks, count the number of abnormal grayscale blocks in each frame of the image. Select The image with the smallest value is used as the analysis image.

4. The machine vision-based aluminum alloy frame weld defect detection device according to claim 3, characterized in that: The defect determination steps in the defect analysis module are as follows: R1: Obtain the serial numbers of the abnormal grayscale blocks in the analysis image, draw the corresponding coordinate points in the coordinate system and connect the adjacent points to form a closed contour; record the length of the contour in the horizontal and vertical directions, and take the midpoint of the line connecting the two farthest points of the contour or the intersection of multiple lines as the reference point; draw extension lines with equal angle intervals outward from the reference point, mark the intersection points of the extension lines and the contour, and calculate the distance between the intersection points and the reference point. If the length difference between adjacent extension lines fluctuates within the preset range, the contour is determined to be a regular shape; otherwise, it is determined to be an irregular shape and marked as a slag inclusion defect. R2: Retrieves the grayscale range of various defects and the aspect ratio of cracks from historical data. If the outline meets the circular equation, it is determined to be a porosity defect. If the outline is long and the aspect ratio is within the crack feature range, it is determined to be a crack defect. Otherwise, it is determined to be an incomplete penetration defect. Records the number of various defects and the position of the reference point, displays them on the display screen in real time, and triggers a buzzer alarm when the total number of defects reaches the preset threshold.

5. The machine vision-based aluminum alloy frame weld defect detection device according to claim 1, characterized in that: The fixing assembly includes a storage slot (4) on the outer wall of one end of the detector (1). Three vertical rods (5) are fixedly connected to the storage slot (4) at equal intervals. Two lifting plates (6) are symmetrically sleeved on the three vertical rods (5). A hinge plate (7) is hinged to the outer end face of the lifting plate (6). Two guide rods are symmetrically fixed to the end of the hinge plate (7) away from the lifting plate (6). A horizontally placed L-shaped plate (8) is slidably sleeved on the two guide rods. An installation hole is opened between the two guide rods on the hinge plate (7). A pull-back spring is provided in the installation hole. One end of the pull-back spring is fixedly connected to the outer wall of the L-shaped plate (8). Multiple first springs (13) sleeved on the vertical rods (5) are fixedly connected between the opposite faces of the two lifting plates (6). The middle section of the first spring (13) is fixedly connected to the outer wall of the vertical rod (5).

6. The machine vision-based aluminum alloy frame weld defect detection device according to claim 5, characterized in that: The lifting plate (6) has a U-shaped cross-section when viewed from above. Two sets of spring blocks (12) are symmetrically arranged on the inner walls of both sides of the lifting plate (6). Unlocking slots that connect the spring blocks (12) are symmetrically opened on both sides of the top surface of the lifting plate (6). The unlocking slots are connected to the spring blocks (12). Unlocking sliders (10) are slidably arranged in the unlocking slots. One end of the unlocking sliders (10) is fixedly connected to the blocks of the spring blocks. The outer walls of both sides of the hinge plate (7) have slots (12) that are compatible with the spring blocks (12).

7. The machine vision-based aluminum alloy frame weld defect detection device according to claim 6, characterized in that: The L-shaped plate (8) has a positioning hole in the middle of the outward-facing end. The inner wall of the storage groove (4) is provided with spring locking steel balls (9) at both the upper and lower ends. The spring locking steel balls (9) are adapted to the positioning hole.

8. The machine vision-based aluminum alloy frame weld defect detection device according to claim 7, characterized in that: The outer wall of the detector (1) is fixed with an elastic rope handle (11) on one side, and the outer wall of the other side of the detector (1) is provided with an oblique insertion hole for inserting the detection camera (3).

9. The machine vision-based aluminum alloy frame weld defect detection device according to claim 7, characterized in that: The detector (1) has a data interface at the lower end of the oblique insertion hole, and a charging port at the lower end of the data interface.

Images