Automatic visual detection imaging device and method for foreign matters in cladding tube

Abstract

The invention belongs to the field of visual inspection, and particularly relates to an automatic visual inspection imaging device and method for foreign matters in a cladding tube, and the device comprises a coaxial light source, an electric zoom lens and a camera; the two ends of the cladding tube are respectively provided with a set of coaxial light sources, the electric zoom lens and the camera, and the coaxial light sources installed at the two ends of the cladding tube are light sources with different colors and different wave bands. According to the invention, foreign matters with the area greater than or equal to 1mm < 2 > can be accurately detected on the premise of no contact, secondary pollution or physical damage to a pipeline is avoided, the problems of depth-of-field limitation and insufficient illumination of a traditional optical system in detection of the interior of the cladding tube are solved, no blind area of detection is ensured, and global visual detection of the interior of the cladding tube with the inner diameter of about 8.5 mm and the length of about 4000 mm can be realized.

Description

Technical Field

[0001] This invention belongs to the field of visual inspection, specifically relating to an automatic visual inspection imaging device and method for foreign objects inside a casing tube. Background Technology

[0002] Existing methods for detecting internal defects in pipelines mainly include endoscopic and video-guided inspection, and fixed-focus imaging inspection using monofocal imaging systems. However, these methods exhibit significant limitations in detecting foreign objects inside clad pipes with an inner diameter of approximately 8.5 mm and a length of 4 meters: the insertion of inspection tools into the 4-meter-long pipe severely restricts operational performance and causes wear on the workpiece; the narrow diameter limits the size of the endoscope, directly affecting image resolution and field of view; the limited depth of field of the fixed-focus camera makes it impossible to clearly capture the entire 4-meter-long inner wall of the pipe, with areas outside the depth of field appearing blurred, significantly reducing the contrast and edge sharpness of foreign objects, increasing the risk of missed detection, and the depth of field problem becomes more severe with increasing distance, significantly reducing the reliability of detection in distant areas. This physical limitation makes it difficult for traditional monofocal imaging systems to meet the high-quality inspection requirements of long pipes.

[0003] Furthermore, the imaging inside the tube is irregular; light rays reach the camera at different angles from different parts of the tube's inner wall, resulting in varying distances to the camera. Consequently, the inner wall of the tube experiences varying degrees of compression and stretching under perspective projection. The tube may contain foreign objects of diverse types, shapes, materials, and sizes, making it impossible for a single algorithm to handle such a complex situation. Summary of the Invention

[0004] The purpose of this invention is to provide an automatic visual detection imaging device and method for foreign objects inside a casing tube, which can accurately detect objects with an area ≥1mm² without contact. 2 This technology removes foreign objects, preventing secondary pollution or physical damage to the pipeline. It overcomes the limitations of traditional optical systems in terms of depth of field and insufficient illumination when inspecting the inside of clad pipes, ensuring no blind spots in the inspection. It can achieve full-area visual inspection of the inside of clad pipes with an inner diameter of approximately 8.5 mm and a length of approximately 4000 mm.

[0005] Technical solution to achieve the purpose of this invention:

[0006] An automatic visual detection imaging device for foreign objects inside a clad tube is provided. The device includes a coaxial light source, an electric zoom lens, and a camera. A set of coaxial light sources, electric zoom lenses, and cameras are installed at each end of the clad tube. The coaxial light sources installed at both ends of the clad tube are light sources of different colors and different wavelengths.

[0007] Furthermore, the coaxial light source is provided with a front-to-back adjustment mechanism and a rear-to-up adjustment mechanism, which are adjusted to align the coaxial light source with the center of the casing tube.

[0008] Furthermore, the focusing position of the motorized zoom lens is adjusted sequentially from the near end to the far end, and multiple shots are taken to cover the entire casing tube. The transition between adjacent focusing positions ensures overlapping depth of field.

[0009] Furthermore, the motorized zoom lens is equipped with a camera lens angle adjustment mechanism for adjusting the angle of the motorized zoom lens.

[0010] Furthermore, the camera simultaneously captures images at both ends of the casing tube, and the captured images are transmitted to the industrial control computer processing unit. The industrial control computer processing unit uses a dual-mode foreign object intelligent detection algorithm to process the images and determines whether there are foreign objects based on the image processing results. If a foreign object is detected in any image, the tube body is determined to be unqualified.

[0011] Furthermore, a dual-mode foreign object intelligent detection algorithm is used for image processing, and the specific method for determining foreign objects is as follows:

[0012] For the near-port area of ​​the cladding tube, traditional visual algorithms are used to quickly identify obvious foreign objects. Binarization is used to extract the foreign object area and extract features such as the target area and grayscale mean. The number of black pixels in the connected component is calculated. If the difference between the area and the number of pixels exceeds the set threshold, it is determined to be a foreign object.

[0013] For the far-port region of the clad tube, gamma correction enhancement is used, and after initial screening using traditional algorithms, detection is performed. Deep learning is used for confidence filtering, and the confidence scores are fused and output to give the judgment result.

[0014] Furthermore, the camera is equipped with a camera up-down adjustment mechanism and a camera left-right adjustment mechanism for adjusting the position of the camera.

[0015] Furthermore, the device also includes: a support and a baffle. The casing tube is laid on the support, and a coaxial baffle is installed on the support. The coaxial baffle intercepts the casing tube, keeping it stationary and ensuring that the center of the casing tube is close to the central coaxial line, ensuring that the entire area inside the tube can be covered by the light source.

[0016] Furthermore, the device also includes a cantilever box, which is installed on the side of the cladding tube inspection area to observe the inspection status of the cladding tube inspection area.

[0017] An automatic visual detection and imaging method for foreign objects inside a clad tube, the method comprising:

[0018] S1: The casing tube is driven downward along the production line and is blocked by a baffle at the detection position. The sensor is installed on the bracket and is aligned with the position of the bracket baffle. When the sensor detects that the casing tube has arrived at the detection mechanism, it sends a signal to the industrial control computer.

[0019] S2: The two motorized zoom lenses are controlled by an industrial control computer to zoom sequentially according to the preset focal length. The camera takes pictures through the motorized zoom lenses and coaxial light source, acquires multiple sets of images, and transmits the images to the industrial control computer processing unit.

[0020] S3: The industrial control computer processing unit uses a dual-mode foreign object intelligent detection algorithm for image processing;

[0021] S4: An alarm is triggered if any foreign object is detected in any image. If no foreign object is detected, the baffle descends and the casing tube continues to flow down, completing the detection of foreign objects inside the casing tube.

[0022] The beneficial technical effects of this invention are as follows:

[0023] 1. This invention provides complete coverage without blind spots: segmented focusing and simultaneous shooting at both ends, with adjacent focusing positions advancing and the depth of field areas physically overlapping, effectively avoiding imaging blind spots caused by "depth of field breaks" in pipe walls or foreign object areas.

[0024] 2. Optimization of image clarity in this invention: By dynamically controlling the focus of the motorized lens, each image segment is captured near its optimal imaging plane, thereby improving the overall signal-to-noise ratio and edge clarity of the image.

[0025] 3. This invention is flexible and adaptable to targets of different depths: Foreign objects of different sizes and depths do not require lens replacement or manual position adjustment. Full coverage of the target area can be achieved simply by adjusting the motorized lens stepping parameters through program settings, thereby improving the system's automation capabilities and adaptability.

[0026] 4. This invention is applied to the detection of foreign objects inside metal cladding tubes in the nuclear industry, preventing process problems and product quality risks that may be caused by foreign objects; at the same time, it meets the efficiency requirements of the production line, and can complete a comprehensive non-contact inspection of a tube within 20 seconds, meeting product quality control requirements. Attached Figure Description

[0027] Figure 1 A schematic diagram of an automatic visual detection imaging device for foreign objects inside a clad tube provided by the present invention; Figure 2 A schematic diagram of the hardware structure for one-sided detection imaging in an automatic visual detection imaging device for foreign objects inside a casing tube provided by the present invention. Figure 3 A schematic diagram of the hardware structure of the detection and imaging device for the automatic visual detection and imaging of foreign objects inside a casing tube provided by the present invention (cantilever box side). Figure 4 Schematic diagram of the foreign object detection principle of this invention; Figure 5 1mm in a specific embodiment of the present invention 2 Imaging image inside the foreign body tube.

[0028] In the diagram: 1-Shell tube; 2-Bracket; 3-Baffle; 4-Coaxial light source; 5-Camera cover; 6-Cantilever box; 7-Light source front-to-back adjustment mechanism; 8-Light source up-to-down adjustment mechanism; 9-Electric zoom lens; 10-Camera; 11-Camera up-to-down adjustment mechanism; 12-Camera left-to-right adjustment mechanism; 13-Camera lens angle adjustment mechanism. Detailed Implementation

[0032] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0033] like Figure 1-3 As shown, the present invention provides an automatic visual detection imaging device for foreign objects inside a casing tube, comprising: a bracket 2, a baffle 3, a coaxial light source 4, a camera cover 5, a cantilever box 6, a front and rear adjustment mechanism for the light source 7, a vertical adjustment mechanism for the light source 8, an electric zoom lens 9, a camera 10, a vertical adjustment mechanism for the camera 11, a horizontal adjustment mechanism for the camera 12, and a lens angle adjustment mechanism for the camera 13.

[0034] The following components are included: a coaxial light source 4, a light source front-to-back adjustment mechanism 7, a light source up-to-down adjustment mechanism 8, an electric zoom lens 9, a camera 10, a camera up-to-down adjustment mechanism 11, a camera left-to-right adjustment mechanism 12, and a camera lens angle adjustment mechanism 13. Each set is located at one end of the bracket 2. Except for the coaxial light source 4, which uses different colors and wavelengths, all others are of the same model. A cantilever box 6 is installed on the side of the cladding tube inspection area. The height and rotation angle of the cantilever box 6 can be flexibly adjusted according to worker operation. This facilitates worker operation and allows observation of the inspection status of the cladding tube inspection area.

[0035] like Figure 4 As shown, the present invention provides an automatic visual detection imaging device for foreign objects inside a casing tube. It employs two-end imaging, with a coaxial light source 4, an electric zoom lens 9, and a camera 10 installed at each end of the casing tube 1. The coaxial light source 4 is installed at both ends of the casing tube 1, and each coaxial light source 4 has a front-to-back adjustment mechanism 7 and a vertical adjustment mechanism 8. Adjusting these mechanisms aligns the coaxial light source 4 with the center of the casing tube. Simultaneously, the casing tube 1 is laid on a support 2, and a coaxial baffle 3 is installed on the support 2. The coaxial baffle 3 intercepts the casing tube, keeping it stationary and ensuring that the center of the casing tube is close to the central coaxial line, thus ensuring that the entire interior of the tube is covered by the light source.

[0036] The coaxial light source 4 is located at both ends of the casing tube and uses different colored light sources to make the foreign object produce obvious color difference shadows under different colored light, thereby enhancing the contrast between the foreign object and the background and avoiding the shadow cancellation problem caused by the same or similar colored light sources.

[0037] The motorized zoom lenses 9 are mounted at both ends of the casing tube 1 to implement a dynamic focusing strategy, solving the depth-of-field limitation problem of traditional fixed-focus systems in long-tube detection and achieving precise electronic control of the focus position. Through lens control software, the focus position is controlled by adjusting the Focus parameter, and the aperture size is controlled by adjusting the Iris parameter. The focus position is sequentially adjusted from the near end (tube opening) to the far end (tube center) after multiple shots, covering the entire casing tube. The motorized zoom lenses 9 are equipped with a camera lens angle adjustment mechanism 13 for adjusting the angle of the motorized zoom lenses 9.

[0038] The transition between adjacent focus positions is calculated and tested to ensure overlapping depth of field, avoid blind spots, and maximize the image sharpness of each area.

[0039] Depth of field is defined as the range of distances in front of or behind an object that can be captured in a clear image at the front of a camera lens or other imager.

[0040] Depth of field calculation formula:

[0041] Foreground depth: ΔL1=(FδL) 2 ) / (f 2 +FδL)

[0042] Depth of field: ΔL2=(FδL) 2 ) / (f 2 -FδL)

[0043] Depth of field: ΔL = ΔL² + ΔL² = (2f 2 FδL 2 ) / (f 4 -F 2 δ 2 L 2 )

[0044] In the formula,

[0045] δ - Diameter of the allowable circle of confusion;

[0046] f - lens focal length;

[0047] F-Aperture value for the lens;

[0048] L - Focusing distance;

[0049] ΔL1 - Deep foreground;

[0050] ΔL2 - Depth of field;

[0051] ΔL - Depth of field.

[0052] With f = 50, δ = 0.0265, and F = 8.7, the overlap between adjacent depth ranges is approximately 10% of their respective depths. The calculation results for adjacent depths are shown in Table 1.

[0053] Table 1 Calculation results of adjacent depth of field

[0054]

[0055]

[0056] Cameras 10 are installed at both ends of the cladding tube 1, and simultaneously capture images at both ends. The captured images are transmitted to the industrial control computer processing unit, which uses a dual-mode foreign object intelligent detection algorithm to process the images. Based on the image processing results, it determines whether a foreign object exists. If a foreign object is detected in either image, the tube is deemed unqualified. Cameras 10 are equipped with a camera up-down adjustment mechanism 11 and a camera left-right adjustment mechanism 12 for adjusting the position of the camera 10.

[0057] The motorized zoom lens 9 and the camera 10 are covered by a camera cover 5 to protect the motorized zoom lens 9 and the camera 10.

[0058] Image processing and foreign object detection specifically include the following steps:

[0059] Image processing determination includes the following steps

[0060] Step 1: Obtain the grayscale image I(x,y).

[0061] Step 2, Preprocessing: Noise Reduction

[0062] I g (x,y)=(I*G σ (x,y)

[0063] Among them, G σ Let σ be the standard deviation of the Gaussian kernel. σ = 1 to 2.

[0064] Step 3: Calculate the threshold T: global threshold (fixed or Otsu).

[0065] Choose a fixed threshold T and binarize:

[0066]

[0067] Step 4: Binarization: Obtain the binary image B(x,y).

[0068] Step 5, Morphological processing: noise reduction, hole filling, and connection.

[0069] Dilation:

[0070]

[0071] Erosion:

[0072]

[0073] Opening operation: (Remove minor bright noise)

[0074] Closing operation: (Fill in the small hole)

[0075] Step 6, Connectivity Analysis: Extract candidate regions and calculate their shape / grayscale features for evaluation.

[0076] By labeling the connected components of B(x, y), we obtain each connected region R. i Calculate commonly used features:

[0077] Area (number of pixels) A i =∣R i |

[0078] Minimum bounding rectangle length and width w i ,h i →Aspect ratio:

[0079]

[0080] Grayscale difference (relative to background / neighborhood):

[0081] Δμ i =μ in,i -μ bg,i

[0082] Where μ in,i μ represents the average gray level within the region. bg,i This represents the average gray level of the outer ring zone (local background) of the region.

[0083] Step 7: Determine whether it is a foreign object based on rules (area, shape, grayscale difference).

[0084] Area threshold (noise removal):

[0085] A i ≥A min

[0086] A min =20 pixels converted to pixels based on physical area.

[0087] Grayscale contrast threshold (ensuring non-weak contrast):

[0088] |Δμ i |≥Δμ min

[0089] Δμ min =10 (grayscale difference).

[0090] Final decision (Boolean logic):

[0091] is_foreign(R i )=(A i ≥A min )(|△μ i ≥Δμ min )

[0092] If any R i If the condition is met, the image is determined to contain a foreign object.

[0093] Output: Binary result + foreign object box

[0094] 1mm 2 Imaging inside the foreign body tube, such as Figure 5 As shown.

[0095] A dual-mode intelligent foreign object detection algorithm is adopted. Based on the imaging conditions of foreign objects at the near and far ports, and considering the differences in imaging characteristics of different regions inside the tube, different algorithm models are used to adapt to the differences in brightness distribution, distortion degree, resolution and target shape of imaging in different regions, so as to achieve high-precision foreign object detection throughout the process.

[0096] Different image processing modes are applied to the pipe opening and both ends within the pipe, as shown in Table 2:

[0097] For the front section (mouth) of the casing tube, traditional visual algorithms are used to quickly identify obvious foreign objects. Binarization is used to extract the foreign object region and extract features such as the target area and grayscale mean. The number of black pixels in the connected component is calculated. If the difference between multiple parameters (area, number of pixels) exceeds the set threshold, it is determined to be a foreign object.

[0098] For the distant section (inside the tube) of the clad tube, where the light is weaker, the contrast is lower, and the noise is greater, gamma correction enhancement is used. After initial screening using traditional algorithms, detection is performed. Deep learning (YOLOv8 algorithm) is used for confidence filtering, and the confidence scores are fused and output to give the judgment result.

[0099] Table 2 Imaging characteristics and image processing methods for different regions within the tube.

[0100]

[0101] This invention provides an automatic visual detection and imaging method for foreign objects inside a clad tube, specifically including the following steps:

[0102] S1: The casing tube 1 is driven downward along the production line and is blocked by the baffle 3 at the detection position. The sensor is installed on the bracket 2 and is in the same position as the bracket baffle 3. When the sensor senses that the casing tube has arrived at the detection mechanism, it sends a signal to the industrial control computer.

[0103] S2: The motorized zoom lenses 9 at both ends are controlled by the industrial control computer to zoom sequentially according to the preset focal length (including focus position and aperture size). The camera 10 takes pictures through the motorized zoom lenses 9 and the coaxial light source 4 to acquire multiple sets of images and transmit the images to the industrial control computer processing unit.

[0104] S3: The industrial control computer processing unit uses a dual-mode foreign object intelligent detection algorithm for image processing.

[0105] S4: An alarm is triggered if any foreign object is detected in any image. If no foreign object is detected, the baffle 3 descends and the casing tube 1 continues to flow down, completing the detection of foreign objects inside the casing tube.

[0106] The present invention has been described in detail above with reference to the accompanying drawings and embodiments. However, the present invention is not limited to the above embodiments, and various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention. All contents not described in detail in the present invention can be derived from existing technologies.

Claims

1. An automatic visual detection and imaging device for foreign objects inside a clad tube, characterized in that, The device includes: a coaxial light source (4), an electric zoom lens (9), and a camera (10); a set of coaxial light source (4), electric zoom lens (9), and camera (10) are installed at each end of the casing tube (1), and the coaxial light source (4) installed at both ends of the casing tube (1) adopts light sources of different colors and different wavelengths.

2. The automatic visual detection imaging device for foreign objects inside a clad tube according to claim 1, characterized in that, The coaxial light source (4) is provided with a front-to-back adjustment mechanism (7) and a rear-to-up adjustment mechanism (8). Adjusting the front-to-back adjustment mechanism (7) and the rear-to-up adjustment mechanism (8) will align the coaxial light source (4) with the center of the casing tube.

3. The automatic visual detection imaging device for foreign objects inside a clad tube according to claim 1, characterized in that, The focus position of the electric zoom lens (9) is adjusted sequentially from the near end to the far end, and multiple shots are taken to cover the entire casing tube. The transition between adjacent focus positions ensures overlapping depth of field.

4. The automatic visual detection and imaging device for foreign objects inside a clad tube according to claim 1, characterized in that, The motorized zoom lens (9) is equipped with a camera lens angle adjustment mechanism (13) for adjusting the angle of the motorized zoom lens (9).

5. The automatic visual detection imaging device for foreign objects inside a clad tube according to claim 1, characterized in that, The camera (10) takes pictures simultaneously at both ends of the casing tube. The pictures are transmitted to the industrial control computer processing unit. The industrial control computer processing unit uses a dual-mode foreign object intelligent detection algorithm to process the images. Based on the image processing results, it determines whether there are foreign objects. If a foreign object is detected in any picture, the tube body is determined to be unqualified.

6. The automatic visual detection imaging device for foreign objects inside a clad tube according to claim 5, characterized in that, A dual-mode foreign object intelligent detection algorithm is used for image processing. The specific method for identifying foreign objects is as follows: For the near-port area of ​​the cladding tube, traditional visual algorithms are used to quickly identify obvious foreign objects. Binarization is used to extract the foreign object area and extract features such as the target area and grayscale mean. The number of black pixels in the connected component is calculated. If the difference between the area and the number of pixels exceeds the set threshold, it is determined to be a foreign object. For the far-port region of the clad tube, gamma correction enhancement is used, and after initial screening using traditional algorithms, detection is performed. Deep learning is used for confidence filtering, and the confidence scores are fused and output to give the judgment result.

7. The automatic visual detection imaging device for foreign objects inside a clad tube according to claim 1, characterized in that, The camera (10) is equipped with a camera up-down adjustment mechanism (11) and a camera left-right adjustment mechanism (12) for adjusting the position of the camera (10).

8. The automatic visual detection imaging device for foreign objects inside a clad tube according to claim 1, characterized in that, The device also includes: a bracket (2) and a baffle (3). The casing tube (1) is laid on the bracket (2). The coaxial baffle (3) is installed on the bracket (2). The coaxial baffle (3) intercepts the casing tube, so that the casing tube (1) remains stationary, ensuring that the center of the casing tube is close to the central coaxial line, and ensuring that the entire area inside the tube can be covered by the light source.

9. The automatic visual detection imaging device for foreign objects inside a clad tube according to claim 1, characterized in that, The device also includes a cantilever box (6), which is installed on the side of the cladding tube inspection area to observe the inspection status of the cladding tube inspection area.

10. An automatic visual detection and imaging method for foreign objects inside a clad tube, comprising the automatic visual detection and imaging device for foreign objects inside a clad tube as described in any one of claims 1-9, characterized in that, The method includes: S1: The casing tube (1) is driven downward along the production line and is blocked by the baffle (3) at the detection position. The sensor is installed on the bracket (2) and is in the same position as the bracket baffle (3). When the casing tube reaches the detection mechanism, it sends a signal to the industrial control computer. S2: The two motorized zoom lenses (9) are controlled by the industrial control computer to zoom sequentially according to the preset focal length. The camera (10) takes pictures through the motorized zoom lens (9) and the coaxial light source (4) to acquire multiple sets of images and transmit the images to the industrial control computer processing unit. S3: The industrial control computer processing unit uses a dual-mode foreign object intelligent detection algorithm for image processing; S4: If any foreign object is detected in any image, an alarm is triggered. If no foreign object is detected, the baffle (3) descends and the casing tube (1) continues to flow down, completing the detection of foreign objects inside the casing tube.

Images