A cotton impurity detection system and method based on multi-feature fusion

Abstract

The application relates to the field of agricultural visual detection technology and discloses a cotton impurity detection system and method based on multi-feature fusion. The system comprises a shell, a multi-modal image acquisition module, a multi-modal information processing module and a cotton conveying module. The multi-modal image acquisition module comprises a plurality of RGB cameras, a polarization camera, a plurality of LED light sources and a UV violet light lamp which are fixed to a camera support. The multi-modal information processing module comprises a lower computer and an industrial computer, and the industrial computer is connected with the lower computer, the plurality of RGB cameras and the polarization camera. Cotton passes through the image acquisition module through a conveying pipeline, and information data is transmitted into the industrial computer through polarization image and RGB image acquisition. Through algorithm processing of the multi-modal information of the cotton, a detection result can be quickly obtained. The application can realize rapid detection, positioning and automatic removal of impurities contained in the cotton, and greatly improves the impurity removal efficiency and precision.

Description

Technical Field

[0001] This invention relates to the field of agricultural visual inspection technology, and in particular to a cotton impurity detection system and method based on multi-feature fusion. Background Technology

[0002] In the planting and harvesting of cotton, plastic film and drip irrigation tape are widely used. However, even with mechanical harvesting, these materials are easily mixed into the cotton, becoming impurities. These impurities not only affect cotton quality but can also damage textile machinery. The ability to effectively identify and sort out impurities in cotton has indirectly promoted the development of the cotton and textile industries.

[0003] Traditional cotton impurity sorting methods primarily rely on manual vision, which suffers from low detection efficiency, poor sorting results, slow speed, high subjectivity, human fatigue, and the tendency to miss small impurities mixed in with the cotton. In recent years, with the development of technologies such as computer vision and deep learning, automated impurity sorting of machine-harvested cotton using image processing and deep learning algorithms has become a research hotspot. However, traditional visual recognition methods have limitations in identifying impurities such as plastic film and polypropylene fibers, especially plastic film covered in soil and polypropylene fibers that are very similar in color to cotton. Most detection schemes cannot accurately detect plastic film and polypropylene fibers. Therefore, there is an urgent need for a multimodal information fusion technology that can effectively improve the accuracy and reliability of cotton impurity detection. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to address the shortcomings of existing technologies by providing a cotton impurity detection system and method based on multi-feature fusion. By combining information from RGB images and polarization images and fusing multi-modal data, the accuracy and reliability of cotton impurity detection are improved.

[0005] Technical Solution: The technical solution adopted by this invention to solve the problem is as follows: a cotton impurity detection system based on multi-feature fusion, comprising a housing, a multimodal image acquisition module, a multimodal information processing module, and a cotton conveying module; the multimodal image acquisition module includes several RGB cameras, polarization cameras, LED light sources, and UV lamps symmetrically distributed on both sides of the cotton conveying pipe; the LED light sources and UV lamps are alternately installed with the cotton conveying pipe as the center; the multimodal information processing module includes an industrial control computer, a lower-level computer, and a PLC controller, the industrial control computer being connected to the RGB cameras, polarization cameras, and the lower-level computer; the cotton conveying module includes a fan and an air duct for controlling the cotton conveying speed.

[0006] Furthermore, the multimodal image acquisition module includes three RGB cameras and three polarization cameras on both the front and back sides. The RGB cameras and polarization cameras are respectively fixed to the camera bracket, and the bracket and the housing are connected by soft contact to eliminate the influence of mechanical vibration.

[0007] Furthermore, the multimodal information acquisition module alternately installs several LED light sources and UV lamps on both sides of the air duct. With the cotton conveying air duct as the center, several LED light sources and UV lamps are installed on both sides of the air duct to ensure that the part scanned by the camera is illuminated by light source in 360°.

[0008] Furthermore, the specific installation positions of the LED light source and the UV lamp are as follows: the first group of light sources is installed on the upper right and upper left sides of the RGB camera line scan position, at a 20° angle to the vertical direction of the pipe; the second group of light sources is installed on the lower right and lower left sides of the RGB camera line scan position, at a 160° angle to the vertical direction of the pipe; the third group of light sources is installed on the upper left and upper right sides of the polarization camera line scan position, at a 20° angle to the vertical direction of the pipe; and the fourth group of light sources is installed on the lower left and lower right sides of the polarization camera line scan position, at a 160° angle to the vertical direction of the pipe.

[0009] This invention also provides a cotton impurity detection method based on multi-feature fusion, comprising the following steps:

[0010] S1: Synchronously trigger the RGB camera and polarization camera to acquire images. The RGB camera outputs a 2048×512 RGB image, and the polarization camera acquires polarization intensity maps at 0°, 45°, 90°, and 135°. The polarization degree map (DoP) and polarization angle map (AoP) are then calculated and generated by the lower-level computer.

[0011] S2: Convert the RGB image to the Lab color space, and identify colored impurities and linear impurities similar to the color of cotton based on L, a, and b channel threshold segmentation and template matching;

[0012] S3: Converts RGB images to HSV color space and detects high-saturation areas based on saturation to identify vibrant color impurities;

[0013] S4: Align the RGB image with the polarization feature map and fuse them into a five-channel image, then input the deep learning network for detection;

[0014] S5: Merge the detection results of S2, S3, and S4, generate an array, and send it to the lower-level machine to control the start of the spray valve used to remove impurities.

[0015] Furthermore, in step S1, "the RGB camera acquires a 2048*512 RGB image, and the polarization camera acquires light intensity maps at polarization angles of 0°, 45°, 90°, and 135°, and processes them through a lower-level machine according to formulas to obtain a degree of polarization (DoP) map and an angle of polarization (AoP) map. Then, the RGB image and the polarization image are input into the image processing module." The specific implementation method is as follows:

[0016] S11: Image acquisition adopts a multi-camera configuration, including an RGB camera and a polarization camera. The core control board sends a trigger signal to the acquisition card to enable the RGB camera and the polarization camera to be triggered synchronously and acquire images respectively. The RGB image is directly sent to the image processing module. The polarization image is processed by the lower-level machine to generate a polarization degree map (DoP) and a polarization angle map (AoP) before being sent to the image processing module.

[0017] Furthermore, the specific implementation method of step S2, "converting the RGB image to a Lab color space image and performing thresholding segmentation and template matching on the image using the L, a, and b values ​​of the three channels," is as follows:

[0018] S21: Convert the RGB color space to the Lab color space and further extract the features of the L, a, and b channels.

[0019] S22: Duplicate the Lab image twice. On one hand, perform threshold segmentation to identify areas with a large difference from the cotton threshold as impurities. On the other hand, perform threshold segmentation to identify suspicious areas with a small difference from the cotton threshold, and then perform multi-template matching to mainly detect linear polypropylene filaments, fine threads, oily cotton, etc.

[0020] S23: The detection results are smoothed and noise is removed through morphological operations (corrosion followed by expansion) to improve the accuracy of impurity region identification.

[0021] Furthermore, in step S3, the RGB image is converted into an HSV channel, and the values ​​of the S channel are detected, with highly saturated portions considered as impurities. The specific operation steps are as follows:

[0022] S31: Convert the RGB color space to the HSV color space, segment the values ​​of the saturation (S) channel, and identify areas with a saturation greater than 160 as highly saturated vivid color impurities.

[0023] As a preferred embodiment, step S4, "fusing the RGB image and the polarization image to generate a five-channel image, which is then input into the deep learning network for detection," is implemented as follows:

[0024] S41: Through calibration and registration, ensure that the RGB image and polarization image are both 2048*512 in size and are fully aligned at the pixel level, which facilitates multimodal feature fusion.

[0025] S42: Combine the aligned RGB image and polarization feature map into a unified format to form a 5-channel multimodal image.

[0026] S43: Detect the fused five-channel image using a deep learning model.

[0027] As a preferred embodiment, step S5 involves "merging the detection results of S2, S3, S4, and S5, generating an array, and sending it to the lower-level machine to control the start of the spray valve." The specific implementation method is as follows:

[0028] S51: The detection results of steps S2, S3, and S4 are converted into a binary image. White pixels are impurities, and black pixels are non-impurities. The results of S2, S3, and S4 are ORed on a pixel-by-pixel basis. The maximum connected component is detected according to the region corresponding to the spray valve. The part of the detected impurity region that is greater than a certain threshold is identified as the final result, which effectively reduces some misidentified scattered pixels. Impurity regions are 1, and non-impurity regions are 0. The three results are merged and output into a 20*512 array and sent to the lower computer to control the spray valve to start spraying out impurities.

[0029] Beneficial effects: (1) The high-precision multimodal information fusion adopted in this invention: RGB camera and polarization camera are deployed on both sides of the cotton conveying pipeline at the same time. Through the deployment of camera and light source positions, multi-angle and multimodal images of cotton and its impurities are obtained; providing a high-precision data source for subsequent detection algorithms, ensuring the accuracy and speed of impurity detection;

[0030] (2) This invention achieves one-stop detection of common impurities in machine-harvested cotton, such as drip irrigation tape, mulch film, packaging film, and polypropylene fibers, by fusing RGB and polarized images. Using a YOLOv8 deep learning model, it performs end-to-end feature learning on the multimodal fused images, automatically extracting complex image features and detecting and classifying various impurity types. Different air volume and air time are selected based on different impurity types to adapt to complex real-world environments, improving the robustness and accuracy of impurity classification. This avoids the problems of low detection accuracy, slow processing speed, and high false recognition rate in traditional methods, greatly improving efficiency and cotton purity in practical production applications.

[0031] (3) This invention uses communication between the host computer and the slave computer to quickly merge the detection results of RGB images and polarization images, ensuring that impurities are accurately separated within one image acquisition cycle from camera image acquisition to spray valve start control, which greatly improves processing efficiency and real-time performance in production applications.

[0032] (4) This invention can handle a variety of impurities in machine-harvested cotton, such as drip irrigation tape, packaging film, polypropylene filaments, and fine thread impurities, and has strong adaptability and practical application value. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0034] Figure 2 This is a schematic diagram of the internal structure of the present invention.

[0035] Figure 3 This is a flowchart of the impurity detection algorithm of the present invention. Detailed Implementation

[0036] like Figure 1 , Figure 2As shown, a cotton impurity detection system based on multi-feature fusion includes a housing 1, an image acquisition module, a multimodal information processing module, a cotton conveying module, and an impurity conveying module. The cotton conveying pipe 2 has a first LED light source 20, a second LED light source 22, a first UV lamp 21, a third LED light source 23, a second UV lamp 23, a fourth LED light source 25, a fifth LED light source 26, a third UV lamp 27, a sixth LED light source 28, a seventh LED light source 29, a fourth UV lamp 30, an eighth LED light source 31, a ninth LED light source 32, a fifth UV lamp 33, a tenth LED light source 34, an eleventh LED light source 35, a sixth UV lamp 36, a twelfth LED light source 37, a thirteenth LED light source 38, a seventh UV lamp 39, a fourteenth LED light source 40, a fifteenth LED light source 41, an eighth UV lamp 42, and a sixteenth LED light source 41 installed on both sides. ED light source 43; the image acquisition module is located on both sides of the cotton conveying pipe 2, with two rectangular openings RGB camera line scan ports 3 reserved on each side of the pipe for the line scan camera to capture images; the image acquisition module includes: RGB camera 6 fixed to the first camera bracket, RGB camera 7 fixed to the second camera bracket, RGB camera 8 fixed to the third camera bracket, RGB camera 9 fixed to the fourth camera bracket, RGB camera 10 fixed to the fifth camera bracket, RGB camera 11 fixed to the sixth camera bracket, polarization camera 13 fixed to the seventh camera bracket, polarization camera 14 fixed to the eighth camera bracket, polarization camera 15 fixed to the ninth camera bracket, polarization camera 16 fixed to the tenth camera bracket, polarization camera 17 fixed to the eleventh camera bracket, and polarization camera 18 fixed to the twelfth camera bracket; the cotton conveying module includes cotton conveying pipe 2 to be inspected, impurity conveying pipe 12, and finished cotton conveying pipe 19. The multimodal information processing module includes an industrial control computer 47, which is connected to RGB cameras 8 and 9, a lower-level computer 1, and a PLC. Lower-level computer 1 is connected to industrial control computer 47, valve plate 1, polarization camera 15, and polarization camera 18. The multimodal information processing module also includes an industrial control computer 48, which is connected to RGB cameras 7 and 10, a lower-level computer 2, and a PLC. Lower-level computer 2 is connected to industrial control computer 48, valve plate 2, polarization camera 14, and polarization camera 17.

[0037] The multimodal information processing module industrial control computer 49 is connected to RGB camera 6, RGB camera 11, slave computer 3, and PLC. The slave computer 3 is connected to industrial control computer 49, valve plate 3, polarization camera 13, and polarization camera 16. The cameras connected to each industrial control computer and slave computer are symmetrical cameras on both sides of the pipeline.

[0038] The first LED light source 20, the second LED light source 22, and the first UV light 21 are installed on the upper right side of the position scanned by the RGB camera, at a 20-degree angle to the vertical direction of the pipe. The third LED light source 23, the second UV light 25, and the fourth LED light source 25 are installed on the upper left side of the position scanned by the RGB camera, at a 20-degree angle to the vertical direction of the pipe. The seventh LED light source 29, the fourth UV light 30, and the eighth LED light source 31 are installed on the lower left side of the position scanned by the RGB camera, at a 160-degree angle to the vertical direction of the pipe. The fifth LED light source 26, the third UV light 27, and the sixth LED light source 28 are installed on the lower right side of the position scanned by the RGB camera, at a 160-degree angle to the vertical direction of the pipe.

[0039] The ninth LED light source 32, the fifth UV light source 33, and the tenth LED light source 34 are installed on the upper left side of the position scanned by the polarization camera, at a 20-degree angle to the vertical direction of the pipe. The eleventh LED light source 35, the sixth UV light source 36, and the twelfth LED light source 37 are installed on the upper right side of the position scanned by the polarization camera, at a 20-degree angle to the vertical direction of the pipe. The fifteenth LED light source 41, the eighth UV light source 42, and the sixteenth LED light source 43 are installed on the lower left side of the position scanned by the polarization camera, at a 160-degree angle to the vertical direction of the pipe. The thirteenth LED light source 38, the seventh UV light source 39, and the fourteenth LED light source 40 are installed on the lower right side of the position scanned by the polarization camera, at a 160-degree angle to the vertical direction of the pipe. The first, second, third, fourth, fifth, and sixth camera supports are located on the same horizontal line and at the center of the first to eighth light sources within the RGB camera's scanned position. The seventh, eighth, ninth, tenth, eleventh, and twelfth camera supports are located on the same horizontal line and at the center of the first to eighth light sources, where the polarization camera scans the image. To ensure that the images captured by the cameras are not affected by external light, the entire image acquisition module is sealed with an opaque aluminum alloy cover.

[0040] The first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth, eleventh, and twelfth camera brackets located on both sides of the cotton conveying pipe are in soft contact with housing 1 to avoid the impact of machine vibration. The covers on both sides of housing 1 can be opened freely for easy adjustment of the cameras, light sources, etc. The openings reserved on both sides of the cotton conveying pipe for adjusting and measuring wind speed can be opened and closed freely to avoid affecting the wind speed inside the pipe during production.

[0041] All cameras acquire images via fixed-frequency trigger signals from lower-level machines. After cotton passes through the camera's scanning position via the conveyor pipe, the image data is sent to the information processing module for processing. RGB images are sent to corresponding industrial control computers 1, 2, and 3 for processing, while polarization images are sent to corresponding lower-level machines 1, 2, and 3 for processing. Finally, the processing results are merged and sent to the valve plate to control the spray valve to eject the identified impurities. The Lab and HSV images converted in the algorithm are processed primarily by the CPU with GPU assistance. The deep learning detection algorithm is mainly accelerated by the GPU to improve processing efficiency, ensuring that impurities are accurately ejected from the image acquisition position to the spray valve opening.

[0042] Since the image acquisition module uses a Matrox image acquisition card, using Matrox's image processing functions not only eliminates the need for image format conversion but also improves processing efficiency and shortens processing time. Matrox image processing functions are used to convert RGB images into Lab and HSV images, and threshold segmentation is performed to identify colored impurities. Simultaneously, threshold segmentation of the Lab image identifies parts that may not belong to cotton. Then, Matrox's shape template matching function is used for template matching to identify impurities such as polypropylene fibers and packaging tape that are similar in color to cotton. While processing the Lab and HSV images, the RGB image is fed into the deep learning detection module, which primarily identifies soil-bearing plastic film that the polarization camera failed to detect. The images acquired by both the RGB and polarization cameras have a resolution of 2048*512.

[0043] like Figure 3 The diagram shows an algorithm flowchart for a cotton impurity sorting method based on multi-feature fusion. First, a camera acquisition program is written to enable the simultaneous rapid acquisition of multi-feature images by both an RGB camera and a polarization camera.

[0044] Using Matrox's image processing functions, RGB images are converted to Lab images for thresholding and template matching. Meanwhile, MIL-type images acquired by the acquisition card are converted to Mat-type images for processing by the deep learning algorithm module.

[0045] The processing of Lab images (thresholding segmentation and template matching) is as follows:

[0046] Threshold segmentation: The threshold range for cotton is white, dark yellow, and the brown area of ​​some cotton leaves and husks. Therefore, other colored items such as drip irrigation tape, colored packaging film, and bright colors like red, blue, and green are considered impurities. Threshold segmentation identifies pixels that do not fall within the L, a, and b threshold ranges of cotton as impurities. This system uses six colors: red, blue, green, purple, orange, and black. Each color has a given range for L, a, and b. For example, the approximate range for red (Lab) is: L: 20-80, a: 40-100, b: 0-60. Pixels that simultaneously meet these conditions are identified as impurities. Furthermore, fine-tuning is performed on images from each camera to ensure recognition accuracy.

[0047] Template matching: For impurities such as polypropylene fibers and hemp ropes that are similar in color to cotton and are mixed in during mechanized cotton harvesting, the areas that are similar in color to cotton are first identified by threshold segmentation, and then template matching is performed using pre-given linear and clump templates to identify linear impurities and clumps of oily cotton.

[0048] Deep learning model detection:

[0049] First, the RGB and polarization images are fused. The specific fusion method is as follows:

[0050] Calculate the polarization characteristics from the acquired light intensity map:

[0051] Degree of polarization (DoP): Used to represent the intensity distribution of polarized light, the formula is:

[0052]

[0053] Polarization angle (AoP): Used to represent the direction of polarized light, the formula is:

[0054]

[0055] The lower-level machine obtains two single-channel images, namely the 4th and 5th channels, by using the light intensity maps with polarization angles of 0°, 45°, 90°, and 135° acquired through the above formula.

[0056] The aligned RGB images (channels 1–3) and polarization feature maps (polarization degree DoP and polarization angle AoP, channels 4 and 5) are stacked along the channel dimension. The fused image format is [H,W,5], where H is the height and W is the width. A storage method suitable for deep learning frameworks is chosen, represented as [B,C,H,W], where B is the batch size and C=5 is the number of channels. To improve the stability of model training, the data for all channels are normalized to a range of [0,1] to reduce the impact of numerical differences on training and optimize feature representation.

[0057] The open-source label-studio annotation tool was used to annotate the fused image with polygonal impurities, the main types of which included: plastic film, oil cotton, polypropylene fibers, etc.

[0058] The fused image and labeled data are input into the improved YOLOv8 algorithm model, and the first convolutional kernel of the model is modified to accommodate the 5-channel input data. A branch network is then used to perform deeper fusion of RGB and polarization features within the model.

[0059] After polarization features are extracted through polarization branching, they are fully integrated with RGB image information, enhancing the model's adaptability to glossy objects and complex scenes. This method fully integrates color, texture, and optical property information, exhibiting strong robustness and high detection accuracy, making it suitable for multimodal detection tasks in complex scenes.

[0060] Finally, the detection results of the above multiple steps are merged to obtain a binary image of size 2048*512. Taking the industrial control computer 47 as an example, the 2048*512 image is divided into 20*512 regions. The impurity area of ​​each region is counted, and the region with an impurity area greater than a certain level is set to 1, that is, the spray valve of the region is activated. This method can effectively avoid false recognition.

[0061] The above-described specific embodiments are merely preferred embodiments of the present invention and are not intended to limit the implementation of the present invention or the scope of the claims. All equivalent changes and modifications made in accordance with the scope of patent protection of the present invention should be included within the scope of the present invention patent application.

Claims

1. A cotton impurity detection method based on multi-feature fusion, characterized in that, A cotton impurity detection system based on multi-feature fusion is adopted. The system includes a housing (1), a multimodal image acquisition module, a multimodal information processing module, and a cotton transport module. The multimodal image acquisition module includes several RGB cameras, polarization cameras, LED light sources, and UV lamps symmetrically distributed on both sides of the cotton transport pipe (2). The LED light sources and UV lamps are installed alternately with the cotton transport pipe as the center. The multimodal information processing module includes an industrial control computer, a lower-level machine, and a PLC controller. The industrial control computer is connected to the RGB cameras, polarization cameras, and lower-level machine. The cotton transport module includes a fan and an air duct for controlling the cotton transport speed. The detection method includes the following steps: S1: Synchronously trigger the RGB camera and polarization camera to acquire images. The RGB camera outputs a 2048×512 RGB image, and the polarization camera acquires polarization intensity maps at 0°, 45°, 90°, and 135°. The polarization degree map and polarization angle map are then generated by the lower-level computer. S2: Convert the RGB image to the Lab color space, and identify colored impurities and linear impurities similar to the color of cotton based on L, a, and b channel threshold segmentation and template matching; S3: Converts RGB images to HSV color space and detects high-saturation areas based on saturation to identify vibrant color impurities; S4: Align the RGB image with the polarization feature map and fuse them into a five-channel image, input it into a deep learning network for detection, and identify soil-bearing plastic film that was not detected by the polarization camera; S5: Merge the detection results of S2, S3, and S4, generate an array, and send it to the lower-level machine to control the start of the spray valve used to remove impurities; The step S4 of fusing the RGB image and polarization feature map includes: confirming that the size of both the RGB image and the polarization feature map is 2048*512; stacking the aligned RGB image and the polarization feature map according to the channel dimension, wherein the three channels of the RGB image are used as channels 1-3, and the polarization degree DoP map and polarization angle AoP map of the polarization feature map are used as channels 4 and 5; the format of the fused image is [H,W,5], where H is the height and W is the width; the deep learning network adopts an improved YOLOv8 model, optimizing the network input dimension of the original network model structure for RGB three channels to a five-channel image suitable for fusion as input, extracting image features through a deep convolutional network Backbone, optimizing the network model structure, and simultaneously extracting features from RGB and polarization information to generate target recognition.

2. The cotton impurity detection method based on multi-feature fusion according to claim 1, characterized in that: The formulas for calculating the degree of polarization (DoP) and angle of polarization (AoP) in S1 are as follows: The formula for calculating the degree of polarization (DoP) is: ， The formula for calculating the polarization angle (AoP) is: 。 3. The cotton impurity detection method based on multi-feature fusion according to claim 1, characterized in that: The template matching in step S2 specifically includes: smoothing the threshold segmentation results through morphological operations; and matching suspicious areas using predefined linear and clump templates to identify polypropylene fibers, hemp ropes, and oily cotton.

4. The cotton impurity detection method based on multi-feature fusion according to claim 1, characterized in that: The merging detection results in step S5 specifically involve: performing pixel-level OR operations on the binarized image generated by threshold segmentation, template matching, and deep learning detection; performing region segmentation on the merged image; and triggering the spray valve only when the impurity area in a certain region exceeds a preset threshold.

5. The cotton impurity detection method based on multi-feature fusion according to claim 4, characterized in that: In step S5, the result is divided into 20*512 regions. The area of ​​impurities in each region is statistically analyzed, and regions with impurities exceeding a certain level are set to 1, meaning that the spray valve in that region is activated.

6. The cotton impurity detection method based on multi-feature fusion according to claim 1, characterized in that: The training data of the deep learning network is labeled using the Label-Studio tool, and the label types include mulch film, oil cotton, polypropylene yarn, and packaging film.

7. The cotton impurity detection method based on multi-feature fusion according to claim 1, characterized in that: The multimodal image acquisition module includes three RGB cameras and three polarization cameras on both the front and back sides. The RGB cameras and polarization cameras are fixed to the camera brackets, and the brackets and the housing (1) are connected by soft contact to eliminate the influence of mechanical vibration.

8. The cotton impurity detection method based on multi-feature fusion according to claim 1, characterized in that: The specific installation positions of the LED light source and UV lamp are as follows: the first group of light sources is installed on the upper right and upper left sides of the RGB camera line scan position, at a 20° angle to the vertical direction of the pipe; the second group of light sources is installed on the lower right and lower left sides of the RGB camera line scan position, at a 160° angle to the vertical direction of the pipe; the third group of light sources is installed on the upper left and upper right sides of the polarization camera line scan position, at a 20° angle to the vertical direction of the pipe; and the fourth group of light sources is installed on the lower left and lower right sides of the polarization camera line scan position, at a 160° angle to the vertical direction of the pipe.

Images