Compact structure artificial intelligence cloth inspection machine

Abstract

The application relates to the field of cloth inspecting machines, and discloses an artificial intelligence cloth inspecting machine with a compact structure. Cloth feeding rollers, cloth opening rollers and tension rollers are sequentially arranged downstream of a plurality of unwinding rollers. Two winding rollers are arranged directly above the plurality of unwinding rollers, and a plurality of conveying rollers are arranged between the two winding rollers and the tension rollers to guide cloth transmission. A back light source irradiates the bottom surface of the cloth, and a front light source irradiates the top surface of the cloth. A detection camera is used to collect image information of the cloth, the shooting direction of the detection camera and the cloth between the tension roller and the adjacent conveying roller form an acute angle, so that the light reflection of the cloth is reduced, and the definition of the image is improved. The compact structure of the cloth inspecting machine reduces the volume of the cloth inspecting machine, the cloth inspecting machine can pass through a conventional passage in a workshop, one or two people can complete the carrying, and the cloth inspecting machine is suitable for the layout of a small and medium-sized factory.

Description

Technical Field

[0001] This invention relates to the field of fabric inspection machines, and in particular to a compact artificial intelligence fabric inspection machine. Background Technology

[0002] Fabric inspection machines are core equipment in the textile industry's fabric quality inspection process. They are mainly used for defect screening and quality control during the fabric production, dyeing, and garment manufacturing processes. They use a motor-driven roller system to smoothly transport the fabric, and with the help of high-brightness LED or fluorescent light sources, provide inspectors with a clear observation environment. Some intelligent models are equipped with high-definition cameras and AI image recognition technology, which can automatically capture common defects on the fabric such as broken yarns, stains, and holes, significantly reducing the rate of missed inspections by manual inspection.

[0003] CN205633256U discloses an automated fabric inspection machine, which includes a defect detection device for detecting defects in fabrics, a heat-shrink packaging device for heat-shrinking the inspected and formed fabric rolls, and a palletizing and transporting device for stacking and transporting the heat-shrinked fabric rolls. The defect detection device, heat-shrink packaging device, and palletizing and transporting device are connected in sequence. This utility model automates defect detection, reduces labor intensity, and lowers the rates of missed and false detections. It also improves the accuracy of fabric defect detection. The automatic winding mechanism, automatic edge-cutting mechanism, and heat-shrink packaging device work together to wind up the inspected fabric, which is then automatically transferred to the heat-shrink packaging device for heat-shrinking. However, existing technologies still have some shortcomings: traditional fabric inspection machines have a large footprint, making them unsuitable for the compact layout of small and medium-sized textile factories, and require specialized equipment for handling and installation. Therefore, we provide a compact, artificial intelligence fabric inspection machine. Summary of the Invention

[0004] The purpose of this invention is to provide a compact artificial intelligence fabric inspection machine to solve the problems existing in the prior art.

[0005] The technical solution adopted by this invention to solve its technical problem is:

[0006] A compact, AI-powered fabric inspection machine includes: a frame, several unwinding rollers, a feed roller, a tension roller, several conveying rollers, two take-up rollers, a back light source, a front light source, an inspection camera, a spreading roller, and a control system. The unwinding rollers are positioned at the front end of the frame and arranged in an arc shape for holding a first fabric roll. At least one take-up roller is connected to a take-up motor, and the spreading roller is connected to a drive motor.

[0007] The feed roller, the spreading roller, and the tension roller are sequentially arranged downstream of the unwinding rollers. Two take-up rollers are positioned directly above the unwinding rollers to hold the second fabric roll. Several conveyor rollers are positioned between the two take-up rollers and the tension rollers to guide fabric transport. The fabric between the tension roller and the adjacent conveyor roller is inclined. A backlight source is positioned below the fabric between the tension roller and the adjacent conveyor roller to illuminate the bottom surface of the fabric. A frontlight source is positioned above the fabric between the tension roller and the adjacent conveyor roller to illuminate the top surface of the fabric. A detection camera is positioned directly above the fabric between the tension roller and the adjacent conveyor roller to collect image information of the fabric. The camera's shooting direction forms an acute angle with the fabric between the tension roller and the adjacent conveyor roller, thereby reducing fabric reflection and improving image clarity. The control system is electrically connected to the drive motor, the take-up motor, the frontlight source, the backlight source, and the detection camera. The fabric unwound from the first fabric roll sequentially passes around the feed roller, the spreading roller, the tension roller, and the several conveyor rollers before being wound onto the second fabric roll. The compact design of the fabric inspection machine reduces its size, allowing it to be moved by one or two people through regular workshop passageways, making it suitable for small and medium-sized factory layouts.

[0008] Furthermore, the frame is made of aluminum alloy, thereby reducing the weight of the fabric inspection machine.

[0009] Furthermore, the tension roller is equipped with a tension detection sensor.

[0010] Furthermore, the surfaces of several unwinding rollers, the feed rollers, the tension rollers, several conveying rollers, and the two take-up rollers are provided with a silicone anti-slip layer, making the fabric transmission more stable.

[0011] Furthermore, the bottom of the frame is equipped with four casters with brakes.

[0012] Furthermore, the detection camera is a line scan camera.

[0013] Furthermore, the control system includes a main control module, an encoder, and an image processing module.

[0014] An encoder is installed on the conveyor roller shaft to collect the fabric conveying speed in real time and transmit the collected conveying speed signal to the main control module. The main control module is connected to the drive motor, tension detection sensor, winding motor, front light source, back light source, and detection camera. It is used to adjust the line frequency of the detection camera according to the conveying speed signal transmitted by the encoder, so that the detection camera is synchronized with the fabric conveying speed, thereby improving the image quality of the fabric.

[0015] The image processing module is connected to the inspection camera. After receiving the image information collected by the inspection camera, it first performs noise reduction processing on the image information to remove interference signals in the image information, and then uploads the noise-reduced image information to the cloud system for defect detection.

[0016] Furthermore, the formula for calculating the line frequency of the detection camera is as follows:

[0017]

[0018] in The line frequency of the camera is used to detect the number of lines scanned per second, measured in Hertz (Hz). The fabric conveying speed is expressed in millimeters per second. The pixel precision is measured in millimeters per pixel, and can be preset according to different needs and fabric types.

[0019] Furthermore, the cloud system is connected to a defect detection module. After the image processing module uploads the denoised image information to the cloud system, the defect detection module uses a defect recognition algorithm to identify defects such as broken yarn, dotted stains, and holes in the image information. In other words, the image processing module performs noise reduction locally, while the defect detection algorithm is executed in the cloud.

[0020] Furthermore, the defect identification algorithm formula is as follows:

[0021]

[0022] In the formula, As a defect judgment indicator, when It was determined to be defective at that time. The threshold value is the decision threshold obtained through sample training. The grayscale difference between the normalized defect and the background is calculated as follows: ,in This represents the actual grayscale value, ranging from 0 to 255. These represent the minimum and maximum grayscale differences in the sample, respectively. The normalized texture complexity is calculated as follows: ,in To calculate the texture entropy based on the gray-level co-occurrence matrix, the parameters of the gray-level co-occurrence matrix are set as follows: distance d = 1 pixel, angle θ = 0°, 45°, 90°, 135°, taking the average of the four directions to avoid directional bias, and the gray level is 256. These are the minimum and maximum values ​​of the texture entropy in the sample, respectively. The normalized shape factor is calculated as follows: ,in The roundness of the defective area is defined as: A represents the defect area, L represents the perimeter of the defect area, and S takes a value between 0 and 1. For a circular shape, S=1, and the more irregular the shape, the closer S is to 0. These are the minimum and maximum values ​​of roundness in the sample, respectively.

[0023] These are the weighting coefficients for grayscale difference, texture complexity, and shape factor, respectively, satisfying... The numerical values ​​are determined using the Analytic Hierarchy Process (AHP), specifically including: constructing a judgment matrix, using grayscale difference, texture complexity, and shape factor as criteria layer elements, and assigning values ​​using a 1-9 scale, where 1 represents equal importance, 3 represents slightly important, 5 represents significantly important, 7 represents strongly important, 9 represents extremely important, and 2, 4, 6, and 8 are intermediate values. Next, the largest eigenvalue of the judgment matrix and its corresponding eigenvector are calculated, and the weight coefficients are obtained after normalization. Finally, a consistency check is performed, calculating the consistency index CI = (λmax - n) / (n - 1), where λmax is the largest eigenvalue and n is the number of elements in the criterion layer. This index is then compared with the random consistency index RI. If CR = CI / RI < 0.1, the weights are considered effective. This algorithm, through normalization and multi-feature weighted fusion, eliminates the influence of different feature dimensions, improving the accuracy of defect identification and more accurately identifying common defects such as broken yarns, spot stains, and holes.

[0024] The beneficial effects of this invention are:

[0025] 1. The present invention reduces the size of the fabric inspection machine by making its structure compact. It can be moved by one or two people through the conventional workshop passage, making it suitable for the layout of small and medium-sized factories.

[0026] 2. This invention adjusts the line frequency of the line scan camera by adjusting the fabric conveying speed, so that the line scan camera and the fabric conveying speed are synchronized, thereby improving the shooting quality of the fabric.

[0027] 3. This invention avoids the influence of differences in the dimensions of different features on the judgment results by weighted fusion of normalized multiple features, and can identify broken yarns, dotted stains and holes in image information, thus improving the efficiency and accuracy of defect detection. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0029] Explanation of reference numerals in the attached diagram: 1. Frame, 2. Unwinding roller, 3. Feeding roller, 4. Spreading roller, 5. Tension roller, 6. Conveyor roller, 7. Rewinding roller, 8. Backlight, 9. Frontlight, 10. Detection camera, 11. Silicone anti-slip layer, 12. Caster wheel, 13. Second roll of fabric, 14. First roll of fabric. Detailed Implementation

[0030] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification.

[0031] It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings of this specification are merely for illustrative purposes to aid those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to the size, without affecting the effects and objectives achieved by the invention, should still fall within the scope of the technical content disclosed herein. Furthermore, the terms such as "upper," "lower," "left," "right," and "middle" used in this specification are merely for clarity and are not intended to limit the scope of the invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention's implementation.

[0032] like Figure 1 As shown, a compact artificial intelligence fabric inspection machine includes: a frame 1, four unwinding rollers 2, a feed roller 3, a tension roller 5, three conveying rollers 6, two take-up rollers 7, a backlight 8, a frontlight 9, an inspection camera 10, a spreading roller 4, and a control system. The four unwinding rollers 2 are located at the front end of the frame 1 and are arranged in an arc shape for holding the first fabric roll 14. At least one take-up roller 7 is connected to a take-up motor, and the spreading roller 4 is connected to a drive motor.

[0033] The feed roller 3, opening roller 4, and tension roller 5 are sequentially positioned downstream of the four unwind rollers 2. Two take-up rollers 7 are positioned directly above the four unwind rollers 2 to hold the second fabric roll 13. Three conveyor rollers 6 are positioned between the two take-up rollers 7 and the tension rollers 5 to guide the fabric transport. The fabric between the tension roller 5 and the adjacent conveyor roller 6 is inclined. A backlight 8 is positioned below the fabric between the tension roller 5 and the adjacent conveyor roller 6 to illuminate the bottom surface of the fabric. A frontlight 9 is positioned above the fabric between the tension roller 5 and the adjacent conveyor roller 6 to illuminate the top surface of the fabric. A detection camera 10 is positioned directly above the fabric between the tension roller 5 and the adjacent conveyor roller 6 to collect image information of the fabric. The shooting direction of the detection camera 10 forms an acute angle with the fabric between the tension roller 5 and the adjacent conveyor roller 6, thereby reducing fabric reflection and improving image clarity. The control system is electrically connected to the drive motor, the take-up motor, the frontlight 9, the backlight 8, and the detection camera 10. The fabric unwound from the first roll 14 passes sequentially around the feed roller 3, the opening roller 4, the tension roller 5, and the three conveyor rollers 6, and is then wound onto the second roll 13.

[0034] The compact design of the fabric inspection machine reduces its size, allowing it to be moved by one or two people through regular workshop passageways, making it suitable for small and medium-sized factory layouts.

[0035] The frame 1 is made of aluminum alloy, which reduces the weight of the fabric inspection machine.

[0036] Tension roller 5 is equipped with a tension detection sensor.

[0037] The surfaces of the four unwinding rollers 2, the feed rollers 3, the tension rollers 5, the three conveying rollers 6, and the two take-up rollers 7 are provided with a silicone anti-slip layer 11, which makes the transmission of the fabric more stable.

[0038] The bottom of the frame 1 is equipped with four casters 12 with brakes.

[0039] The detection camera 10 is a line scan camera.

[0040] The control system includes a main control module, an encoder, and an image processing module.

[0041] An encoder is installed at axis 6 of the conveyor roller to collect the fabric conveying speed in real time and transmit the collected conveying speed signal to the main control module. The main control module is connected to the drive motor, tension detection sensor, winding motor, front light source 9, back light source 8, and detection camera. It is used to adjust the line frequency of the detection camera according to the conveying speed signal transmitted by the encoder, so that the detection camera is synchronized with the fabric conveying speed, thereby improving the image quality of the fabric.

[0042] The image processing module is connected to the inspection camera. After receiving the image information collected by the inspection camera, it first performs noise reduction processing on the image information to remove interference signals in the image information, and then uploads the noise-reduced image information to the cloud system for defect detection.

[0043] The formula for calculating the line frequency of the camera is:

[0044]

[0045] in The line frequency of the camera is used to detect the number of lines scanned per second, measured in Hertz (Hz). The fabric conveying speed is expressed in millimeters per second. The pixel precision is measured in millimeters per pixel, and can be preset according to different needs and fabric types.

[0046] The cloud system is connected to a defect detection module. After the image processing module uploads the noise-reduced image information to the cloud system, the defect detection module uses a defect recognition algorithm to identify defects such as broken yarn, dotted stains, and holes in the image information.

[0047] The defect identification algorithm formula is:

[0048]

[0049] In the formula, As a defect judgment indicator, when It was determined to be defective at that time. The threshold value is the decision threshold obtained through sample training. The grayscale difference between the normalized defect and the background is calculated as follows: ,in This represents the actual grayscale value, ranging from 0 to 255. These represent the minimum and maximum grayscale differences in the sample, respectively. The normalized texture complexity is calculated as follows: ,in To calculate the texture entropy based on the gray-level co-occurrence matrix, the parameters of the gray-level co-occurrence matrix are set as follows: distance d = 1 pixel, angle θ = 0°, 45°, 90°, 135°, taking the average of the four directions to avoid directional bias, and the gray level is 256. These are the minimum and maximum values ​​of the texture entropy in the sample, respectively. The normalized shape factor is calculated as follows: ,in The roundness of the defective area is defined as: A represents the defect area, L represents the perimeter of the defect area, and S takes a value between 0 and 1. For a circular shape, S=1, and the more irregular the shape, the closer S is to 0. These are the minimum and maximum values ​​of roundness in the sample, respectively.

[0050] These are the weighting coefficients for grayscale difference, texture complexity, and shape factor, respectively, satisfying... The numerical values ​​are determined using the Analytic Hierarchy Process (AHP), specifically including: constructing a judgment matrix, using grayscale difference, texture complexity, and shape factor as criteria layer elements, and assigning values ​​using a 1-9 scale, where 1 represents equal importance, 3 represents slightly important, 5 represents significantly important, 7 represents strongly important, 9 represents extremely important, and 2, 4, 6, and 8 are intermediate values. Next, the largest eigenvalue of the judgment matrix and its corresponding eigenvector are calculated, and the weight coefficients are obtained after normalization. Finally, a consistency check is performed, calculating the consistency index CI = (λmax - n) / (n - 1), where λmax is the largest eigenvalue and n is the number of elements in the criterion layer. This index is then compared with the random consistency index RI. If CR = CI / RI < 0.1, the weights are considered effective. This algorithm, through normalization and multi-feature weighted fusion, eliminates the influence of different feature dimensions, enabling efficient and accurate identification of common defects such as broken yarns, spot stains, and holes.

[0051] Work methods:

[0052] The encoder collects the fabric conveying speed in real time and transmits the collected speed signal to the main control module. The main control module adjusts the line frequency of the detection camera based on the speed signal transmitted by the encoder. The formula for calculating the line frequency of the detection camera is as follows:

[0053]

[0054] in The line frequency of the camera is used to detect the number of lines scanned per second, measured in Hertz (Hz). The fabric conveying speed is expressed in millimeters per second. Pixel precision, unit is millimeters per pixel.

[0055] The camera takes pictures of the fabric and then transmits the collected image information to the image processing module. The image processing module first performs noise reduction on the image information to remove interference signals. Then, the noise-reduced image information is uploaded to the cloud system for defect detection. The defect detection module uses defect recognition algorithms to identify defects such as broken yarn, spot stains, and holes in the image information.

[0056] The defect identification algorithm formula is:

[0057]

[0058] In the formula, As a defect judgment indicator, when It was determined to be defective at that time. The threshold value is the decision threshold obtained through sample training. The grayscale difference between the normalized defect and the background is calculated as follows: ,in This represents the actual grayscale value, ranging from 0 to 255. These represent the minimum and maximum grayscale differences in the sample, respectively. The normalized texture complexity is calculated as follows: ,in To calculate the texture entropy based on the gray-level co-occurrence matrix, the parameters of the gray-level co-occurrence matrix are set as follows: distance d = 1 pixel, angle θ = 0°, 45°, 90°, 135°, taking the average of the four directions to avoid directional bias, and the gray level is 256. These are the minimum and maximum values ​​of the texture entropy in the sample, respectively. The normalized shape factor is calculated as follows: ,in The roundness of the defective area is defined as: A represents the defect area, L represents the perimeter of the defect area, and S takes a value between 0 and 1. For a circular shape, S=1, and the more irregular the shape, the closer S is to 0. These are the minimum and maximum values ​​of roundness in the sample, respectively.

[0059] These are the weighting coefficients for grayscale difference, texture complexity, and shape factor, respectively, satisfying... The numerical values ​​are determined using the Analytic Hierarchy Process (AHP), specifically including: constructing a judgment matrix, using grayscale difference, texture complexity, and shape factor as criteria layer elements, and assigning values ​​using a 1-9 scale, where 1 represents equal importance, 3 represents slightly important, 5 represents significantly important, 7 represents strongly important, 9 represents extremely important, and 2, 4, 6, and 8 are intermediate values. Next, the largest eigenvalue of the judgment matrix and its corresponding eigenvector are calculated, and the weight coefficients are obtained after normalization. Finally, a consistency check is performed, calculating the consistency index CI = (λmax - n) / (n - 1), where λmax is the largest eigenvalue and n is the number of elements in the criterion layer, and comparing it with the random consistency index RI. When CR = CI / RI < 0.1, the weights are confirmed to be effective.

[0060] The above embodiments are only some embodiments of the present invention, and not all embodiments. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the scope of protection of the present invention.

Claims

1. A compact artificial intelligence fabric inspection machine, comprising: The machine includes a frame (1), several unwinding rollers (2), a feed roller (3), a tension roller (5), several conveyor rollers (6), two take-up rollers (7), a back light source (8), a front light source (9), a detection camera (10), and a control system. The several unwinding rollers (2) are located at the front end of the frame (1) and are arranged in an arc shape for placing the first roll of fabric (14). At least one take-up roller (7) is connected to a take-up motor. The feature is that it further includes a spreading roller (4), which is connected to a drive motor; The feed roller (3), the opening roller (4), and the tension roller (5) are sequentially arranged downstream of the unwinding rollers (2); two take-up rollers (7) are arranged directly above the unwinding rollers (2) for placing the second fabric roll (13); several conveyor rollers (6) are arranged between the two take-up rollers (7) and the tension rollers (5); the fabric between the tension roller (5) and the adjacent conveyor roller (6) is inclined; a backlight (8) is arranged below the fabric between the tension roller (5) and the adjacent conveyor roller (6); a frontlight (9) is arranged between the tension roller (5) and the adjacent conveyor roller (6). Above the fabric between 6); the detection camera (10) is set directly above the fabric between the tension roller (5) and the adjacent conveyor roller (6) to collect image information of the fabric. The shooting direction of the detection camera (10) forms an acute angle with the fabric between the tension roller (5) and the adjacent conveyor roller (6); the control system is electrically connected to the drive motor, the winding motor, the front light source (9), the back light source (8) and the detection camera (10); the fabric unwound from the first fabric roll (14) passes through the feed roller (3), the opening roller (4), the tension roller (5) and several conveyor rollers (6) in sequence, and is wound onto the second fabric roll (13); The control system includes a main control module, an encoder, and an image processing module; The encoder is set at the shaft of the conveyor roller (6) to collect the conveying speed of the fabric in real time and transmit the collected conveying speed signal to the main control module. The main control module is connected to the drive motor, tension detection sensor, winding motor, front light source (9), back light source (8) and detection camera (10) to adjust the line frequency of the detection camera according to the conveying speed signal transmitted by the encoder. The image processing module is connected to the inspection camera. After receiving the image information collected by the inspection camera, it first performs noise reduction on the image information, and then uploads the noise-reduced image information to the cloud system for defect detection. The cloud system is connected to a defect detection module. After the image processing module uploads the noise-reduced image information to the cloud system, the defect detection module uses a defect recognition algorithm to identify defects such as broken yarn, dotted stains, and holes in the image information. The defect identification algorithm formula is: In the formula, As a defect judgment indicator, when It was determined to be defective at that time. The decision threshold is obtained through sample training; The grayscale difference between the normalized defect and the background is calculated as follows: ,in This represents the actual grayscale value, ranging from 0 to 255. These represent the minimum and maximum grayscale differences in the sample, respectively. The normalized texture complexity is calculated as follows: ,in To calculate the texture entropy based on the gray-level co-occurrence matrix, the parameters of the gray-level co-occurrence matrix are set as follows: distance d = 1 pixel, angle θ = 0°, 45°, 90°, 135°, taking the average of the four directions to avoid directional bias, and the gray level is 256. These are the minimum and maximum values ​​of the texture entropy in the sample, respectively; The normalized shape factor is calculated as follows: ,in The roundness of the defective area is defined as: A represents the defect area, L represents the perimeter of the defect area, and S takes a value of 0-1. For a circular shape, S=1, and the more irregular the shape, the closer S is to 0. These are the minimum and maximum values ​​of roundness in the sample, respectively; These are the weighting coefficients for grayscale difference, texture complexity, and shape factor, respectively, satisfying... The numerical values ​​are determined using the Analytic Hierarchy Process (AHP), specifically including: constructing a judgment matrix, using grayscale difference, texture complexity, and shape factor as criteria layer elements, and assigning values ​​using a 1-9 scale, where 1 represents equal importance, 3 represents slightly important, 5 represents significantly important, 7 represents strongly important, 9 represents extremely important, and 2, 4, 6, and 8 are intermediate values; then, the largest eigenvalue of the judgment matrix and its corresponding eigenvector are calculated, and the weight coefficients are obtained after normalization. Finally, a consistency check is performed, and the consistency index CI = (λmax - n) / (n - 1) is calculated, where λmax is the largest eigenvalue and n is the number of elements in the criterion layer. The consistency index CI is then compared with the random consistency index RI. When CR = CI / RI < 0.1, the weights are confirmed to be effective.

2. The compact artificial intelligence fabric inspection machine according to claim 1, characterized in that: The frame (1) is made of aluminum alloy.

3. The compact artificial intelligence fabric inspection machine according to claim 1, characterized in that: The tension roller (5) is equipped with a tension detection sensor.

4. The compact artificial intelligence fabric inspection machine according to claim 1, characterized in that: The surfaces of several unwinding rollers (2), several feed rollers (3), several tension rollers (5), several conveying rollers (6), and two take-up rollers (7) are provided with a silicone anti-slip layer (11).

5. The compact artificial intelligence fabric inspection machine according to claim 1, characterized in that: The frame (1) is equipped with four casters (12) with brakes at the bottom.

6. The compact artificial intelligence fabric inspection machine according to claim 1, characterized in that: The detection camera (10) is a line scan camera.

Images