Industrial fabric net disease data detection system based on image recognition analysis

Abstract

The invention discloses an industrial fabric mesh disease data detection system based on image recognition analysis, relates to the technical field of mesh disease data detection, solves the technical problem that automatic focusing regulation and control cannot be carried out when industrial fabric defects are collected and detected in the prior art, and particularly relates to the technical field of automatic focusing regulation and control when the industrial fabric defects are collected and detected. Whether the current industrial fabric has defects or not is deduced according to real-time fabric picture collection, so that quality evaluation of the industrial fabric is facilitated, the processing quality of the industrial fabric is ensured, the processing effectiveness of the industrial fabric is improved according to real-time processing defect detection, and the yield of fabric output is ensured; the collected pictures are subjected to net disease type detection and classification, and net diseases existing in the industrial fabric are deduced according to real-time picture defect detection, so that net disease prevention and control reorganization can be carried out in time, the net disease generation rate of an industrial fabric production line is reduced, and the yield and qualification rate of the industrial fabric are ensured.

Description

Technical Field

[0001] The invention relates to the technical field of web disease data detection, and in particular to an industrial fabric web disease data detection system based on image recognition and analysis. Background Art

[0002] As an industrial fabric with a special textile structure, the forming net is interwoven with polyester yarns of different diameters. In the process of weaving the yarns into nets, due to the quality problems of the polyester yarns themselves, the vibration of textile machinery, and improper operation of workshop workers, net defects such as stains and yarn breakage are prone to occur.

[0003] However, in the existing technology, it is not possible to collect and detect defects in industrial fabrics during production, and it is not possible to perform automatic focus adjustment during detection, so the detection accuracy cannot be guaranteed. At the same time, it is not possible to classify the types of network diseases, and it is impossible to increase training samples for classification inspection. In addition, it is impossible to trace the source of network diseases and prevent and control them.

[0004] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0005] The purpose of the present invention is to solve the above-mentioned problems and to propose an industrial fabric web disease data detection system based on image recognition and analysis.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] The industrial fabric web disease data detection system based on image recognition analysis includes a fabric detection center, wherein the fabric detection center is connected to:

[0008] The defect collection and detection unit is used to collect and detect defects on industrial fabrics. The image collection device is set. At the start time of defect collection, the center of the collection area is used as the focus center of the image collection device to collect images of the moving industrial fabrics. During the image collection, the real-time moving industrial fabrics are collected and tested. The collected images of the industrial fabrics are compared with the standard completed fabric images, and the collected images are divided into defective images and non-defective images.

[0009] The mesh defect type classification unit is used to detect and classify the mesh defect types of the collected images, and uniformly mark the inner weft, bottom weft and continuous warp of industrial fabrics as fabric components; distinguish the mesh defect types by position positioning, and send the corresponding parts to the fabric testing center. The mesh defect types are external force defects, material defects and stain defects;

[0010] The sample adds a detection unit, which is used to add training samples based on the collected pictures, and perform lightening, darkening, horizontal flipping, vertical flipping, and center symmetry processing on the collected pictures of industrial fabrics to obtain the newly added pictures, and infer whether the classification of network disease types is consistent through comparison. If yes, trace the source, otherwise, re-collect and classify;

[0011] The network disease tracing and analysis unit is used to perform tracing and analysis on the repaired network diseases.

[0012] As a preferred embodiment of the present invention, for the collected pictures, the focus center of the industrial fabric picture is taken as the midpoint, and the pixel points of the surrounding pictures are counted, and the corresponding azimuth pixels are guaranteed to be on the same track, and the numerical fluctuation of the corresponding pixel value of the real-time pixel point on the current track is carried out, and the fluctuation curve is constructed according to the pixel value of each pixel point;

[0013] If the current fluctuation curve shows a reciprocating floating trend and the peak-to-valley value fluctuation is within the set range, the pixel points corresponding to the downward trend of the curve are set as occasional fluctuations, and the adjacent collected pictures of the current picture are detected. If there is no pixel value fluctuation at the corresponding pixel point, it is determined to be occasional fluctuation;

[0014] If the current fluctuation curve shows a continuous downward trend and the valley value fluctuation exceeds the set range, the pixel points corresponding to the downward trend of the curve will be set as influential fluctuations, and the adjacent collected pictures of the current picture will be detected. If the corresponding pixel points all have pixel value fluctuations, they are determined to be influential fluctuations.

[0015] As a preferred embodiment of the present invention, if the pixel floating of the pixel points corresponding to the moving track orientation has not decreased, and the pixel floating of the pixel points in the non-moving track orientation has not increased, the moving speed of the focus center is increased; if the pixel floating of the pixel points corresponding to the moving track orientation has not decreased, and the pixel floating of the pixel points in the non-moving track orientation has increased, the movement of the focus center is stopped and the specifications of the image acquisition device are adjusted or the area of ​​the acquisition area is controlled;

[0016] If the pixel floating of the pixel point corresponding to the moving track direction decreases, and the pixel floating of the pixel point in the non-moving track direction does not increase, the moving speed of the focus center remains constant and stops moving after the pixel point in the non-moving track direction has pixel floating;

[0017] If the pixel float of the pixel point corresponding to the moving track orientation decreases, and the pixel float of the pixel point in the non-moving track orientation increases, the moving speed of the focusing center is reduced, and the movement is stopped when the pixel float of the pixel point in the non-moving track orientation is at the red line value. At this time, the pixel float of the pixel point corresponding to the moving track orientation is not controlled within the red line value range, then the image acquisition device is adjusted or the acquisition area is controlled.

[0018] As a preferred embodiment of the present invention, the network disease type classification unit process is as follows:

[0019] According to the real-time missing position, the fabric component endpoints of the fabric component adjacent to the missing position are collected. If the cut of the corresponding fabric component endpoint is rough and the existing part of the current fabric component is deformed by force, the corresponding defect is set as a material defect; if the cut of the corresponding fabric component endpoint is smooth and the existing part of the current fabric component is not deformed by force, the corresponding defect is set as an external force defect;

[0020] The occluded position of the collected image in the non-defective image is located. If the current occluded position corresponds to an image composed of multiple adjacent fabrics, and the displayed fabric components are connected but there is an occluded position in the area, the corresponding occluded position will be marked as a stain defect.

[0021] As a preferred embodiment of the present invention, the process of adding a detection unit to a sample is as follows:

[0022] The newly added pictures and the collected pictures are classified into network disease types. If the network type classifications of the newly added pictures and the collected pictures are consistent, the network disease type received by the fabric testing center will determine the part of the corresponding collected picture as the network disease part, and the fabric testing center will send a network disease repair instruction to the administrator terminal; if the network type classifications of the newly added pictures and the collected pictures are inconsistent, the industrial fabrics corresponding to the collected pictures will be re-collected and a non-unique number of environmental conditions will be set during the collection process.

[0023] As a preferred embodiment of the present invention, the network disease tracing and analysis unit process is as follows:

[0024] Distance information and quantity information are collected, and the distance information and quantity information are respectively the distance deviation of the corresponding web defect position during the image collection process of industrial fabrics of the same specification but different batches, and the proportion of industrial fabrics with the same processing step corresponding to the web defect position of industrial fabrics of the same specification but different batches.

[0025] As a preferred embodiment of the present invention, if the distance information exceeds the distance deviation threshold, and the quantity information exceeds the quantity ratio threshold, a process traceability control signal is generated; if the distance information exceeds the distance deviation threshold, and the quantity information does not exceed the quantity ratio threshold, a processing environment impact signal is generated;

[0026] If the distance information does not exceed the distance deviation threshold, and the quantity information does not exceed the quantity proportion threshold, a material adjustment signal is generated; if the distance information difference does not exceed the distance deviation threshold, and the quantity information exceeds the quantity proportion threshold, an adaptation scheduling signal is generated.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1. In the present invention, defect collection and detection are performed on industrial fabrics, and whether the current industrial fabrics have defects is inferred based on real-time fabric image collection, so as to facilitate quality assessment of industrial fabrics, ensure the processing quality of industrial fabrics, improve the effectiveness of industrial fabric processing based on real-time processing defect detection, and ensure the qualified rate of fabric output;

[0029] The collected images are detected and classified by mesh defect type, and the mesh defects existing in industrial fabrics are inferred based on real-time image defect detection, so as to facilitate timely mesh defect prevention and control, reduce the mesh defect rate of industrial fabric production lines, and ensure the qualified rate of industrial fabric production.

[0030] 2. In the present invention, training samples are added on the basis of collected pictures, so that the sample addition based on the collected pictures improves the accuracy of network disease detection, avoids the deviation of detection caused by factors such as the environment in the current picture collection process, thereby reducing the accuracy of network disease detection, and improves the detection accuracy by increasing samples, ensuring the credibility of the corresponding collected pictures in various environments;

[0031] Conduct source tracing analysis on the network disease being repaired to infer whether the cause of the current network disease type persists so as to facilitate timely control. Ensure that when repairing the network disease, source tracing control is performed based on the cause of the network disease, effectively reducing the probability of a secondary occurrence of the network disease at the current location. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0033] Figure 1 It is a system principle block diagram of the present invention;

[0034] Figure 2 It is a schematic diagram of the operation method of the system of the present invention;

[0035] Figure 3 It is a schematic diagram of the image stretching scheme 1 in the present invention;

[0036] Figure 4 This is a schematic diagram of the second image stretching scheme in the present invention. DETAILED DESCRIPTION

[0037] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0039] See also Figure 1-2 As shown, the industrial fabric web disease data detection system based on image recognition analysis includes a fabric detection center, wherein the fabric detection center is connected to a defect collection detection unit, a web disease type classification unit, a sample additional detection unit, and a web disease tracing analysis unit; and each unit cooperates with each other and performs the next system processing according to the real-time generated signal to complete a comprehensive detection method in the detection system;

[0040] The fabric inspection center generates a defect collection detection signal and sends the defect collection detection signal to the defect collection detection unit. After receiving the defect collection detection signal, the defect collection detection unit performs defect collection detection on the industrial fabric, and infers whether the current industrial fabric has defects based on real-time fabric image collection, so as to facilitate the quality assessment of the industrial fabric and ensure the processing quality of the industrial fabric. The effectiveness of industrial fabric processing is improved based on real-time processing defect detection, and the qualified rate of fabric output is ensured;

[0041] Set up the image acquisition device. The image acquisition device in this application is: the camera model is Hikvision MV-CH120-10GM / GC, an array camera with a resolution of 4096×3000 and an FPS of 9.4; the lens model is Hikvision SA5028M-20MP; the light source model is Xiangrui Technology PDC3-10024-4-232, the sampled image format is BMP, and the size is about 12m;

[0042] According to the collection area reserved by the image collection device, the industrial fabric is filled in the area and the passing speed is set, the industrial fabric is controlled to pass through the collection area at the rated speed, and the collection frequency of the image collection device is set according to the real-time set speed to ensure the collection clarity. It can be understood that in the actual detection process, the passing speed of the industrial fabric is controlled according to the real-time progress requirements and the collection frequency of the image collection device is accelerated to increase the detection rate to ensure the completion of the real-time progress requirements;

[0043] At the start time of defect collection, the center of the collection area is used as the focus center of the image collection device to collect images of the moving industrial fabric; the industrial fabric is composed of the surface weft, bottom weft and continuous warp;

[0044] During image collection, real-time mobile industrial fabrics are collected and tested;

[0045] For the collected pictures, the focus center of the industrial fabric picture is taken as the midpoint, and the pixel points of the surrounding pictures are counted, and the corresponding azimuth pixels are guaranteed to be on the same track. The corresponding pixel values ​​of the real-time pixel points on the current track fluctuate, and the fluctuation curve is constructed according to the pixel values ​​of each pixel point;

[0046] If the current fluctuation curve shows a reciprocating floating trend and the peak-to-valley value fluctuation is within the set range, the pixel points corresponding to the downward trend of the curve are set as occasional fluctuations, and the adjacent collected pictures of the current picture are detected. If there is no pixel value fluctuation at the corresponding pixel point, it is determined to be occasional fluctuation;

[0047] If the current fluctuation curve shows a continuous downward trend, and the valley value fluctuation exceeds the set range, the pixel points corresponding to the downward trend of the curve are set as influential fluctuations, and the adjacent collected pictures of the current picture are detected. If the corresponding pixel points all have pixel value fluctuations, they are determined to be influential fluctuations; the focus center of the pixel points is moved with the direction of the trajectory where the current pixel point is located, and the pixel fluctuations of the pixel points corresponding to the moving trajectory direction and the pixel fluctuations of the pixel points in the non-moving trajectory direction are monitored;

[0048] It can be understood that if the pixel floating of the pixel points corresponding to the moving track orientation does not decrease, and the pixel floating of the pixel points in the non-moving track orientation does not increase, the moving speed of the focus center is increased; if the pixel floating of the pixel points corresponding to the moving track orientation does not decrease, and the pixel floating of the pixel points in the non-moving track orientation increases, the movement of the focus center is stopped and the specifications of the image acquisition device are adjusted or the area of ​​the acquisition area is controlled;

[0049] If the pixel floating of the pixel point corresponding to the moving track direction decreases, and the pixel floating of the pixel point in the non-moving track direction does not increase, the moving speed of the focus center remains constant and stops moving after the pixel point in the non-moving track direction has pixel floating;

[0050] If the pixel floating of the pixel point corresponding to the moving track orientation decreases, and the pixel floating of the pixel point in the non-moving track orientation increases, the moving speed of the focus center is reduced, and the movement is stopped when the pixel floating of the pixel point in the non-moving track orientation is within the red line value. At this time, if the pixel floating of the pixel point corresponding to the moving track orientation is not controlled within the red line value range, the image acquisition device is adjusted or the acquisition area is controlled;

[0051] Compare the collected pictures of industrial fabrics with the pictures of standard completed fabrics, and divide the collected pictures into defective pictures and non-defective pictures; send them to the fabric testing center for backup, and generate a network disease type analysis signal, and send the network disease type analysis signal to the network disease type analysis unit. After receiving the network disease type analysis signal, the network disease type analysis unit performs network disease type detection and classification on the collected pictures, and infers the network diseases existing in the industrial fabrics based on the real-time picture defect detection, so as to facilitate timely network disease prevention and control, reduce the network disease generation rate of the industrial fabric production line, and ensure the qualified rate of industrial fabric production;

[0052] The inner weft, bottom weft and connecting warp of industrial fabrics are uniformly marked as fabric components;

[0053] The missing position of a single fabric component in the defect image is located, and the fabric component endpoints of the fabric component adjacent to the missing position are collected according to the real-time missing position. If the cut of the corresponding fabric component endpoint is rough and the existing part of the current fabric component is deformed by force, the corresponding defect is set as a material defect, where the force deformation is represented by the increase in the length or decrease in the thickness of the component due to the force on both ends of the fabric component; specifically, the material defect is that the fabric component material itself cannot withstand the current force, and the fibers of the fabric component are gradually broken;

[0054] If the cut at the endpoint of the corresponding fabric component is smooth and the existing part of the current fabric component does not have stress deformation, the corresponding defect is set as an external force defect; specifically, the defect is caused by external force defects such as hard object damage;

[0055] The occluded position of the collected image in the non-defective image is located. If the current occluded position corresponds to an image of multiple adjacent fabrics, and the fabrics are connected but there is an occluded position in the area, the corresponding occluded position is marked as a stain defect. Based on the number of layers constructed by the fabric composition and the occluded position area at different viewing angles, the coverage of the current stain defect is inferred and sent to the fabric inspection center.

[0056] External force defects, material defects and stain defects are marked as mesh defect types; the corresponding parts of the collected images are sent to the fabric testing center according to the mesh defect types;

[0057] At the same time, a sample addition detection signal is generated and sent to a sample addition detection unit. After receiving the sample addition detection signal, the sample addition detection unit adds training samples on the basis of the collected images, so that the sample addition is based on the collected images to improve the accuracy of network disease detection, avoid the current image collection process The environment and other factors cause the detection deviation, thereby reducing the accuracy of network disease detection, and improve the detection accuracy by adding samples, ensuring the credibility of the corresponding collected images under various environments;

[0058] The collected pictures of industrial fabrics are brightened, darkened, horizontally flipped, vertically flipped, and centrally symmetrically processed, and the processed collected pictures are set as new pictures; the horizontal flipping, vertical flipping, and centrally symmetrical processing are obtained by moving the collected pictures;

[0059] Gamma transform is used to brighten and darken the image to simulate the situation of excessive light and insufficient light. The principle is as follows:

[0060] represents the input pixel value, Represents the output pixel value. The three channels of the image are 0-255 for a total of 256 gray levels. When the image is set to 1, the low-brightness part is compressed, the high-brightness part is stretched, and the overall image brightness is reduced; The low brightness part is stretched and the high brightness part is compressed, and the overall brightness of the image is improved. To achieve three different degrees of brightening effect: strong, medium and weak; when darkening, Choose to achieve different degrees of darkening effect; this application is as follows Figure 3 and Figure 4 Add pictures after collecting and processing the pictures;

[0061] The newly added pictures and the collected pictures are classified into network disease types. If the network types of the newly added pictures and the collected pictures are consistent, the network disease type received by the fabric testing center will be determined as the network disease part corresponding to the part of the collected picture, and the fabric testing center sends a network disease repair instruction to the administrator terminal;

[0062] If the network type classification of the newly added image and the collected image is inconsistent, the industrial fabric corresponding to the collected image will be re-collected and a non-unique number of environmental conditions will be set during the collection process. The environmental conditions are represented by environmental conditions with different environmental parameters and influence, such as the light intensity of the fill light, etc.;

[0063] During the repair process of the network disease, the fabric testing center generates a network disease traceability analysis signal and sends the network disease traceability analysis signal to the network disease traceability analysis unit. After receiving the network disease traceability analysis signal, the network disease traceability analysis unit conducts a traceability analysis on the repaired network disease to infer whether the cause of the current network disease type persists, so as to facilitate timely control and ensure that when repairing the network disease, traceability control is carried out according to the cause of the network disease, effectively reducing the probability of the network disease reappearing at the current location;

[0064] The distance deviation of the corresponding web defect position in the process of collecting images of industrial fabrics of the same specification but different batches is obtained, and the proportion of industrial fabrics with the same processing procedure corresponding to the web defect position of industrial fabrics of the same specification but different batches is obtained, and the distance deviation of the corresponding web defect position in the process of collecting images of industrial fabrics of the same specification but different batches and the proportion of industrial fabrics with the same processing procedure corresponding to the web defect position of industrial fabrics of the same specification but different batches are marked as distance information and quantity information, respectively, and compared with the distance deviation threshold and the quantity proportion threshold, respectively:

[0065] If the distance deviation of the corresponding web defect position during the image collection process of industrial fabrics of the same specification but different batches exceeds the distance deviation threshold, and the proportion of industrial fabrics with the same processing process corresponding to the web defect position of industrial fabrics of the same specification but different batches exceeds the quantity proportion threshold, it is inferred that there is an impact on the corresponding process of the current industrial fabric processing production line, and the specific impact is the equipment impact or the current process processing impact. A process traceability control signal is generated and sent to the fabric testing center. The fabric testing center pre-adjusts the process of the current industrial fabric processing production line. If the web defect rate decreases after the pre-adjustment, the corresponding pre-adjustment will be comprehensively rectified;

[0066] If the distance deviation of the corresponding web defect position during the image collection process of industrial fabrics of the same specification but different batches exceeds the distance deviation threshold, and the proportion of industrial fabrics with the same processing process corresponding to the web defect position of industrial fabrics of the same specification but different batches does not exceed the quantity proportion threshold, it is inferred that there is an impact on the current industrial fabric processing production line, and a processing environment impact signal is generated and sent to the fabric testing center. After receiving the processing environment impact signal, the fabric testing center controls the processing environment of the current industrial fabric to avoid external impacts;

[0067] If the distance deviation of the corresponding web defect position during the image collection process of industrial fabrics of the same specification but different batches does not exceed the distance deviation threshold, and the proportion of industrial fabrics with the same processing process corresponding to the web defect position of industrial fabrics of the same specification but different batches does not exceed the quantity proportion threshold, it is inferred that the material of the current batch of industrial fabrics has an impact, and a material adjustment signal is generated and sent to the fabric testing center. After receiving the material adjustment signal, the fabric testing center replaces the material for the processing of the industrial fabric of the current specification to avoid the failure to meet the force at the same position, resulting in defects;

[0068] If the distance deviation of the corresponding web defect position during the image collection process of industrial fabrics of the same specification but different batches does not exceed the distance deviation threshold, and the proportion of industrial fabrics with the same processing process corresponding to the web defect position of industrial fabrics of the same specification but different batches exceeds the quantity proportion threshold, it is inferred that the current industrial fabric processing process is not compatible with the current material, and an adaptation scheduling signal is generated and sent to the fabric detection center. After receiving the adaptation scheduling signal, the fabric detection center adjusts the material of the current processing production line. If the material cannot be adjusted, the process execution process is optimized to reduce the stress on the fabric components produced under this process;

[0069] When the present invention is used, the defect collection and detection unit performs defect collection and detection on the industrial fabric, sets a picture collection device, and at the start time of defect collection, takes the center of the collection area as the focus center of the picture collection device to collect pictures of the moving industrial fabric; during the picture collection, the real-time moving industrial fabric is collected and detected; the collected pictures of the industrial fabric are compared with the standard completed fabric pictures, and the collected pictures are divided into defect pictures and non-defect pictures;

[0070] The network disease type classification unit detects and classifies the network disease types of the collected images; distinguishes network disease types through position positioning; the sample addition detection unit adds training samples based on the collected images, brightens, darkens, horizontally flips, vertically flips, and performs center symmetry processing on the collected images of industrial fabrics to obtain newly added images, and infers whether the network disease type classification is consistent through comparison. If yes, it traces the source, otherwise, it re-collects and classifies; the network disease tracing analysis unit is used to trace the source of the repaired network disease.

[0071] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. Industrial fabric web disease data detection system based on image recognition analysis, characterized in that: Including fabric testing center, where the fabric testing center communication connections are: The defect collection and detection unit is used to collect and detect defects on industrial fabrics, set up a picture collection device, and at the start time of defect collection, use the center of the collection area as the focus center of the picture collection device to collect pictures of the moving industrial fabrics; During the image collection, the real-time mobile industrial fabrics are collected and tested; the collected images of the industrial fabrics are compared with the standard finished fabric images, and the collected images are divided into defective images and non-defective images; The mesh defect type classification unit is used to detect and classify the mesh defect types of the collected images, and uniformly mark the inner weft, bottom weft and continuous warp of industrial fabrics as fabric components; distinguish the mesh defect types by position positioning, and send the corresponding parts to the fabric testing center. The mesh defect types are external force defects, material defects and stain defects; The sample adds a detection unit, which is used to add training samples based on the collected pictures, and perform lightening, darkening, horizontal flipping, vertical flipping, and center symmetry processing on the collected pictures of industrial fabrics to obtain the newly added pictures, and infer whether the classification of network disease types is consistent through comparison. If yes, trace the source, otherwise, re-collect and classify; The network disease tracing and analysis unit is used to perform tracing and analysis on the repaired network diseases.

2. The industrial fabric web disease data detection system based on image recognition analysis according to claim 1 is characterized in that: For the collected pictures, the focus center of the industrial fabric picture is taken as the midpoint, and the pixel points of the surrounding pictures are counted, and the corresponding azimuth pixels are guaranteed to be on the same track. The corresponding pixel values ​​of the real-time pixel points on the current track fluctuate, and the fluctuation curve is constructed according to the pixel values ​​of each pixel point; If the current fluctuation curve shows a reciprocating floating trend and the peak-to-valley value fluctuation is within the set range, the pixel points corresponding to the downward trend of the curve are set as occasional fluctuations, and the adjacent collected pictures of the current picture are detected. If there is no pixel value fluctuation at the corresponding pixel point, it is determined to be occasional fluctuation; If the current fluctuation curve shows a continuous downward trend and the valley value fluctuation exceeds the set range, the pixel points corresponding to the downward trend of the curve will be set as influential fluctuations, and the adjacent collected pictures of the current picture will be detected. If the corresponding pixel points all have pixel value fluctuations, they are determined to be influential fluctuations.

3. The industrial fabric web disease data detection system based on image recognition analysis according to claim 2 is characterized in that: If the pixel floating of the pixel points corresponding to the moving track orientation has not decreased, and the pixel floating of the pixel points in the non-moving track orientation has not increased, the moving speed of the focus center is increased; if the pixel floating of the pixel points corresponding to the moving track orientation has not decreased, and the pixel floating of the pixel points in the non-moving track orientation has increased, the movement of the focus center is stopped and the specifications of the image acquisition device are adjusted or the area of ​​the acquisition area is controlled; If the pixel floating of the pixel point corresponding to the moving track direction decreases, and the pixel floating of the pixel point in the non-moving track direction does not increase, the moving speed of the focus center remains constant and stops moving after the pixel point in the non-moving track direction has pixel floating; If the pixel float of the pixel point corresponding to the moving track orientation decreases, and the pixel float of the pixel point in the non-moving track orientation increases, the moving speed of the focusing center is reduced, and the movement is stopped when the pixel float of the pixel point in the non-moving track orientation is at the red line value. At this time, the pixel float of the pixel point corresponding to the moving track orientation is not controlled within the red line value range, then the image acquisition device is adjusted or the acquisition area is controlled.

4. The industrial fabric web disease data detection system based on image recognition analysis according to claim 1 is characterized in that: The process of classifying net disease types is as follows: According to the real-time missing position, the fabric component endpoints of the fabric component adjacent to the missing position are collected. If the cut of the corresponding fabric component endpoint is rough and the existing part of the current fabric component is deformed by force, the corresponding defect is set as a material defect; if the cut of the corresponding fabric component endpoint is smooth and the existing part of the current fabric component is not deformed by force, the corresponding defect is set as an external force defect; The occluded position of the collected image in the non-defective image is located. If the current occluded position corresponds to an image composed of multiple adjacent fabrics, and the displayed fabric components are connected but there is an occluded position in the area, the corresponding occluded position will be marked as a stain defect.

5. The industrial fabric web disease data detection system based on image recognition analysis according to claim 1 is characterized in that: The process of adding a test unit to a sample is as follows: The newly added pictures and the collected pictures are classified into network disease types. If the network type classifications of the newly added pictures and the collected pictures are consistent, the network disease type received by the fabric testing center will determine the part of the corresponding collected picture as the network disease part, and the fabric testing center will send a network disease repair instruction to the administrator terminal; if the network type classifications of the newly added pictures and the collected pictures are inconsistent, the industrial fabrics corresponding to the collected pictures will be re-collected and a non-unique number of environmental conditions will be set during the collection process.

6. The industrial fabric web disease data detection system based on image recognition analysis according to claim 1 is characterized in that: The process of the network disease tracing analysis unit is as follows: Distance information and quantity information are collected, and the distance information and quantity information are respectively the distance deviation of the corresponding web defect position during the image collection process of industrial fabrics of the same specification but different batches, and the proportion of industrial fabrics with the same processing step corresponding to the web defect position of industrial fabrics of the same specification but different batches.

7. The industrial fabric web disease data detection system based on image recognition analysis according to claim 6 is characterized in that: If the distance information exceeds the distance deviation threshold, and the quantity information exceeds the quantity ratio threshold, a process traceability control signal is generated; if the distance information exceeds the distance deviation threshold, and the quantity information does not exceed the quantity ratio threshold, a processing environment impact signal is generated; If the distance information does not exceed the distance deviation threshold, and the quantity information does not exceed the quantity proportion threshold, a material adjustment signal is generated; if the distance information difference does not exceed the distance deviation threshold, and the quantity information exceeds the quantity proportion threshold, an adaptation scheduling signal is generated.

Images