A deformity detection system and a method for detecting fabric deformities and management thereof

An integrated deformity detection and management system using CNN with autoencoders and ambient light conditions addresses the inefficiencies of current systems by accurately identifying and separating defective web material sections in real-time, enhancing efficiency and reducing costs by integrating defect detection and management.

WO2024231880A1PCT designated stage expired Publication Date: 2024-11-14LOHIA CORP LTD

Patent Information

Application Number
PCT/IB2024/054529
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-10
Filing Date
2024-05-09
Publication Date
2024-11-14

AI Technical Summary

Technical Problem

Current systems for detecting fabric deformities in web materials are inefficient, requiring manual inspection, high computational power, and separate defect detection and management systems, leading to increased costs and machine downtime, while failing to accurately identify and separate defective sections in real-time without stopping the web processing line.

Method used

An integrated deformity detection and management system using a single device with a camera module, processing unit, and cloud interface, employing Convolutional Neural Networks (CNN) with autoencoders, capable of operating in ambient light conditions, which captures images, analyzes defects, and communicates their coordinates to the web-processing line controller for automatic separation, reducing the need for additional lighting and computational power.

Benefits of technology

The system effectively detects and separates defective web material sections in real-time, reducing machine downtime and material wastage, improving efficiency and accuracy while lowering costs by integrating defect detection and management, and allowing for online tuning of the algorithm in the cloud.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure IB2024054529_14112024_PF_FP_ABST
    Figure IB2024054529_14112024_PF_FP_ABST
Patent Text Reader

Abstract

The present invention relates to a deformity detection system (4) and a method for detecting the deformities (2) in online web materials and managing them on the web-processing lines Such web processing line along with deformity detection system may comprise of data transmission (5) to cloud (3) which may include intranet, wifi, or can be any means of online data transfer and a processing line's control unit (10) includes human machine interface HMI, display screens, control boxes. The system (4) has system has a camera module interface (7) for capturing images; a processing unit (8) to detect deformities; an interface for cloud (3) for storing the deformities data; and an external unit interface (13) to accommodate external GPU accelerators for enhanced image processing speed and to facilitate seamless information exchange between said camera module interface (7), said processing unit (8) and said interface for cloud (3).
Need to check novelty before this filing date? Find Prior Art

Description

[0001] A DEFORMITY DETECTION SYSTEM AND A METHOD FOR

[0002] DETECTING FABRIC DEFORMITIES AND MANAGEMENT THEREOF

[0003] Field of Invention

[0004] The present invention is relates to a deformity detection system and a method for detecting the deformity in the fabric in an online web materials and managing them on the web-processing lines. The invented device identifies the web's defective portion and stops the machine for the respective defect. Then, the operator can either reject or process the fabric in the downstream line.

[0005] Background of Invention

[0006] Today, many powdery or flowable products such as sand, fertilizers, and plastic granules are filled, stored, and transported in an industrial container of flexible fabric. On the other hand, FIBCs are often made of thick woven strands of oriented polypropylene, either coated or uncoated. Therefore, bags made of web material must meet the quality requirements for the respective purpose of use.

[0007] FIBCs, or “Bulk Bags,” are widely acknowledged and approved as the most economical type of packing. These bags are the best packing choices for storing and shipping up to 5000 pounds of dry flowable goods. Furthermore, such bags meet strict requirements in terms of occupational safety as well as provisions regarding the resistance of the material used against weather influences.

[0008] The web material may have different defects along its lengths, including weaving, holes or surface defects, uneven coating edges, printing defects, or other similar defects. Some defects may be cosmetic; however, some others may compromise the integrity and strength of the web material or affect its other desirable properties, such as leak-proofness, ultimately rendering it unsuitable for the purpose for which it is designed. Mainline processing of the defective web results in the material jam in the main web-processing line, increasing machine downtime and material wastage and reducing machine efficiency. Sometimes, these jams may also result in wear and tear on the mechanical parts due to sudden stops.

[0009] Original equipment manufacturers of packaging bags made of web material are increasingly facing the problem of automated filling plants detecting flawed bags randomly and returning the lot to the supplier. In addition, the conventional defect measurement systems present in the market are costing a fortune to OEMs and forcing them to invest hugely or use additional manpower to check the defect manually, which is time-consuming and inefficient enough to produce without any manual or automatic system to save additional cost. Some available systems may cost approximately the same as machine cost.

[0010] WO 2005 / 085813 Al discloses a device for monitoring a moving web of material at a textile machine, wherein a line sensor is arranged for scanning the web of material along its width. A sensor signal of the line sensor, for example, a CCD— optoelectrical line sensor, is then processed by an electronic circuit into a configurable starting signal that triggers an action. Such a starting signal is suited to trigger, for example, the drive's shutdown for the web of material or the initiation of an optical or acoustic alarm.

[0011] From DE 43 12 452 Al, a method for optically determining quality-defining parameters of textile surfaces using imaging sensors is known. Therefore, the textile surface is detected section-wise by a line camera, and the digital image is subsequently analysed regarding homogeneity, orientation, and number of the meshes and trebles or weft and warp threads, respectively, of the periodically structured web of material or knitted goods.

[0012] DE 33 04 817 Al discloses a method for automatically detecting flaws in textile web materials using image filtration to increase the contrast between the global texture considered normal for the textile to be examined and the local deviations deemed flawed. Therefore, at least two spatial filters for image transformation are used that are adjusted to the detection of straight-lined contour elements. Error signals obtained via digital image filtration are thus stored and analysed by a downstream processor. One of the conventional approaches of individual bag producers is to detect defects in the material already during the production stage of a plastic woven textile, like printing or coating and to label these defects using coloured marks or print marks as given in W02009121841, which have not been entirely successful due to several other factors. Firstly, defects are marked at the printing or coating stage before the conversion machine. Then, special units are installed on the conversion machine to detect the marking / labels posted in the previous operation and the reject defective web accordingly. However, this system involves modification or changes in at least two different machines. Further, the rejected web is a lot of waste, and the system needs to be more reliable.

[0013] A coloured marking of a defect, for example, at a flat web of material, which in the subsequent process steps of the bag production is formed or closed, respectively, into a web material tube, may thus later on be situated on the inside of this web material tube and will thus not be detected in a down-stream optical quality control. The same is effective for coloured marks, which, if the plastic webs in the course are printed later on, will not be detectable anymore. Furthermore, depending on the customer's requirements, it is important to distinguish which defect sizes are to be eliminated. Large-scale or extended defects occurring along several metres of a section of the plastic woven web material are more prone to be detected by the operating personnel than small-scale defects.

[0014] According to US patent application US20120133763, an inspection camera is provided for taking images of the plastic woven web material, and an analysis unit, which detects irregularities in the weaving pattern of the plastic woven web material on the basis of the images taken by the inspection camera and which, when detecting irregularities in the weaving pattern indicating a defected section of the plastic woven web material, triggers an error signal, which actuates, for example, a flashing light or a horn, or which can be fed to a higher-level machine controller. An inspection system is further provided for monitoring defects of a plastic woven web material and variants of a method for labelling or monitoring, respectively, defects of a plastic woven web material.

[0015] A drawback of this approach is that labelling defective web material areas on a high-speed processing unit is a simple operation but further processing of the areas of web material identified as defective, particularly separating them from the defect-free sections, is manual - which will involve stoppage of the advancing web. In addition, identifying the defective section is also important for the smooth processing of web material in the web-processing line. For this, knowledge of coordinate in the machine controller is missing, which causes the said stoppage where the operator or user has to see where the defect has occurred and accordingly remove it manually.

[0016] The prior art does not yet provide a simple and reliable system for detecting a defective section of web material. The sensitivity of this system can be adjusted by tuning physical or other parameters. These adjustments help customers segregate critical defects from acceptable defects for the production of bags before entering the production line.

[0017] Thus, there is a need to provide a system / device and method of web material deformity detection and management, including their detection and separating the defective areas of web material from defect-free web material without stopping the advancing web material such that it overcomes the known disadvantages of the prior art web inspection system. There is also a need to provide a device for managing defects in a web material like plastic woven web material made of monoaxially drawn polymer ribbons, in particular polyolefin ribbons, preferably polypropylene ribbons, which is optionally coated with a thermoplastic material on one or two sides, which are characteristic of defected sections of a typical plastic woven web material.

[0018] A further need for the state of the art is to provide a device that can identify the deformities on the running web material and send them to a rejects collection zone or mark them as defective without allowing them to enter the downstream bag manufacture units. Finally, there is a need to provide a device that meets the needs mentioned earlier by providing a device for detecting defective areas of web material, wherein the coordinates of the defective areas are noted in the control unit. The progress of the defective areas of web material down the conveyor belt is monitored through the control of the web cutting unit; the web material piece / length containing defects is identified and sorted out accordingly. Finally, the cut / marked web containing defects are sent to a rejects collection unit, whereas normal (defect-free) web is sent to downstream processing units.

[0019] In the conventional devices, defect detection and defect management was done in separate systems which adds complexity to the operation. To overcome this drawback, the detecting device and managing device is integrated in single module to optimise the overall system.

[0020] The device proposed in the present invention is designed to accommodate said defects in the producing stage. The device can be extended to measure web conveying rates of up to 200 m / min. The invented device can be mounted on any part of the machine where defects to be identified and significant to reject at that point. According to patent EP3678858, the method for detecting and managing defects uses two separate devices, which adds complexity and cost to the system. In order to improve the system efficiency and reduce complexity, one integrated device with defect detection and management is proposed for this invention. Such an integrated device can send the defect signal to the PLC controller or other third-party systems without requiring additional hardware. This device can work in varying environmental conditions without necessitating additional lightning.

[0021] Many systems use classical image processing or other machine learning algorithms for the quality control system, which requires high computational power at the processing level; however, in the present invention, the model with trained algorithm and tuned parameters are tried in a low processing power system, and it is found working as desired therefore training part of the model is shifted to the cloud to save the processing power. The invention disclosed in the present application was tried using the Convolutional Neural Network (CNN) with autoencoders. The details are given in the invention description. The parameters used to make the model are tuned to achieve higher accuracy.

[0022] Objects of Invention

[0023] The main object of the present invention is to provide a device for detecting the deformity at any location before a transverse cut is made in the material.

[0024] Another object of the present invention is to provide a low-cost edge intelligent device to detect deformities irrespective of the change in web material.

[0025] The other object of the present invention is to use the natural environment condition to find the defects by using the device's capability to reduce the need for additional lighting.

[0026] A further object of the present invention is to alert the user of the individual defect and allow them to separate the defective web material from the line.

[0027] Another object of the present invention is to find the accurate coordinates of the defects on the web material and feed them to the web-processing line controller to decide on handling the defective web materials. Yet another object of the present invention is to remove the need for parameter tuning on the edge device and preserve computational power on the integrated module / edge device. Optional turning can be done on a cloud-based system, and the model can be deployed locally.

[0028] Brief description of Figures:

[0029] Figure 1 shows a plan-view schematic of the application where the Deformity detection device / system of present invention is used Figure 2 shows a side-view schematic of the application where the device of the present invention is used

[0030] Figure 3 shows a method of communication between the various units of deformity detection system in the web-processing line.

[0031] Figure 4 shows the computer images of the defective portions of the web material, with defects detected in the middle region.

[0032] Figure 4a shows the computer image captured during various experiments which are marked with the defect after getting processed by the device.

[0033] Figure 5 shows the schematic of the deformity detection system of the present invention and its various parts which are camera module interface, processing unit, interface for cloud and external interface for processing lines.

[0034] Figure 5A shows the deformity detection system and interface among various units of deformity detection system

[0035] List of Parts: 1. Web material 5. Data Transmission to la- defective piece Cloud

[0036] 2. Defects or Deformities 6. Cutting station

[0037] 3. Interface for cloud 7. Camera Module interface

[0038] 4. Deformity detection 35 8. processing unit system or device 9. Web-processing line 10. Web-processing line 12. Feed rollers control unit 5 12a- Feed zone

[0039] 11. Web material feed roller

[0040] 13 External interface for processing line or external unit interface

[0041] Summary of Invention

[0042] The present invention provides a deformity detection and management system (4) to detect the deformities (2) or to detect the deformities in advancing the web material processing in web processing lines including conversion machines, printing machines, lamination machines etc. Such web processing line along with deformity detection system may comprise of data transmission to cloud 5) which may include intranet, wifi, or can be any means of online data transfer and a processing line’s control unit (10) includes human machine interface HMI , display screens, control boxes , wherein said web processing line having processing stations for web material, said processing stations comprising at least one or more combination of any of following units like a cutting station, a stitching / seaming / sealing station, a stacking station. Depending on the purpose of the web processing line such as bag making, fabric printing, fabric lamination etc. there can be multiple processing stations.

[0043] As a key aspect of the present invention, the deformity detection system (4) which comprises a single on-chip board device having standard camera module to capture and store the images These captured images are analysed / processed by a customized algorithm running by processing unit on an invented system / device for deformity detection. The said algorithm is tuned for specific types of web processing lines in a cloud-based system, and the tuned weights are imported to the invented system / device for web processing. The system / device of the present invention also has the provision of reading the position of the defect on the travelling web material the invented system / device has the capability to communicate with: a. The web processing line's control unit to sort or mark the defect out from the main line and increase the quality of the produced material. b. The stored defects can be transmitted to cloud based system store the defect and tune the algorithm for various custom needs.

[0044] The invented system / device can be connected to any web processing line to communicate the defect position. Based on the application type, the control unit can mark or sort out the defect either online or offline.

[0045] The invented system / device has made it possible to use machine learning algorithms on edge by using normal processor and storage options. The device of the present invention is an edge device which can perform specific task such as defect detection without requiring high computing equipment. The system adapts online changes in fabric print and roll condition, which marks the robustness of deformity detection.

[0046] The cloud system is optional for storing the defects and changing the tuning parameters to tweak the model as per the requirement.

[0047] Detail Description of Invention

[0048] Figure 1 shows a plan-view schematic of the application where the deformity detection device / system of present invention is used. The term system or device or system / device has been used interchangeably to describe the invention.

[0049] The present invention discloses a reliable deformity detection system to detect deformities defects in web material as shown in figures 3 & 5 which has the following components. 1. Camera Modul interface (7) : This unit is responsible for capturing or recording the images of the advancing web and also enables the various standard camera modules with different lens systems. During testing, a camera featuring a Sony IMX 219 PQ CMOS image sensor was utilized, boasting an 8 -megapixel resolution and a 30fps transfer rate. Operating within a temperature range of -20 to 60 degrees Celsius, the camera’s lens measures 14” with dimensions of 23.86 x 25 x 9mm and weighs 3g. Notably, the interface dynamically adjusts light intensity according to environmental conditions, negating the need for supplementary light sources. Further for optimal image capturing, the camera module can be minimum 2-megapixel resolution which can be made by any of the brand or make, one should not restrict to Sony. Further it can have variable shutter speed with variable specification of camera.

[0050] 2. Processing Unit (8): Responsible for executing the deformity detection algorithm, this unit is equipped with a Broadcom BCM2712 processor featuring a 2.4 GHz quad-core 64-bit Arm Cortex A76 CPU architecture.

[0051] 3. Interface for cloud (3): Optionally, this unit facilitates the storage of defect data in the cloud and supports the deployment of algorithm updates wirelessly. Compatibility extends to any cloud provider, with the test configuration utilizing Microsoft’s Azure cloud platform.

[0052] 4. External Unit Interface (13): Designed to accommodate external GPU accelerators for enhanced image processing speed, this interface also facilitates seamless information exchange between processing units, particularly beneficial for integration into high-performance systems.

[0053] The deformity detection system (4) under invention has a microcontroller, camera interface and communication module for interfacing with web-processing lines. The microcontroller is capable of handling the data rate of 160MBPS and can handle signal rate of 75MHz which makes it capable of process captured images at the speed of 200 mpm. By changing the processor and clock of the device, the same device can also be used for speeds up to 600mpm.

[0054] The system / device (4) of the present invention can be used in any web processing system under ambient light condition. The deformity detection system (4) can be placed after the web gets unwounded from the roll. The range of environment light can vary between 200 and 400 lux, preferably 400 lux. The deformity detection system (4) monitors the travelling web material (1) containing defects processed right after being unwounded or before getting into the main processing line. The information on the defect position is sent to the main controller unit of the processing line, which can decide how to handle the defect. Optionally, the image of the deformity can be stored locally or on the cloud, which can be used for further modification of the algorithm. The web processing line can automatically reject the cut piece of web containing defect and prevent it from entering the main line or stop the machine and remove the defective cut piece of web manually.

[0055] The deformity detection system (4) of the invention runs the machine learning algorithm, which detects the web deformity based on the configured or trained models. If necessary, any new web material used for the processing can be trained online. The trained algorithm on the deformity detection system can be modified over the optional cloud-based interface whenever required. The deformity detection system sends the image to the cloud, and the cloud-based deformity management system tests the existing algorithm for its accuracy.

[0056] The algorithm used in this device of invention is based on a neural network-based deep learning algorithm. The described application is tried with Convolutional Neural Networks (CNN) using autoencoders. The images of web material are fed into this model, and it detects the defective area on the image. CNN has multiple layers such as the convolution layer- where several filters are applied, Rectified Linear Unit (ReLU)- used for image normalization; the pooling layer- to resize the image; the fully connected layer used to identify the defect type on the web materials. The main parameter of this network can be adjusted and tuned to set the detection sensitivity. The other models can also be tried to fine-tune the system for separate applications depending on their accuracy.

[0057] Parameters contributing to defect detection accuracy using a tried model may include at least the following.

[0058] 1. Threshold value: In the neural network, the threshold value activates a neuron and sends the signal to another neuron. This value is adjusted during training to optimize the network's performance.

[0059] 2. Confidence threshold value: The confidence threshold value describes the least score that the model considers its prediction to be true. Lowering this value reduces the performance of the model.

[0060] 3. Number of filters: The number of filters represents the number of neurons in the convolutional layer. Different filter values are to be tried to find a suitable value for the network.

[0061] 4. Compression ratio: This parameter was used to squeeze the convolution layer size to reduce the computation and evaluation times at different stages of the CNN network. Increasing the compression ratio will decrease the evaluation time but reduce the model accuracy.

[0062] After a few experiments, it has been concluded that the accurate result of the detection of defects can be achieved when

[0063] 1. Threshold value is in the range of 0.7-1.3 or preferably 0.8-1.0 or more preferably 0.9- 1.0

[0064] 2. Confidence threshold value is in the range of 0.4-60 or preferably 0.8-55 or more preferably 1-50

[0065] 3. Number of filters is in the range of 10-80 or preferably 15-70 or more, preferably 20-65 4. Compression ratio is in the range of 0.5-15 or preferably 0.8-12 or more preferably 0-10

[0066] Finally, the deformity detection system sends the signal to the control unit of the web-processing line to stop the line or use the automatic separation of the fabric as described in patent no EP3678858 so that the detected fabric can be removed or fed to the machine.

[0067] The deformity detection system (4) includes a camera / infrared sensor / thickness sensor or image processing devices that are light sensing devices integrated with an embedded single board processing unit capable of running machine learning algorithms. The deformity management system comprised of a cloud-based system that runs a machine learning algorithm to improve the existing algorithm running on the deformity detection system. The deformity detection system stores the time of defect and calculates and stores the position of the defect (2) along the traverse web, where a cut needs to be made to convert the web material (1) into required processed form. According to the speed of travel of the web material, the deformity detection system calculates the speed and position of the travelling defects (2). According to the type of web material (1), i.e. printed or plain, and the final product requirement, the deformity detection device calculates the position of the transverse cuts to be made in the web material (1). Finally, the deformity detection system stops the machine or automatically mark / sorts out the defect and uses operator input to accept or reject the input.

[0068] The web deformity detection system of the present invention works without additional equipment and can be ported to another place easily based on ambient light conditions.

[0069] The term defect or deformity (2) refers to any unacceptable irregularities in the web material from the perspective of the suitability for the purpose for which the web material is intended. Defects (2) may be uneven thickness variation, holes, uneven pattern, missing warp end, missing weft, crushing of weft, uneven coating edge trim, web joints, one or more web layers, foreign particles or any combination of these defects. The terms web material, web and fabric are used interchangeably in the following description.

[0070] The web material (1) may be in the form of plastic film, non-woven material, paper, woven web material, woven plastic web material, or a composite material made from any of these materials. It may also be coated or laminated with a plain or printed film, or printed, or any combination thereof.

[0071] Web materials - either in flat form or tubular form - are used for many purposes, including making storage bags. Storage bags are typically made using automatic or semi-automatic web-processing lines (9) (which convert web materials into processed goods). These systems may have any or all individual machines, such as a cutting machine, folding / forming machine, stitching machine, printing machine, lamination machine and valve forming machine. The web-processing line (9) may include a bag or fabric-cutting system machine. In particular, the placement of the invention device on any of these conversion systems may be according to the user’s choice. However, it should be placed before doing any processing on the web.

[0072] The bag conversion system that makes products using web materials (1) may have individual stations to perform operations such as making perforations, D cut or C cut, folding operation, stitching operation or stacking unit. The detected defect on the web material can be taken out manually or automatically, depending on the operation type. The defective portion / section of web material is separated from the defect-free section at the cutting position and taken out of the process. This ensures that the process flow moves smoothly without damaging any of the process stations. Figure 3 shows the communication protocol between some of the various units of the interface for cloud (3). Upon detection of any defects (2) or irregularities in any area of the advancing web material (1), a signal is sent by deformity detection system (4) to the control unit (10). The location of the defect (2) at the time of its detection is noted.

[0073] As the web material (1) travels in the web material feed zone (12a) and if it is defect-free, then the web material (1) gets cut into pieces for product-making. The changing location of the defect (2) detected by the deformity detection system (4) on the advancing web material (1) is noted in the memory of the deformity detection system (4) and also sent to the interface for cloud(3), finally leading to the identification of the piece of web material containing the defect, or defective cut pieces (la).

[0074] As seen from Figure 2, web (1) from the web material feed roller (11) are supplied to the conversion systems fitted with the web material deformity detection system (4), where a web material feed (la) roller facilitates controlled entry of the web material (1) into the deformity detection system (4). Additional feed rollers facilitate the movement of the web material (1) between various stations (not assigned numbers or shown) of the web-processing line (9), such as a cutting station (6) or a sealing station, of the web-processing line. The deformity detection system (4) may be provided between any two feed rollers depending on the user’s requirement.

[0075] One of the aspects of the present invention is that a provision is also made for a suitable system for monitoring the speed of the fabric feed roller using an encoder or similar device that communicates with the deformity detection system (4). The feed roller (12) speed determines the speed of web material (1) advancing into the web-processing line (9). This determines the speed of the web material (1) travelling through the various parts of the conversion system. It is important to know the speed of web material (1) so that once the defective parts of web material (1) are detected and identified, the movement of the cutter that separates the defective sections of web material from the defects-free sections of web material (1), can be controlled to make transverse cuts in the web material (1) at desired locations.

[0076] The sensitivity or accuracy of the interface for cloud (3) may be set according to individual customer requirements.

[0077] Once the defective piece is identified, a PLC logic of the web-processing line control unit (10) stops / sort the defect out from the machine to prevent it from getting into production. Next, the operator decides whether the stopped bag should be processed.

[0078] According to another aspect of the present invention, the feeding device mentioned in this application is controlled by a servo-drive unit, which enables PLC to store the position of the defect and resume the process after operator intervention. In addition, the defect detection system does not require additional computational power, which benefits users monetarily as it has edge intelligence, one aspect of the invention.

[0079] Another aspect of the present invention is that the deformity detection system does not require an additional light source to measure the defect. Ambient light with an intensity can varies between 200 and 400 lux, preferably 400 lux. lux is sufficient to measure the defect efficiently.

[0080] The identification of defects (2) using the deformity detection system (4) of the present invention is carried out as follows. The deformity detection system (4) comprises a camera module interface (7) and a processing unit (8) that communicates with the control unit (10) of the web-processing line. The images of the traversing web material (1) are captured by the camera module and stored at processing unit (8), which runs the machine learning algorithm and signals the defects. Ambient light is sufficient to sense the defect using the proposed deformity detection system. The algorithm running the defect detection system can identify the type of defect (2).

[0081] The web material deformity detection system (4) of the present invention is capable of the exact time and / or location of the transverse cut that must be made in the web material to convert it into one or more pieces once a defective length is detected. One key aspect of the invention is to reduce the cost of the defect detection / management device, which can be used to improve the quality aspect of the process engineering in the printing / web-processing line.

[0082] The method of describing the defect detection in the printed pattern is in conjunction with patent no EP3678858.

[0083] According to one aspect of the present invention, a method to detect deformities by deformity detection system (4) and assure the quality of the converted bag from FIBC or another web-processing line where the method comprises the steps as shown in figure 3 & figure 5a: a. using ambient light conditions to capture the defects in the main web material by camera module interface (7); b. measuring the speed of web material within the deformity detection system itself by the processing unit (10); c. registering / storing data for the defects by recording planar co-ordinates in the detection device by interface for cloud (3); d. enhancing image processing speed and to facilitate seamless information exchange between said camera module interface (7), said processing unit (8) and interface for cloud (3) by external unit interface (13), e. controlling the defect detection algorithm using a deformity management system residing on the cloud whenever necessary; f. stopping or automatically rejecting the web piece by communicating to the controller of the web-processing line control unit (10) by data transmission to cloud (5) ; g. allowing users to decide whether to reject or accept the defective piece in a machine like FCM / BCS or other converting machines.

[0084] In one of the aspects, a deformity detection system (4) detects and manages the defects in advancing web material provided on a web-processing line; said web processing lines can have multiple stations depending on type of web processing where web traves between several feed rollers that facilitate the movement of said web material between two stations.

[0085] In another aspect of the present invention, the deformity detection system does not require an additional light source to detect the defects on the web material. The ambient conditions are enough to detect the defect when the device is operated in the range of environment light which can vary between 200 and 400 lux, preferably 400 lux.

[0086] Another aspect of the device of the present invention is that said deformity detection system is independent of the scanning width, so multiple detection devices can be placed over a width to cover the wider range of the main material width.

[0087] Figures 4 and 4a show the digital images of the defective portions of web material as scanned by the device under ambient light conditions for web material of white background. As seen in Figure 4, the inventive device has detected defects in the middle region. As also seen in the figure 4a, the device of the present invention has detected a defect and created a mark on the defect.

Claims

Claims:

1. A deformity detection system (4) for detecting fabric deformities and management thereof in online web materials and managing them on the webprocessing lines, characterised in that said system has the following components: a. camera module interface (7) for capturing images in ambient light; b. a processing unit (8) to detect deformities c. an interface for cloud (3) for storing the deformities data in cloud; d. an external unit interface (13) to accommodate external GPU accelerators for enhanced image processing speed and to facilitate seamless information exchange between said camera module interface (7), said processing unit (8) and said interface for cloud (3), wherein said camera module interface (7), said processing unit (8), said interface for cloud, and said external unit interface are integrated with each other.

2. The deformity detection system (4) as claimed in claim 1 wherein said system is placed after said web gets unwound from a roll on which it has been wound.

3. The deformity detection system (4) as claimed in claim 2, wherein said system (4) preferably operates in absence of external light source other than said ambient light.

4. The deformity detection system (4) as claimed in claims 1-3, wherein said ambient light is between 200 and 400 lux, preferably 400 lux.

5. The deformity detection system (4) as claimed in claim 1 wherein the said processing unit (8) has the following parameters: an adjustable threshold value in the range of 0.7-1.3 or preferably 0.8- 1.0; a confidence threshold value in the range of 0.4-60 or preferably 0.8-55; number of filters that represents the number of neurons in the convolutional layer is in the range of 10-80 or preferably 15-70 ); and a compression ratio that determines the size of said convolutional layer is in the range of 0.5-15 or preferably 0.8-12.

6. The deformity detection system (4) as claimed in claims 1 to5, wherein said system (4) is independent of the scanning width, so multiple detection devices can be placed over a width to cover the wider range of the main material width.

7. A method for detecting the deformities and management thereof by deformity detection system (4) comprising the following steps : a. using ambient light conditions to capture the defects in the main web material by camera module interface (7); b. measuring the speed of web material within the deformity detection system itself by the processing unit (10); c. registering / storing data for the defects by recording planar co-ordinates in the detection device by interface for cloud (3); d. enhancing image processing speed and to facilitate seamless information exchange between said camera module interface (7), said processing unit (8) and interface for cloud (3) by external unit interface (13), e. controlling the defect detection algorithm using a deformity management system residing on the cloud whenever necessary; f. stopping or automatically rejecting the web piece by communicating to the controller of the web-processing line control unit (10) by data transmission to cloud (5) ; g. allowing users to decide whether to reject or accept the defective piece in a machine like FCM / BCS or other converting machines.

Citation Information

Patent Citations

  • High-definition camera cloth defect detection method based on convolutional neural network

    CN111047655A

  • Fabric defect detection method and system based on over-complete convolutional neural network

    CN114494236A

Cited By

  • Textile fabric tearing strength detection method and system

    CN120352277A