Field of the Invention

The invention relates to an arti ficial vision system developed to detect and classi fy fabric de fects in circular knitting machines in real time during production . The system comprises equipment such as cameras , lighting units , an encoder and a computer, and can be adapted for circular knitting machines . The system of the invention comprises an image processing software for detecting defects on knitted fabric . In addition, an arti ficial intelligence algorithm has been developed to automatically classi fy the detected defects by their type .

Known State of the Art

In knitted fabric production, no intervention is made to fabric defects during production, and fabric defects are controlled by quality control personnel on an i lluminated surface after production .

In the art , the quality control process on produced fabrics is performed with human eyes on an illuminated table . The fabric is wrapped by passing it over a surface illuminated from the bottom and top . An experienced quality control personnel stands at the front side of this surface , and detects defective areas by following the fabric with its eyes during the winding of the fabric . During this process , the quality control personnel needs to thoroughly scan an area of about 2 meters wide . Therefore , manual operation is both exhausting and time-consuming . For example , an inspection of fabric defects is carried out by an average of 30 workers in an enterprise that produces approximately 100 , 000 meters per day . An experienced quality control personnel can detect only 60-70% of the defects , and control a fabric up to 2 meters wide . While the quality control process performed as such is not obj ective , the evaluation of defects cannot be performed statistically . Since the quality of the fabric varies according to the evaluator, there is a di f ference of opinion and lack of agreement between the parties . Additionally, this process requires signi ficant machine investment costs and labor costs . Also , since fabric defects cannot be detected with high ef ficiency, production cost increases and undetected defects may cause loss of prestige .

Arti ficial vision systems are available in the art to detect fabric defects in real time on a circular knitting machine .

Examples of the state of the art include a non-patent document titled " FABRIC DEFECT DETECTION SYSTEM BASED ON IMAGE PROCESS ING FOR CIRCULAR KNITTING MACHINES (Kazim Hanbay) " .

This document relates to a fabric defect detection system based on image processing developed for circular knitting machines . The system of said document detects and classi fies fabric defects that may occur during fabric production in knitting machines in real time . The system uses a camera, a lens and a lighting unit for image acquisition process , and an encoder . In the system, the inner surface of the fabric is scanned by the camera placed in interior of the circular knitting machine , and the system i s stopped when a default is detected . Di f ferent fabric structures can be produced in circular knitting machines , and there might be defects on both the front and back surfaces of some fabrics . Since the mentioned system can scan only single side of the fabric, it can not detected the defects on the opposite side due to the position of the camera . In addition, it is understood that said system does not provide data regarding position information of the defects . Another example of the known state of the art is the patent document WO20079493A1 , wherein said document relates to a defect detection system developed to detect fabric defects in a circular knitting machine in real time . The system of the document allows detection of a defect , saving information and images of detected defect for reporting or analysis , sending an alert i f a defect is detected and stopping the system . Said system has hardware parts such as a camera, a lighting unit located on the exterior of the knitting machine . The system inspect a single surface of the fabric due to the position of the camera and the defects occur on the other surface are not to be detected . In addition, the system of the document requires many modi fications in order to be adapted for a knitting machine .

The patent document TR2020/ 08719 relates to a circular knitting machine and an automatic control system developed for controlling defects on a fabric in real time, reducing the margin of error, intervening in the defects and reporting data on the defects . In said system, there is at least one camera that can be located on di f ferent regions of the circular knitting machine . However, said automatic control system only performs defect detection . Therefore, it is understood that the defects are not classi fied and their positions are not determined . In addition, an alarm unit that gives warning in case of a defect detection and a unit that can synchroni ze the rotation speed of the circular knitting machine with the arti ficial vision system is not mentioned . In addition, one of the most important parameters in image acquisition and image processing operations is lighting conditions . However, the use of a light source is not mentioned in said document . In order to eliminate the disadvantages of the above documents, an artificial vision system has been developed that has an assembly that follows both the inner and outer surface of fabric, saves position data of the defects detected on the fabric surfaces and performs real-time inspection and classification of fabric defects, which can be easily adapted for circular knitting machines.

Detailed Description of the Invention

The invention relates to an artificial vision system that is particularly adapted for circular knitting machines (7) and allows automatic detection and classification of surface defects of knitted fabrics.

The invention is an artificial vision system for knitting machines (7) , comprising in general

- at least two cameras (9) such that one of the camera (9) being located to acquire an image frame of one surface of a fabric sample (8) and the other camera (9) being placed to acquire an image frame of other surface of the fabric sample ( 8 ) ,

- at least two lighting units (6) providing lighting to an area where the image is captured by the cameras (9) ,

- at least two encoders (5) for synchronizing image acquisition speed of the cameras (9) with movement of the fabric surface, and

- a mechanical component being mounted on the knitting machine (7) so as to adjust its distance to the knitting machine (7) , carrying at least one of the cameras (9) , at least one of the lighting units (6) and at least one of the encoders (5) ,

- a computer (10) for transferring, displaying and saving images from the cameras (9) , and

- an image processing software being contained in said computer (10) , having an algorithm for processing images, detecting and classifying defects, identifying position data and stopping the knitting machine (7) when the defect is detected.

An embodiment of the invention comprises an alarm unit that gives warning when a defect is detected.

An embodiment of the invention comprises a CCD camera system.

An embodiment of the invention comprises a connector providing connection between the cameras (9) and the computer (10) . The connection is preferably provided by an interface (GigE) or USB-3.0 .

In an embodiment of the invention, the camera (9) is contained in a carrying apparatus/cabinet (1) . The lighting unit (6) is located at the front of said cabinet (1) .

In an embodiment of the invention, the camera (9) and the lighting unit (6) are in the same position.

In an embodiment of the invention, the lighting unit (6) comprises a LED light.

In an embodiment of the invention, the encoder (5) comprises an encoder arm (3) which allows its position to be adjusted with respect to the fabric surface, and an encoder wheel (4) which contacts the fabric surface.

In an embodiment of the invention, the mechanical component comprises mechanical arms (2) .

An embodiment of the invention is an artificial vision system that can be adapted for circular knitting machines (7) , which comprises an image processing and defect classification software .

An embodiment of the invention consists of a CCD camera system, a lighting unit (6) , a camera (9) , a carrying apparatus/cabinet (1) , an encoder (5) for synchronizing the image acquisition process with movement of fabric surface, and a computer (10) .

The artificial vision system of an embodiment of the invention is suitable for the structure and operation system of the knitting machine (7) . The camera (9) and the lighting unit (6) needs to be adapted for the circular knitting machine (7) such that they can see the fabric surface properly and receive high quality images. Therefore, the artificial vision system includes an apparatus or mechanical component that can be easily adapted for the circular knitting machine (7) .

An embodiment of the invention comprises a mechanical component necessary for adapting the camera (9) and the lighting units (6) to the body of the circular knitting machine (7) . After the materials to be used for the artificial vision system of the invention are provided, mounting thereof and system installation takes place. The success rate of the image processing algorithm depends on the algorithm developed as well as the quality of the processed image. The hardware properties of the image acquisition system and the characteristics of the light source are the most important factors determining the quality of the captured image. Therefore, the invention comprises a suitable lighting unit (6) that will provide uniform lighting using different light sources for the image acquisition system to be installed. A suitable position for the lighting unit (6) is determined on the artificial vision system. A position for the camera (9) where the best image can be taken and which will not allow shadings is determined. In addition, the system is related to the mounting of the cameras (9) on the knitting machine (7) so as to scan the front and back sides of the fabric surface, wherein the positions of the cameras (9) are determined so as not to prevent the operation of the knitting machine (7) and not to affect the production performance. For this purpose, an appropriate mechanical component is developed to carry the camera (9) and to connect it to the machine body. The images taken by the camera (9) are transferred to the computer (10) for processing in the algorithms developed. This process is carried out via GigE or USB-3 connectors.

The mechanical component comprising the mechanical arms (2) carrying the camera (9) and the encoder (5) , which is an important part of the artificial vision system of an embodiment of the invention is shown in Figures 1 and 2, wherein the mechanical arms (2) allows the camera (9) to be located with respect to the fabric surface and lens settings. The camera (9) is placed in a cabinet (1) . The LED lighting unit (6) placed at the front of the cabinet (1) provides uniform lighting to the fabric surface. The encoder arm (3) allows the position of the encoder (5) to be adjusted with respect to the fabric surface, and the encoder wheel (4) to contact the fabric surface. Accordingly, the speed of fabric surface and the image acquisition speed of the camera (9) are synchronized.

The circular knitting machine (7) comprising the artificial vision system of the invention is also within the scope of the invention. Figure 3 shows a circular knitting machine (7) comprising the artificial vision system of the invention. Accordingly, the artificial vision system controlling the images of the knitted fabrics advancing on the circular knitting machine (7) comprises a camera (9) contained in a carrying apparatus/cabinet (1) receiving the images of the fabric produced in the knitting machine (7) , a lighting unit (6) providing uniform lighting to the environment where the image is taken, a mechanical arm (2) for adjusting the distance of the camera (9) and the lighting unit (6) to the circular knitting machine (7) , an image processing software with an algorithm for processing the images, an encoder (5) triggering the speed of the machine to the camera (9) in order to take images of the circular knitting machine (7) by the camera (9) , and a computer (10) for displaying the images taken .

A detailed embodiment of the invention comprises a camera (9) located in a cabinet (1) , a lighting unit (6) located at the front of the cabinet (1) , an encoder (5) connected to the cabinet (1) by the encoder arm (3) , and an encoder wheel (4) contacting the fabric sample (8) , a mechanical component for mounting said cabinet (1) to the circular knitting machine (7) . In the said configuration, there are two mechanical components, one located in the interior and the other on the exterior of the circular knitting machine (7) . Thus, both the inner surface and the outer surface of the fabric sample (8) are followed by the camera (9) . The images taken by the cameras (9) are sent to the computer (10) .

The system of the invention can detect defects on knitted fabrics by means of the image processing software contained therein, and the classify them with defect classification algorithm. This is an important part of the system.

Thanks to the computer (10) in the system of the invention, the user can see the images of the knitted fabrics in the artificial vision system, which are taken by the camera (9) under the light source. In an embodiment of the invention, the camera (9) is located inside the carrying apparatus/cabinet (1) . Thus, it takes the image of the fabric samples (8) from the surface that is completely perpendicular to the fabric produced in the circular knitting machine (7) .

In an embodiment of the invention, the lighting unit (6) is located just above the cabinet (1) that is the carrying apparatus of the camera (9) , does not affect the field of view of the camera (9) and is placed in the same position as the camera ( 9 ) .

In an embodiment of the invention, the artificial vision system is located on the interior and exterior of the circular knitting machine (7) for detecting defects on knitted fabric in real time during the production phase.

The operation method of an artificial vision system for the knitting machine (7) of the invention is also within the scope of protection. This method comprises the process steps of

- taking images of the inner and outer surface of the fabric sample (8) produced in the knitting machine (7) by the cameras ( 9 ) ,

- illuminating the image area of the cameras (9) ,

- synchronizing the image acquisition speed of the cameras (9) with the surface movement of the fabric sample (8) ,

- transferring the acquired image to the computer (10) ,

- processing the transferred images,

- detecting, classifying defects and determining position data thereof,

- stopping the knitting machine (7) and giving a warning when a defect is detected.

With this method, the knitted fabric formed on the circular knitting machine (7) is also inspected during its production by the arti ficial vision system . Since the machine rotates continuously due to the position where the arti ficial vision system is placed, it is ensured that the entire region of the fabric enters the field of view of the camera ( 9 ) and the system is automatically triggered and the surface image of the fabric sample ( 8 ) is taken . Analysis of the taken image by the developed image processing and classi fication algorithm, and determination of whether it contains a defective area, identi fication of the fabric images determined to be defective as a result of the analysis and a warning on the machine are provided .

The predetermined position in which the image of the fabric sample ( 8 ) will be taken in the arti ficial vision system of the invention is a position in the interior of the circular knitting machine ( 7 ) . This position varies according to where the camera ( 9 ) is located in the arti ficial vision system .

In an embodiment of the invention, a defect control is performed for both sides of the knitted fabric by a second camera ( 9 ) to be placed on the exterior of the circular knitting machine ( 7 ) . Thus , it is ensured that both surfaces of the fabrics with di f ferent structural properties and knitting patterns are scanned, and a defect inspection is performed with high precision .

In an embodiment of the invention, in addition to archiving images of defects detected, position data is also saved as fabric quantity .

The invention allows that image frames are taken in high quality, defects are detected with higher success , their positions are determined properly, and classi fication is performed with a high accuracy rate . For this purpose , the image acquisition speed and the movement speed of the fabric surface (the speed of the circular knitting machine) are synchronized by using the encoder (5) .

The artificial vision system of the invention can be mounted on circular knitting machines (7) without requiring further modification .

With the invention, fabric defects are detected in real time during the knitting process. Production is stopped when a continuous or periodic defect is detected. Thus, production of defective fabrics will be reduced. After fabric production, products can be classified directly according to the quality class without the need for a defect inspection. Therefore, the production speed will be increased. The number of workers working in the quality control department will be significantly decreased. Thus, the cost of production will be reduced. The images and positions of the defects will be saved automatically. A significant archive will be obtained related to the defects encountered in the production. Defect detection is performed objectively.

As a result of the effective use of the system of the application, it is aimed to reduce the production costs by at least 1% in the short term, and to increase the efficiency by 4%. After a determined trial period, it is aimed to reconstruct the fabric defect control unit gradually, and minimize the unit by downsizing. It is envisioned that the system will provide significant savings from labor costs in the long term. Additionally, automatic and precise inspection of fabric defects will significantly reduce the reclamation costs of the company arising from fabric defects, and will provide the company with a remarkable prestige.

The camera (9) , the LED lighting unit (6) and the encoder (5) will be placed suitably with the mechanical arm (2) apparatus designed to be in interior and exterior of the circular knitting machine ( 7 ) . The positions and classes of the defective regions will be determined by processing the fabric images taken by the arti ficial vision system in the developed algorithms . Production will be stopped when defects of a determined number or si ze are detected . Thus , production of defective fabrics will be minimi zed . There will be no need for a further quality control process after production . Thus , signi ficant savings on quality control labor and quality control machine investments will be achieved .

Description of the Figures

Figure-1 : Perspective view of the camera, the lighting unit , the encoder, and the mechanical component carrying them of the inventive system

Figure-2 : Side view of the camera, the lighting unit , the encoder, and the mechanical component carrying them of the inventive system

Figure-3 : A view of a circular knitting machine mounted with the arti ficial vision system of the invention

Description of Reference Numbers in Figures

1 . Cabinet

2 . Mechanical arm

3. Encoder arm

4 . Encoder wheel

5 . Encoder

6. Lighting unit

7 . Knitting machine

8 . Fabric sample

9. Camera

10 . Computer