TWI731565B - Rapid defect detection integration system for flake materials - Google Patents

Abstract

A rapid defect detection integration system for flake materials, which is provided for testing a fabric and includes an optical detection module for photographing the fabric, a system host connected to the optical detection module and provided with a data receiving unit for receiving photos taken by the optical detection module, a defect discrimination server for picking up a defective photo and a positioning unit for binding the defective photo to the corresponding coordinate position of the defect in the flake material, a storage host connected to the system host and provided for receiving the defective photos and coordinate positions binding to defective photos, an artificial intelligence identification unit connected to the storage host and has a training module for receiving various sample photos and a defect classification module for receiving the defective photos.

Description

Sheet material quick defect detection integration system and use method thereof

The present invention relates to a technique of optical inspection, and particularly relates to a technique of integrating optical inspection and deep learning.

The so-called roll-to-roll (Roll-to-Roll) is a high-efficiency and continuous production method that specializes in processing flexible films, soft boards or fabrics. Such films, soft boards or fabrics are from the original cylindrical shape. After the roll of material is rolled out, the film, soft board or fabric is tested, and then rolled into a cylindrical shape.

Due to the roll-to-roll (Roll-to-Roll) method, the object to be inspected is moved quickly. However, the object to be inspected, such as cloth, will vibrate due to its own weight during the rapid movement, which in turn causes Optical inspection equipment is difficult to accurately detect. Even if the optical inspection equipment has picked up the flaws on the fabric, the vibration of the fabric will deform the flaws detected by the optical inspection equipment, which will affect the accuracy of the inspection. For this reason, there is an urgent need for a A detection system that can accurately inspect the tiny flaws on the fabric, and can further determine the type of flaws on the fabric.

The present invention provides a system for quickly detecting defects in sheet materials, the main purpose of which is to be able to accurately inspect the tiny defects on the fabric and to further determine the types of defects on the fabric.

In order to achieve the foregoing objectives, the system for rapid detection of defects in sheet materials of the present invention is used to detect a cloth, and the integrated system for rapid detection of defects in sheet materials includes:

An optical detection module for shooting the cloth;

A system host, which is connected to the optical inspection module for information, the system host has a data receiving unit, a positioning unit, and a defect discrimination server for information connection, and the data receiving unit is used to receive photos taken by the optical inspection module , The defect judgment server is used to capture a defect photo, and the positioning unit is used to bind each of the defect photo and the corresponding coordinate position of the defect in the sheet material;

A storage host connected to the system information of the host, and the storage host is used to receive the defective photo and the coordinate position bound to the defective photo;

An artificial intelligence recognition unit is connected to the storage host for information. The artificial intelligence recognition unit has a training module and a defect classification module connected to the information. The training module is used to receive various template photos, and the training module uses the Some sample photos form a judgment logic. The defect classification module is used to receive the defective photo. The defect classification module classifies according to the judgment logic, and then binds each defective photo to its defect classification, and then sends it back to the storage Host.

The present invention provides a method for using an integrated system for rapid detection of defects in sheet materials, which includes:

A step of capturing photos for continuous photography of sheet materials to obtain multiple photos;

A judgment step for judging whether each photo is a defective photo, and extracting the defective photo;

A position obtaining step, obtaining the relative coordinates of the defective photo on the sheet material, and the coordinates are bound to the defective photo;

A storage step, record the defective photos and bound coordinates;

A training step is to classify through plural model photos to form a discriminant logic;

In a classification step, the defective photos are classified by the aforementioned judgment logic, and the photos are stored according to the aforementioned classification.

From the foregoing, the present invention mainly uses the system host to initially determine the defective photos, and then stores the defective photos in the storage host, and then uses the artificial intelligence identification unit to determine the judgment logic obtained after training on the defective photos stored in the storage host. Accurate classification. In this way, the artificial intelligence recognition unit only needs to recognize a small number of flawed photos, instead of recognizing all the photos taken by the optical inspection module, so as to improve the performance of the artificial intelligence recognition unit. In addition, the artificial intelligence recognition unit can also be used Provide accurate analysis, even if the defects detected by the optical inspection module are deformed, the types and sizes of the defects can be accurately identified through the artificial intelligence recognition unit.

The present invention provides a sheet-shaped material rapid detection defect integration system, please refer to Figures 1 to 6, 11 to 12, and the sheet-shaped material rapid detection defect integration system is used to detect a cloth F, and the sheet-shaped material rapid detection Defect integration system, including:

An optical detection module 10 for photographing the cloth;

Preferably, the optical detection module 10 has at least one photographing group 10A, the photographing group 10A has a light source 10B and a photographing lens 10C, the light source 10B is used to illuminate the cloth F, and the photographing lens 10C is provided for the cloth F F for continuous photography;

A system host 20 is connected to the optical detection module 10 for information. The aforementioned information connection can be achieved through wireless or wired means. The system host 20 has an information connection: a data receiving unit 21, a positioning unit 22, and a defect determination The server 23, the data receiving unit 21 for receiving photos taken by the first camera lens 112 and the second camera lens 122, the defect judging server 23 retrieves the defective photos based on the appearance of the defects in the photos, the aforementioned defective photos It means that the cloth in the photo has a defect, and the positioning unit 22 is used to bind each photo of the defect to the corresponding coordinate position (X, Y) of the defect located in the sheet material, and then record the position of each defect in the cloth;

A storage host 30 is connected to the host system 20 for information. The aforementioned information connection can be achieved through wireless or wired means. The storage host 30 is used to receive the defective photo and the coordinate position bound to the defective photo;

An artificial intelligence recognition unit 40 is connected to the storage host 30 for information. The aforementioned information connection can be achieved through wireless or wired means. The artificial intelligence recognition unit 40 has an information connection training module 41 and a defect classification module 42. The training module 41 is used to receive various model photos, each of the model photos can be photos of various defects, such as photos of missing edges of fabric, photos of fabric hooks, photos of fabrics with indentations, photos of fabrics with holes, The training module 41 forms a judgment logic through the sample photos, the defect classification module 42 is used to receive the defect photos of the storage host 30, the defect classification module 42 classifies according to the judgment logic, and then classifies each defect The photo is bound to its defect classification, and then sent back to the storage host 30.

In a preferred embodiment, the fabric F has a first face F1 and a second face F2 opposite to each other. The first face F1 faces a first side K1, and the second face F2 faces a second side K2. The number of the photography group 10A is plural, and the photography group 10A can be divided into at least one first photography group 11 and at least one second photography group 12. The light source 10B of the first photography group 11 is a first light source 111, and the first The photographic lens 10C of the photographic group 11 is a first photographic lens 112, the light source 10B of the second photographic group 12 is a second light source 121, and the photographic lens 10C of the second photographic group 12 is a second photographic lens 122. The first light source 111 is arranged on the first side K1 of the cloth F, the first light source 111 is used to illuminate the first surface F1 of the cloth F to form a front light source, and the second light source 121 is arranged on the cloth F The second side K2, the second light source 121 is used to illuminate the second surface F2 of the cloth F to form a back light source, and the first camera lens 112 and the second camera lens 122 are both arranged on the first side of the cloth K1, the first photographic lens 112 and the second photographic lens 122 continuously photograph the first surface F1 of the fabric F. The first photographic lens 112 photographs through the front light source, and the second photographic lens 122 photographs through the back Light source for shooting;

In a preferred embodiment, the storage host 30 or the system host device 20 has a tabulation unit 50, and each tabulation unit 50 binds according to the defect photo and the defect coordinates and defect classification bound to it to form a Record form T.

In another preferred embodiment, the tabulation unit 50 binds the defect photo and the defect number, pattern, defect coordinates, and defect classification bound to the defect photo to form a record form T.

In a preferred embodiment, the storage host 30 or the system host 20 has an imaging unit 60. The imaging unit 60 forms a defect schematic P according to the bound defect coordinates and defect classification. A mark P1 is displayed on the position corresponding to the defect coordinate.

Preferably, the mark P1 forms different patterns according to the classification of the defect for visual indication. For example, the defect at coordinates (765,1455) is a hole, and the coordinates of coordinates (6541,888) are missing edges. Then, the imaging unit 60 marks a circle at the position corresponding to the coordinates (765, 1455) on the icon, and a triangle at the position corresponding to the coordinates (6541, 888) on the icon.

Preferably, the system host 20 further has a preliminary classification module 24 for preliminary classification of the defective photos, and then returns the classification results to the storage host 30, and if the defects are classified When the classification of the module 42 and the preliminary classification module 24 have different classifications for the same defective photo, the classification result of the preliminary classification module 24 is replaced by the classification result of the preliminary classification module 24, and the preliminary classification module 24 is adjusted. The algorithm enables the classification accuracy of the preliminary classification module 24 to be closer to that of the defect classification module 42.

A method of using the integrated system for rapid defect detection of sheet materials, please refer to Figures 7 to 10, including:

A step S1 of capturing photos: continuous photography of the sheet material to obtain multiple photos;

A judgment step S2 for judging whether each photo is a defective photo, and extracting the defective photo. The aforementioned judgment mechanism can be recognized through the appearance of the defect in the photo, but is not limited to this;

A position obtaining step S3, obtaining the relative coordinates of the defective photo on the sheet material, and the coordinates are bound to the defective photo;

A storage step S4, record the defective photos and the coordinates where the defects occurred, and back them up accordingly to avoid loss;

In a training step S5, classification and definition are performed through plural template photos to form a discrimination logic. Furthermore, the accuracy of the discrimination logic can be improved by increasing the number of template photos;

Preferably, the aforementioned classification definition can be drawn up with the assistance of personnel;

In a classification step S6, the defective photos are classified by the aforementioned judgment logic, and the photos are stored according to the aforementioned classification.

Preferably, it further includes a tabulation step S7 to create a record sheet T according to the bound defect photos, coordinates, and defect classification.

Preferably, a drawing step S8 is further included to form a defect schematic P according to the bound defect photos, coordinates, and defect classification, and a mark P1 is displayed on the defect schematic P at a position corresponding to the defect coordinates.

Preferably, the mark P1 forms different patterns according to the classification of the flaws to visualize the schematic diagram. For example, the flaw at the coordinates (765,1455) is a hole, and the coordinates (6541,888) is a missing edge. Then the imaging unit 60 marks a circle at the position corresponding to the coordinates (765, 1455) on the icon, and a triangle at the position corresponding to the coordinates (6541, 888) on the icon.

From the foregoing, the present invention mainly uses the system host 20 to initially determine the defective photo, and then stores the defective photo in the storage host 30, and then stores the judgment logic pair obtained by the artificial intelligence identification unit 40 in the storage host 30. In this way, after the preliminary screening of the system host 20, the artificial intelligence recognition unit 40 only needs to recognize a small number of defective photos, and the recognition amount can be reduced by as much as 99%, without the need for the optical inspection module 10. All photos are recognized to improve the performance of the artificial intelligence recognition unit 40. In addition, the artificial intelligence recognition unit 40 can also provide accurate analysis. Even the defect and deformation detected by the optical inspection module 10 can be accurately analyzed by the artificial intelligence recognition unit 40 Identify the type and size of flaws.

10: Optical inspection module 10A: Photography Team 10B: light source 10C: Photography lens 11: The first photography group 111: The first light source 112: The first photographic lens 12: The second photography group 121: second light source 122: second photographic lens 20: System host 21: Data receiving unit 22: positioning unit 23: Defect Judgment Server 24: Preliminary classification module 30: storage host 40: Artificial Intelligence Identification Unit 41: Training Module 42: Defect Classification Module 50: Tabulation unit 60: Imaging unit F: Cloth K1: First side K2: second side P: Schematic diagram of defects P1: Mark F1: First side F2: Second side T: record form S1: Steps to capture photos S2: Discrimination steps S3: Steps to obtain location S4: Storage steps S5: training steps S6: Classification steps S7: Tabulation steps S8: Drawing steps

Figure 1 is a schematic diagram of the optical inspection module of the integrated system for rapid defect detection of sheet materials according to the present invention. Figure 2 is a schematic diagram of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 3 is a schematic diagram of the system host of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 4 is a schematic diagram of the system host of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 5 is a schematic diagram of a preferred embodiment of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 6 is a schematic diagram of a preferred embodiment of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 7 is a method of using the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 8 is a schematic diagram of a preferred embodiment of the use method of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 9 is a schematic diagram of a preferred embodiment of the use method of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 10 is a schematic diagram of a preferred embodiment of the use method of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 11 is a schematic diagram of a preferred embodiment of the integrated system for rapid defect detection of sheet materials according to the present invention. Fig. 12 is a schematic diagram of a preferred embodiment of the integrated system for rapid defect detection of sheet materials according to the present invention.

10: Optical inspection module

20: System host

30: storage host

40: Artificial Intelligence Identification Unit

41: Training Module

42: Defect Classification Module

F: Cloth

Claims (6)

A sheet-shaped material rapid detection defect integration system for detecting a cloth, and the sheet-shaped material rapid detection defect integration system includes: an optical detection module for shooting the cloth; a system host, and the optical The detection module information connection. The system host has a data receiving unit, a positioning unit, and a defect judgment server for information connection. The data receiving unit is used to receive photos taken by the optical detection module. The defect judgment server provides To capture a defect photo, the positioning unit is used to bind each defect photo and the corresponding coordinate position of the defect on the sheet material; a storage host is connected to the host system information, and the storage host is used to receive the defect photo and The coordinate position bound to the defective photo; an artificial intelligence recognition unit connected to the storage host for information, the artificial intelligence recognition unit has a training module for information connection and a defect classification module for receiving Various model photos, the training module uses the model photos to form a judgment logic, the defect classification module is used to receive the defect photos, the defect classification module classifies according to the judgment logic, and then each defect photo and its defect Classification and binding, and then return to the storage host; the cloth has a first side and a second side opposite, the first side faces a first side, the second side faces a second side, the optical detection The module has at least one first photographing group and at least one second photographing group, the first photographing group has a first light source and a first photographing lens, and the second photographing group has a second light source and a second photographing lens , The first light source is arranged on the first side of the cloth, the first light source is used to illuminate the first surface of the cloth to form a front light source, the second light source is arranged on the second side of the cloth, the The second light source is used to illuminate the second side of the cloth to form a back light source. The first photographing lens and the second photographing lens are both arranged on the first side of the cloth, the first photographing lens and the second photographing lens The first side of the cloth is photographed, the first photographic lens is photographed through the front light source, the second photographic lens is photographed through the back light source; the data receiving unit is used to receive the first photographic lens and the second photographic lens A photo taken by a photographic lens. For example, the integrated system for rapid detection of defects in sheet materials as described in item 1 of the scope of patent application, wherein the storage host or the system host has a tabulation unit, and the tabulation unit is based on the defect photo and the defect coordinates bound to it. Defect classification forms a record form. For example, the sheet-like material rapid defect detection integration system described in item 1 of the scope of patent application, wherein the storage host or the system host has a tabulation unit, and the tabulation unit is based on the defect photo and the defect number bound to it. Pattern, defect coordinates, and defect classification to form a record form. For the sheet material rapid detection and defect integration system described in item 1 of the scope of patent application, the storage host or the system host has an imaging unit, and the imaging unit forms a defect according to the bound defect coordinates and defect classification In the schematic diagram, a mark is displayed on the position corresponding to the defect coordinates on the defect schematic diagram. As described in item 4 of the scope of patent application, the sheet-like material rapid detection defect integration system, the mark forms different patterns according to the classification of the defects, so as to visualize the schematic diagram. For example, the integrated system for rapid detection of defects in sheet materials described in the first item of the patent application, the system host also has a preliminary classification module for preliminary classification of defective photos.

Images