TW202323801A - Artificial leather defect detection system and method wherein the system includes a retracting device, a defect detection device and a management platform - Google Patents

Abstract

An artificial leather defect detection system includes a retracting device, a defect detection device and a management platform. The defect detection device includes a camera and an image recognition module. The defect detection method includes: retracting artificial leather by the retracting device; capturing a real-time image of the artificial leather by a camera; and generating and sending a defect data corresponding to a defective part to the management platform when the image recognition module recognizes the defective part in the real-time image. The management platform records the defect data in a database. The management platform can receive an inquiry command and obtain the defect data from the database according to the inquiry command. Therefore, the invention can replace manual visual inspections and can be used to query defect data as a source of defect traceability.

Description

Artificial leather defect detection system and detection method

The present invention is related to the detection of artificial leather; in particular, it refers to a method using deep learning to detect defects in artificial leather with a color industrial camera that can keep up with the speed of the retracting device.

Artificial leather is widely used in daily life. Artificial leather includes a base fabric layer and a rubber layer. During the production process, the surface of the rubber layer may have defects such as uneven coloring, wrinkles, damage, and holes. If these defects are not detected After inspection, it will continue to be sent to the processing process, which will cause waste of production capacity and cost. And if the shipped product contains defects, it will lead to the customer's request to return or modify the product, and it is more likely that it will need to be remade. Therefore, in the production process of artificial leather, defect detection is a very important part, not only to avoid the waste of production capacity and cost, but also to ensure the quality of the shipped products.

Traditional artificial leather defect detection relies on personnel's visual inspection, which will be affected by the subjective judgment and personal experience of the inspector during the visual inspection process, resulting in inconsistent inspection standards. In addition, long-term visual inspection by personnel will also lead to a decrease in detection accuracy. During production, due to the need to cooperate with the speed of visual inspection, the speed of artificial leather moving in the production line should not be too fast, otherwise it will be difficult to meet the demand for accurate detection in real time, and the production capacity of artificial leather defect detection will not be increased.

In view of this, the purpose of the present invention is to provide a detection system and detection method for artificial leather flaws, which can replace human visual detection.

Another object of the present invention is to provide an artificial leather flaw detection system and detection method, which can be used to query the detected flaw data.

In order to achieve the above object, the present invention provides a defect detection system for artificial leather, comprising a retracting device, including a detection area, the retracting device is used to retract an artificial leather, and the artificial leather is detected after passing through the detection area Rolling; a defect detection device, including a camera and an image recognition module, the camera is set up in the detection area of the winding device and captures a real-time image of the artificial leather passing through the detection area; the image recognition module is electrically When the image recognition module recognizes at least one defective part in the real-time image, the image recognition module generates a defect data corresponding to the at least one defective part; and a management platform, signal The defect detection device is connected, the management platform is connected to a database, and the management platform receives the defect data and records the defect data in the database; wherein, the management platform receives an inquiry instruction, and according to the inquiry The information command obtains the defect data from the database.

In addition, the present invention further provides a method for detecting artificial leather defects, which uses an artificial leather defect detection system, and the artificial leather defect detection system includes a retracting device, a defect detecting device and a management platform; wherein, the retracting device is used To roll up an artificial leather, the defect detection device includes a camera and an image recognition module; the management platform signal is connected to the defect detection device; the method of the artificial leather defect detection system includes the following steps: A. B. The camera captures a real-time image of the artificial leather being rolled up in step A; C. The image recognition module identifies whether there is at least one defect in the real-time image, and if so, the image recognition module The group generates a defect data corresponding to the at least one defective part; D. The management platform receives the defect data, and records the defect data in a database; E. The management platform receives a query instruction, and according to the query The information command obtains the defect data from the database.

The effect of the present invention is that the flaw detection device can detect flaws on the artificial leather instead of visual inspection by personnel, and can generate flaw data and record them in a database for querying the flaw data of the flaws.

For the sake of convenience and a clearer description, a preferred embodiment is given and described in detail in conjunction with the drawings as follows. Please refer to FIG. 1 , which is an artificial leather defect detection system 1 according to the first preferred embodiment of the present invention, including a retracting device 10 , a defect detection device 40 and a management platform 50 . Hereinafter, the artificial leather flaw detection method of this embodiment will be described together. The method includes the steps shown in FIG. 2 .

The retracting device 10 includes a retracting shaft 101, a yardage counter 102 and a warning light 103, wherein the retracting shaft 101 can be driven by a drive motor (not shown) to retract an artificial leather 100, the The artificial leather 100 includes a base fabric layer and a rubber layer. The rubber layer is combined with the base fabric layer. The rubber layer is an example of a PVC rubber layer. The yardage counter 102 is used to count a yardage of rolling the artificial leather 100 . The body of the retracting device 10 has a detection zone 103 , and the artificial leather 100 passes through the detection zone 103 and is retracted by the retracting shaft 101 during the retraction process. The warning light 104 is used for warning.

The defect detection device 40 includes a camera 20 and an image recognition module 30, wherein the camera 20 is mounted on the detection area 103 of the body of the retracting device 10 and the imaging range of the camera 20 can cover the rubber layer to continuously capture A real-time image of the rubber layer of the artificial leather 100 passing through the detection zone 103 is taken. In this embodiment, the camera 10 is an industrial camera, which has a higher image capture rate than ordinary cameras, so that clear real-time images can still be obtained when the artificial leather 100 is moving at high speed. The image recognition module 30 is electrically connected to the camera 20 , and is signally connected to the retracting device 10 and the management platform 50 . The image recognition module 30 receives the real-time images output by the camera 20, and continuously recognizes whether there are flaws in the obtained real-time images, so as to greatly improve the negligence, misjudgment and standardization of personnel status in the visual inspection of personnel. Inconsistent situations.

In this embodiment, the image recognition module 30 includes a storage unit 301 and a comparison unit 302, wherein a plurality of defect models are stored in the storage unit 301, and the defect models obtain their respective values for the plurality of defect images. After the features are used, a usable defect model is generated, wherein the defect model includes defect types such as deformation defect, damage defect, dirty defect, color difference defect, wrinkle defect and indentation defect, and the defect model is stored in the storage unit 301. The comparison unit 302 obtains the real-time image of the camera 20, extracts a feature in the real-time image and compares it with the defect model in the storage unit 301, and when it is judged that the feature in the real-time image conforms to any of the defect models , judging that the feature is at least one defect, and generating a defect data corresponding to the defect. The comparison unit 302 transmits the currently generated defect data to the management platform 50 .

More specifically, when the comparison unit 302 determines that there is a defective part in the real-time image, it records a static image with the defective part in the real-time image in the defect data, and obtains from the retracting device 10 when the defective part is found. The current yardage of the yardage counter 102 is recorded in the defect data. In addition, the comparison unit 302 further records the defect type of the defect model that the defect part conforms to in the defect data. In other words, the defect data includes still images, yardages, and defect types.

The management platform 50 is connected to a database 60, wherein the database 60 stores historical data related to the production process of the artificial leather 100. The historical data includes the manufacturing number, manufacturing process, material details, color formula and process parameters, etc. The management platform 50 records the manufacturing number of the artificial leather 100 currently being rolled by the retracting device 10 . Whenever the management platform 50 receives the defect data transmitted from the image recognition module 30, it establishes the corresponding relationship between the manufacturing number and the defect data, that is, adds the defect data to the artificial leather 100 in the database 60. in the biographical data.

Then the user can use an electronic device, such as a mobile phone, a computer, or any communicable device to connect to the management platform 50, and send an inquiry command to the management platform 50, and the management platform 50 will automatically The defect data is obtained from the database 60 . In this embodiment, the inquiry instruction includes an inquiry number, which may be, for example, a specific manufacturing number. The management platform 50 will compare the inquiry number of the inquiry order with the manufacturing number in the history data, find the history data that matches the inquiry number, and when the inquiry number matches the corresponding manufacturing number, it will immediately Biographical data containing the defective data can be obtained. The management platform 50 transmits the found resume data to the electronic device for the user to view. In this way, the history data of the artificial leather 100 in the production process can be obtained, so it can be used for traceability of defects.

In addition, in this embodiment, if the comparison unit 302 in the image recognition module 30 judges a defect during the comparison process, the comparison unit 302 simultaneously sends a stop signal to the retracting device 10. When the retracting device 10 receives the stop signal, it stops the retracting shaft 101 from retracting the artificial leather 100 and lights up the warning light 104, so that personnel can carry out follow-up processing on the defective parts of the artificial leather 100. If no defect has been detected, the comparison unit 302 does not send a stop signal, and the retracting device 10 continues to retract.

In this embodiment, the defect models stored in the storage unit 301 in the image recognition module 30 are compared and detected, and by judging whether the feature in the real-time image is a defect part, the defect can be correctly and quickly judged. It also significantly reduces the need for manpower.

In a second preferred embodiment of the present invention, the defect models can be classified into at least a first type of defect model and a second type of defect model, wherein the first type of defect model is unacceptable Defects, such as deformation defects, damage defects, stain defects, chromatic aberration defects, wrinkle defects, and indentation defects, the second type of defect models are acceptable small defects, such as small holes. When the comparison unit 302 of the image recognition module 30 judges that the feature in the real-time image conforms to the first type of defect model, it sends a stop signal to the retracting device 10, so that the retracting device 10 stops retracting, And at the same time light the stop warning light 104 to remind the user to follow up; and when the comparison unit 302 judges that the feature in the real-time image conforms to the second type of defect model, it will not generate a stop signal, allowing the volume to The retracting device 10 continues to retract the artificial leather 100.

Fig. 3 shows the artificial leather defect detection system 3 of the third preferred embodiment of the present invention, which has substantially the same structure and detection method as the first embodiment, the difference is that the image recognition module 30 further includes a manual The intelligence training unit 303 can use the artificial intelligence training unit 303 to perform deep learning training on the defect model in the storage unit 301 . Since the types and sizes of the artificial leather 100 flaws are complex and diverse, and the leather texture varies greatly with different products, which also increases the difficulty of flaw detection. Therefore, the real-time image of the camera 20 is used to match the image recognition. The artificial intelligence training unit 303 of the module 30 learns the different complex backgrounds, textures and environmental changes of various defect parts. After training, these defect models are generated and stored in the storage unit 301 .

To further explain, these flawed models may have problems such as Overfitting, Vanishing gradient, and Exploding gradient when trained using deep learning methods, so it is necessary to use a large amount of data to train the model Training, the method of deep learning can overcome the problem that the background is complex or the object type is not fixed, so it cannot be detected. Compared with the image processing method, it has better adaptability. Therefore, the comparison unit of this embodiment uses the object detection model YOLOv4 to detect the flawed part of the artificial leather 100 being rolled up, and then inputs the detected flawed subject image to the second deep learning model GoogLeNet , to identify the defect subject category. In this way, the comparison unit can more accurately identify whether the feature in the real-time image is a defect, so as to improve the detection accuracy. In addition, the artificial intelligence training unit 303 of this embodiment can also be applied in the second embodiment.

According to the above, the artificial leather defect detection system and detection method of the present invention can automatically detect the defective parts on the artificial leather 100, replace the visual detection by personnel, and solve the shortcomings of traditional defect detection relying on visual detection by personnel. The detected defect data can also be recorded in the database 60 as a history traceability of the artificial leather 100 production process.

The above description is only a preferred feasible embodiment of the present invention, and all equivalent changes made by applying the description of the present invention and the scope of the patent application should be included in the scope of the patent of the present invention.

[this invention] 1,3: Artificial leather defect detection system 10: retracting device 101: Take-up shaft 102: Code number counter 103: detection area 104: warning lights 20: Camera 30: Image recognition module 301: storage unit 302: comparison unit 303: Artificial Intelligence Training Unit 40: Defect detection device 50:Management platform 60: Database 100: artificial leather

FIG. 1 is a schematic diagram of an artificial leather defect detection system according to a first preferred embodiment of the present invention. Fig. 2 is a flow chart of the artificial leather defect detection method of the present invention. FIG. 3 is a schematic diagram of an artificial leather defect detection system according to a third preferred embodiment of the present invention.

1: Artificial leather defect detection system

10: retracting device

101: Take-up shaft

102: Code number counter

103: detection area

104: warning lights

20: Camera

30: Image recognition module

301: storage unit

302: comparison unit

40: Defect detection device

50:Management platform

60: Database

100: artificial leather

Claims (15)

A system for detecting artificial leather defects, comprising: A retracting device, including a detection area, the retracting device is used to retract an artificial leather, and the artificial leather is retracted after passing through the detection area; A defect detection device, including a camera and an image recognition module, the camera is set up in the detection area of the retracting device and captures a real-time image of the artificial leather passing through the detection area; the image recognition module is electrically connected to the The camera receives the real-time image, and when the image recognition module recognizes at least one defective part in the real-time image, the image recognition module generates a defect data corresponding to the at least one defective part; and A management platform, the signal is connected to the defect detection device, the management platform is connected to a database, and the management platform receives the defect data and records the defect data in the database; wherein, the management platform receives an inquiry instruction , and obtain the defect data from the database according to the query instruction. The artificial leather defect detection system as described in claim 1, wherein the database records a manufacturing number of the artificial leather, and the management platform establishes a corresponding relationship between the manufacturing number and the defect data after receiving the defect data; the inquiry The command includes an inquiry number, and the management platform obtains the defect data when comparing the inquiry number with the corresponding manufacturing number. The artificial leather defect detection system as described in claim 1, wherein the retracting device has a yardage counter, and the yardage counter counts a yardage of the artificial leather; the defect detection device is connected to the retracting device, And when the image recognition module recognizes the at least one defect in the real-time image, record the current code number in the defect data. The artificial leather defect detection system as described in claim 3, wherein when the image recognition module recognizes the at least one defective part in the real-time image, the image recognition module sends a stop signal to the retracting device, and the retracting device The device stops rolling the artificial leather according to the stop signal. The artificial leather flaw detection system as claimed in Claim 4, wherein the retracting device includes a warning light, and when the retracting device receives the stop signal, the warning light is turned on. The artificial leather defect detection system as described in Claim 1, wherein the image recognition module of the defect detection device includes a storage unit and a comparison unit, the storage unit stores a plurality of defect models, and the comparison unit judges the real-time When a feature in the image conforms to any of the defect models, it is determined that the feature is the at least one defect part. The artificial leather defect detection system as described in claim 6, wherein the image recognition module includes an artificial intelligence training unit, and the artificial intelligence training unit performs training based on a plurality of images to generate the defect models. The artificial leather defect detection system as described in claim 6, wherein the defect detection device is connected to the retracting device; the defect models include a first type of defect model and a second type of defect model; when the comparison When the unit determines that the feature in the real-time image conforms to the defect model of the first type, it determines that the feature is the at least one defect part, and generates a stop signal, and the retracting device stops retracting the artificial leather according to the stop signal; When the comparison unit determines that the feature in the real-time image conforms to the second type of defect model, it determines that the feature is the at least one defect part, and does not generate the stop signal. A method for detecting artificial leather defects, which is to apply an artificial leather defect detection system, the artificial leather defect detection system includes a retracting device, a defect detection device and a management platform; wherein, the retracting device is used to retract an artificial leather , the defect detection device includes a camera and an image recognition module; the management platform signal is connected to the defect detection device; the artificial leather defect detection system method includes the following steps: A. The retracting device retracts the artificial leather; B. The camera captures a real-time image of the artificial leather being rolled up in step A; C. The image recognition module identifies whether there is at least one defective part in the real-time image, and if so, the image recognition module generates a defect data corresponding to the at least one defective part; D. The management platform receives the defect data and records the defect data in a database; E. The management platform receives an inquiry instruction, and obtains the defect data from the database according to the inquiry instruction. The artificial leather defect detection method as described in claim 9, wherein in step D, the management platform establishes a correspondence between the manufacturing number and the defect data after receiving the defect data; in step E, the inquiry command includes an inquiry number, The management platform obtains the defect data when comparing the inquiry number with the corresponding manufacturing number. The artificial leather defect detection method as described in Claim 9, wherein step A includes, the retracting device counts the number of yards of the artificial leather rolled; step C includes, the image recognition module records the current state in the defect data of the yardage. In the artificial leather defect detection method as described in Claim 9, in step C, when the image recognition module recognizes the at least one defect in the real-time image, the retracting device stops retracting the artificial leather. The artificial leather defect detection method as described in Claim 9, wherein the image recognition module of the defect detection device includes a storage unit and a comparison unit; the method of the artificial leather defect detection system includes: establishing a plurality of Defect model: in step C, when the comparison unit determines that a feature in the real-time image conforms to any of the defect models, it is determined that the feature is the at least one defect part. The artificial leather flaw detection method as described in claim 13, wherein the image recognition module includes an artificial intelligence training unit; the artificial leather flaw detection system method includes: the artificial intelligence training unit performs training based on a plurality of images to generate the some flawed models. The artificial leather defect detection method as described in claim 13, including classifying these defect models into at least a first type of defect model and a second type of defect model; in step C, when the comparison unit judges that the real-time When the feature in the image conforms to the defect model of the first type, stop the retracting device from retracting the artificial leather; when the comparison unit judges that the feature in the real-time image conforms to the defect model of the second type, The retracting device maintains retracting the artificial leather.

Images