TWI818668B - Method for checking defects of products - Google Patents

Abstract

A method for checking defects of products is used to solve the problem of misjudgment caused by manual checking. The method includes gripping a product to check by a gripping device to deliver the product to check to an image-capture unit. An appearance image of the product to check is captured by the image-capture unit and is delivered to a processing unit. The processing unit compares the appearance image of the product to check with an image of types of defects and/or an image of locations of defects to determine whether a defect is present on the product to check.

Description

Product appearance defect inspection methods

The present invention relates to a method for inspecting defects in product appearance, particularly an inspection method for detecting and eliminating defects in processed and manufactured products.

In the processing manufacturing industry, when a semi-finished product or finished product is formed by casting or forging, etc., the appearance and weight of the semi-finished product or finished product will be checked to see if they do not meet the specifications to prevent the defective product from being further processed and the processed manufacturing process The quality of products results in inferior products and affects production efficiency. The existing defect detection method is to visually inspect the appearance of the semi-finished product or finished product manually to determine whether there are defects in the appearance. However, performing the above inspection manually requires a lot of time and labor costs, and there will also be differences in interpretation when inspected by different personnel, and defective products may not be detected due to human errors.

In view of this, there is still a need to improve the existing defect inspection methods for product appearance.

In order to solve the above problems, the purpose of the present invention is to provide a method for inspecting defects in product appearance, which can automatically conduct defect inspection to reduce labor costs.

A secondary purpose of the present invention is to provide a method for inspecting product appearance defects, which can unify defect judgment standards and reduce the occurrence of misjudgments.

The use of the quantifier "a" or "an" in the elements and components described throughout the present invention is only for convenience of use and to provide a common sense of the scope of the present invention; in the present invention, it should be interpreted as including one or at least one, and single The concept also includes the plural unless it is obvious that something else is meant.

The product appearance defect inspection method of the present invention includes: using a clamping device to clamp a product to be inspected, and transporting the product to be inspected to an image capture unit; using the image capture unit to capture an image of the product to be inspected Appearance image, obtain an appearance image information, and transmit the appearance image information of the product to be inspected to a computing unit; store the defect shape image or/and defect location image of the product to be inspected in advance, and convert the image The appearance image information captured by the acquisition unit is compared and identified with the defect shape image or/and defect location image pre-stored in the computing unit to determine whether the appearance image of the product to be inspected has defects; The product to be inspected after unit interpretation is transported to a marking unit. If the interpretation result is that the product to be inspected is defective, the marking unit will mark the product to be inspected. The defects on the product to be inspected are repairable defects. The marking unit marks the product to be inspected as a product to be repaired; and sends the product to be repaired to a logistics module, the logistics module has at least one conveyor belt, and the at least one conveyor belt is used to transport the product to be repaired, The product to be repaired is transported to a processing area, the defects of the product to be repaired are repaired in the processing area, and a repaired product is obtained.

Therefore, the product appearance defect inspection method of the present invention uses the clamping device to clamp the product to be inspected, and sends the product to be inspected to the image capture unit, and the image capture unit captures the product to be inspected The appearance image of the product. The appearance image of the product to be inspected can be sent to the computing unit and compared with the defect shape image and defect location image stored in the computing unit to determine whether the appearance of the product to be inspected has defects. and eliminated. The system can replace manual work with computer automation to reduce labor costs and avoid errors caused by manual interpretation. The system can achieve the effect of checking accuracy. In addition, it can be achieved to facilitate the classification of defective products. The effect can be intuitively known that the inspected product is a product to be repaired, and repairing the product to be repaired to become a good product has the effect of reducing the amount of waste and reducing production costs.

Wherein, the product to be inspected has at least one inspection surface, and the clamping device rotates the product to be inspected so that at least one inspection surface of the product to be inspected is located on the image capture unit. In this way, each inspection surface of the product to be inspected can be respectively aligned with the image capture unit, which has the effect of inspecting each inspection surface of the product to be inspected.

Among them, the image of the defect shape of the product to be inspected is at least one image of holes, cracks, water lines, knocks, abrasions and color difference. In this way, the product to be inspected can be inspected for various defect states.

Among them, the computing unit has an artificial intelligence module, and the defect shape image or/and defect location image of the inspection surface is pre-stored by the computing unit for the artificial intelligence module to memorize and learn, so that the artificial intelligence module of the computing unit can Identify defect shape images or/and defect location images that are not stored in the computing unit. In this way, the accuracy of identifying the defective product can be improved.

Among them, the repaired product is sent to the image capture unit through the logistics module, and the image capture unit captures the appearance image of the repaired product, and the computing unit interprets the image captured by the image capture unit. Whether the appearance of the repaired product is no longer defective. In this way, you can ensure that the repaired product has been restored to a good quality.

The product appearance defect inspection method of the present invention may also include sending the product to be inspected after interpretation by the computing unit to a logistics module. The logistics module has at least one conveyor belt, and the at least one conveyor belt is used for transportation. The product to be inspected is transported to at least one of a good product area, a processing area and a waste area. In this way, the production line can be automated to reduce labor costs.

Among them, the logistics module has a weight detection device, and the weight detection device measures The weight of the product to be checked is then placed on the at least one conveyor belt. In this way, the products to be inspected are further classified by weight.

1: Gripping device

2:Image capture unit

3: Mark unit

4: Logistics module

41,41a,41b,41c: Conveyor belt

42: Clamping piece

43: Weight detection device

S1: Image capture steps

S2: Interpretation steps

S3: Marking steps

S4: Logistics steps

H: Products to be inspected

H1: Inspection surface

H2: Hole defects

C: Operation unit

[Figure 1] A flow chart of the first embodiment of the present invention.

[Figure 2] Schematic diagram of the detection device according to the first embodiment of the present invention.

[Figure 3] Schematic diagram of the present invention used to inspect a defective product.

[Figure 4] A flow chart of the second embodiment of the present invention.

[Figure 5] Schematic diagram of the marking steps of the second embodiment of the present invention.

[Figure 6] A three-dimensional view of the logistics module according to the second embodiment of the present invention.

In order to make the above and other objects, features and advantages of the present invention more obvious and understandable, preferred embodiments of the present invention are illustrated below and described in detail with reference to the accompanying drawings; in addition, the same symbols are used in different drawings. are considered to be the same and their description will be omitted.

Please refer to Figure 1, which is a first embodiment of the product appearance defect inspection method of the present invention, and includes an image capturing step S1 and an interpretation step S2.

Referring to Figure 2, the image capturing step S1 includes gripping a product H to be inspected by a gripping device 1 to transport the product H to be inspected to an image capturing unit 2. The invention does not limit the type of the product H to be inspected. In this embodiment, the product H to be inspected is a golf club head. The clamping device 1 can be connected to a power supply source, so that the power supply source provides power for the operation of the clamping device 1 . The gripping device 1 can be a robotic arm, and the operator can automatically control the gripping device 1 to grip the product H to be inspected according to the program set by the central processing system. For example, The gripping device 1 can be a six-axis mechanical arm, so that the gripping device 1 can reach any position and angle in the three-dimensional space, so as to facilitate the gripping of the product H to be inspected and move it to any position. In addition, the positioning accuracy of the robotic arm can reach 0.05mm. Specifically, when the robot arm clamps the product H to be inspected to a target position, the distance error between the actual position of the product H to be inspected and the target position can be reduced to 0.05 mm. Robotic arms are not as prone to fatigue problems as manual labor, and the errors are generally more consistent without sudden increases or decreases, making them suitable for a large number of inspection operations. In one embodiment, the positioning accuracy of the robotic arm is about 0.05 mm to about 1.4 mm. In another embodiment, the positioning accuracy of the robotic arm is 0.05mm, 0.08mm, 1.1mm or 1.4mm.

Please refer to Figures 2 and 3. The image capturing step S1 continues to include capturing the appearance image of the product H to be inspected through the image capturing unit 2. The image capturing unit 2 can be a camera or video camera, so as to Obtain the appearance image information of the product H to be inspected. For example, the image capturing unit 2 can capture 2D images through monocular line scanning, or capture 3D images through binocular line scanning. In the image capturing step S1, images of various parts of the product H to be inspected can be captured as required. Specifically, the product H to be inspected can have at least one inspection surface H1, and the clamping device 1 can rotate the product H to be inspected so that at least one inspection surface H1 of the product H to be inspected is aligned with the image capture Unit 2, so that the image capturing unit 2 can capture the image of the at least one inspection surface H1. In addition, when there are several inspection surfaces H1, the clamping device 1 can rotate the product H to be inspected several times to align each inspection surface H1 with the image capture unit 2 respectively.

In this embodiment, the golf club head of the product H to be inspected has six inspection surfaces H1, such as Crown, Sole, Face, Back, and Heel. Heel) and toe (Toe). Specifically, after the clamping device 1 aligns an inspection surface H1 of the product H to be inspected with the image capture unit 2, the clamping device 1 continues to rotate the product H five times to separate the golf balls. The inspection surface H1 of each part of the club head is aligned with the image capture unit 2, thereby completing the image capture of each part of the product H to be inspected, that is, the aforementioned six Check the image capture of surface H1 (hereinafter referred to as six-surface capture). Moreover, the time taken for the six-sided imaging is about 30 seconds to about 50 seconds. In some embodiments, the inspection surface H1 may further include the shaft, skirt or other parts of the golf club head.

The interpretation step S2 includes transmitting the appearance image information of the inspection surface H1 to a computing unit C, and using the computing unit C to identify defects on the appearance image information to determine whether the inspection surface H1 has defects. The computing unit C can be a computer, computer or other device with storage and computing processing functions. The computing unit C pre-stores the defect shape image of the inspection surface H1 and the defect location image of the inspection surface H1 for comparison with the appearance image information of the inspection surface H1 captured by the image capture unit 2 . Among them, the computing unit C can pre-store the defect shape image of the inspection surface H1 and the defect position image of the inspection surface H1, which include the defect shapes that are known to occur during the manufacturing or processing of the product H to be inspected and the defects that have been inspected. The invention does not limit the location of the defect that is known to occur; or the form of the defect that is estimated to occur and the location of the defect that is estimated to occur. For example, as far as the actual production process of golf club heads is concerned, when the yield rate is difficult to reach 100%, at least one golf club head with defects will inevitably be produced. In this disclosure, at least one of the aforementioned defective golf club heads is selected, an image of the defect is captured, and the image is input into a computer for storage and artificial intelligence training (see below for details). Therefore, the computer can establish a basis for interpretation based on the image of the golf club head with defects, so as to identify defects in the golf club heads to be subsequently inspected.

Following the above, in one embodiment, the product H to be inspected is a golf club head to be inspected, and the above-mentioned defect shape images may be images of holes, cracks, water lines, knocks, abrasions, and color differences. The above-mentioned defect location images may be images of the same defect at different locations on the same inspection surface H1. For example, the hole defect H2 is located at the edge or in the middle of the hitting surface of the inspection surface H1. When the appearance image information of the inspection surface H1 captured by the image capture unit 2 shows a defect appearance that is generally consistent with the above-mentioned defect shape image and defect location image, the inspection surface captured by the image capture unit 2 can be H1 is judged to be defective, otherwise the image captured by capture unit 2 will be The inspection surface H1 is judged as a good product. It is worth noting that the image of the defective shape that can be repaired and the image of the defective shape that cannot be repaired can be pre-stored in the computing unit C. In this way, the computing unit C can further determine whether the defect on the inspection surface H1 is a repairable defect, and treat the product H to be inspected as a product to be repaired; or, determine whether the defect on the inspection surface H1 is irreparable. If there is a defect, the product H to be inspected will be regarded as a discarded product. For example, if the diameter of the hole on the hitting surface is larger than 5mm, it is no longer productive in terms of processing and repair. Therefore, several images of hole defects with a diameter larger than 5mm can be trained in the computing unit C in advance; after training, if the computing unit C encounters a hole with a diameter larger than 5mm, it can be interpreted as an irreparable defect and classified as Classified as waste.

In this embodiment, the computing unit C has an artificial intelligence module (Artificial Intelligence). The artificial intelligence module has a self-learning function, which is understandable to those in the art and will not be described in detail here. The plurality of defect shape images of the inspection surface H1 and the defect position images of the inspection surface H1 pre-stored by the computing unit C can be used for memory learning by the artificial intelligence module. In this way, the artificial intelligence module of the computing unit C can identify the defect shape image and defect location image that are not stored in the computing unit C. That is, the artificial intelligence module of the computing unit C can identify the hole defect H2 image at any position on the inspection surface H1, and is not limited to the hole defect H2 image pre-stored in the computing unit C. Alternatively, the artificial intelligence module can identify defect images on the inspection surface H1 that are similar to the hole defect H2, for example, but are not limited to images that are exactly the same as the hole defect H2 pre-stored in the computing unit C.

In one embodiment, known defects used for artificial intelligence module training and defects to be identified may be located on different inspection surfaces H1. For example, among the known defect images used for training, the defect image is captured from the crown of a golf club head, and this defect image shows a distribution of multiple holes. Correspondingly, the image of the defect to be identified is taken from the hitting face of the golf club head, which also has a plurality of holes distributed. After training by the artificial intelligence module, the computing unit C not only Hole-shaped defects on the crown can be identified, as well as hole-shaped defects on the hitting surface. In other words, the computing unit C is trained based on the image of the hole defect in the crown; after training, even if the hole defect to be identified is located in a different environment (i.e., the hitting surface), the computing unit C can still identify the hole defect. .

Please refer to Figures 4 and 5. The product appearance defect inspection method of the present invention may additionally include a marking step S3. The marking step S3 includes using the clamping device 1 to inspect the product H after being interpreted by the computing unit C. Transported to a marking unit 3, the product H to be inspected is marked by the marking unit 3. The marking unit 3 may be a spray gun, and the marking unit 3 may be aligned with the product H to be inspected by a nozzle to perform inkjet marking on the product H to be inspected. The marking unit 3 is coupled to the computing unit C to receive the interpretation result of the computing unit C, and to mark the product H to be inspected based on the interpretation result. For example, if the interpretation result is that the product H to be inspected is defective, the product H to be inspected will be marked; otherwise, the product H to be inspected will be regarded as a good product and will not be marked. In addition, when the product H to be inspected has defects, the marking step S3 can perform inkjet in different colors or different patterns on several defects on the inspection surface H1 to further mark different types of defects or defects at different locations. flaw. For example, if the defect on the inspection surface H1 is an irreparable defect, the marking unit 3 will mark the product H to be inspected with a predetermined color or a predetermined pattern to indicate it is a waste product; alternatively, the defect on the inspection surface H1 is For repairable defects, the marking unit 3 marks the types of defects with different predetermined colors or different predetermined patterns, and uses the marks as a basis for classification of subsequent repair processes. In one embodiment, the marking pattern can be a simple shape such as a circle, a square, or a triangle. Such simple shapes can reduce the time required for marking.

In addition, the marking unit 3 can be fixed on a moving mechanism, and the moving mechanism performs linear motion along a single axis. When the product H to be inspected has defects and needs to be marked, the clamping device 1 can position the product H to be inspected below the marking unit 3, and then the marking unit 3 performs line inkjet marking through a moving mechanism. In another embodiment, the marking unit 3 is set to be stationary. When performing inkjet marking, the clamping device 1 clamps the product H to be inspected and moves linearly along a uniaxial direction, thereby further integrating the action path involving mobility into the clamping device 1 .

Please refer to Figures 4 and 6. The product appearance defect inspection method of the present invention can also include a logistics step S4, which is to send the interpreted product H to be inspected to a logistics module 4 for processing by the logistics module. 4. Transport the interpreted product H to be inspected. The product H to be inspected after performing the interpretation step S2 can be sent to the logistics module 4 , or the product H to be inspected after the marking step S3 has been performed can be sent to the logistics module 4 . The logistics module 4 may have at least one conveyor belt 41, and the at least one conveyor belt 41 is used to transport the product H to be inspected.

In addition, the logistics module 4 may have a clamping member 42 , and the clamping member 42 may be a robotic arm to use the clamping member 42 to handle workpieces (such as but not limited to the above-mentioned qualified products, products to be repaired, or waste), so as to place the above-mentioned workpiece on the at least one conveyor belt 41. For example, when there are several conveyor belts 41, the good products, the products to be repaired and the discarded products can be placed on different conveyor belts 41a, 41b, 41c respectively, so that the good products can be shipped collectively from one of the conveyor belts 41a, or the products to be repaired can be shipped together. The repaired products are sent back for repair by another conveyor belt 41b, and the discarded products are collected and discarded by another conveyor belt 41c. In addition, the workpiece can be transported inertially by the clamping device 1 . For example, the product H to be inspected is transported from the image capture unit 2 and the marking unit 3 to the logistics module 4 to form an automated operation. In addition, the logistics module 4 can have a weight detection device 43, which can be a scale. The clamp 42 can place the product H to be inspected after being weighed by the weight detection device 43. According to the weight detection device 43, the product H to be inspected can be weighed. Check the weight of product H and choose to place it on different conveyor belts 41a, 41b, 41c. For example, two weights of products H to be inspected are transported by two conveyor belts 41a and 41b respectively, whereby appropriate additional accessories can be selected according to the weight of the products H to be inspected to control the final total weight of the finished product. In this way, the product H to be inspected is further classified by weight.

In another embodiment, when the computing unit C determines that the product H to be inspected is the product to be repaired, the product to be repaired can be transported to a processing area through one of the conveyor belts 41 for processing. The processing area repairs the defects of the product to be repaired. Or the products to be repaired with different defects can be placed on different conveyor belts 41a, 41b, 41c respectively, so that the products to be repaired with different defects can be subjected to processing procedures such as grinding and painting respectively. In addition, the repaired product can also be sent to the image capture unit 2 again through the logistics module 4. The image capture unit 2 captures the appearance image of the repaired product, and the computing unit C interprets the image capture. Check whether the repaired product captured in unit 2 has been repaired (that is, it has no defects). In this way, you can ensure that the repaired product has been restored to a good quality.

To sum up, the product appearance defect inspection method of the present invention uses the clamping device to clamp the product to be inspected, and sends the product to be inspected to the image capture unit, and is captured by the image capture unit The appearance image of the product to be inspected can be sent to the computing unit and compared with the defect shape image and defect location image stored in the computing unit to determine the appearance of the product to be inspected. Whether there are defects and eliminate them. The system can replace manual work with computer automation to reduce labor costs and avoid errors caused by manual interpretation. The system can achieve the effect of checking accuracy.

Although the present invention has been disclosed using the above-mentioned preferred embodiments, they are not intended to limit the invention. Anyone skilled in the art can make various changes and modifications to the above-described embodiments without departing from the spirit and scope of the invention. The technical scope protected by the invention, therefore, the protection scope of the invention shall include all changes within the literal and equivalent scope described in the appended patent application scope.

S1: Image capture steps

S2: Interpretation steps

Claims (7)

A method for inspecting product appearance defects, including: using a clamping device to clamp a product to be inspected, transporting the product to be inspected to an image capture unit; and capturing an appearance image of the product to be inspected by the image capture unit , obtain an appearance image information, and transmit the appearance image information of the product to be inspected to a computing unit; pre-store the defect shape image or/and defect location image of the product to be inspected in the computing unit, and capture the image The appearance image information captured by the unit is compared and identified with the defect shape image or/and defect location image pre-stored in the computing unit to determine whether the appearance image of the product to be inspected is defective; after the computing unit has interpreted The product to be inspected is transported to a marking unit. If the interpretation result is that the product to be inspected is defective, the marking unit will mark the product to be inspected. The defect on the product to be inspected is a repairable defect, and the marking unit will The product to be inspected is marked as a product to be repaired; and the product to be repaired is sent to a logistics module. The logistics module has at least one conveyor belt, and the at least one conveyor belt is used to transport the product to be repaired to transport the product to be repaired. The product to be repaired is transported to a processing area, where the defects of the product to be repaired are repaired, and a repaired product is obtained. The product appearance defect inspection method of claim 1, wherein the product to be inspected has at least one inspection surface, and the clamping device rotates the product to be inspected so that at least one inspection surface of the product to be inspected is located at the image capture unit. For example, the method for inspecting product appearance defects according to claim 1, wherein the image of the defect shape of the product to be inspected is at least one image of holes, cracks, water lines, knocks, abrasions, and color differences. For example, the product appearance defect inspection method of claim 2, wherein the computing unit has an artificial intelligence module, and the computing unit pre-stores the defect shape image of the inspection surface or/ and defect location images for the artificial intelligence module to memorize and learn, so that the artificial intelligence module of the computing unit can identify defect shape images or/and defect location images that are not stored in the computing unit. For example, if the defect inspection method of the product appearance of item 1 is requested, the repaired product is sent to the image capture unit through the logistics module, and the image capture unit then captures the appearance image of the repaired product, and the image is captured by the image capture unit. The computing unit determines whether the appearance image of the repaired product captured by the image capturing unit no longer has defects. For example, the product appearance defect inspection method of claim 1 further includes sending the product to be inspected after interpretation by the computing unit to a logistics module. The logistics module has at least one conveyor belt, and the at least one conveyor belt is To transport the product to be inspected, the product to be inspected is transported to at least one of a good product area, a processing area and a waste area. The product appearance defect inspection method of claim 6, wherein the logistics module has a weight detection device that measures the weight of the product to be inspected and then places it on the at least one conveyor belt.

Images