KR20220065545A - Apparatus for checking defect of object having atypical surface - Google Patents

Abstract

According to an embodiment of the present invention, an apparatus for checking a defect in a product having an atypical exterior comprises: a rotating mount on which an atypical product is mounted and which rotates the atypical product according to a selected algorithm; a plurality of cameras which are disposed in each area to surround the atypical product mounted on the rotating mount and photograph the atypical product in each area according to the rotating algorithm of the rotating mount; a plurality of lights disposed in each area where the plurality of cameras are disposed, to correspond to the respective cameras to emit light to the atypical product; a rotation photography control unit which controls rotation of the rotating mount according to the algorithm of the rotating mount and controlling photography of each of the plurality of cameras in response thereto; and a defect checking unit for detecting a defect in the atypical product through images captured by the plurality of cameras. Therefore, the apparatus can more accurately detect defects, such as foreign matter, dents, and burrs, in the exterior of a product with a curved surface.

Description

Apparatus for inspecting defects of products having atypical appearance

The present invention relates to an apparatus for inspecting defects of products having an atypical appearance, and more particularly, while rotating a product subject to defect inspection according to a specified routine algorithm, acquiring a plurality of images through a plurality of cameras, It relates to an apparatus for more accurately inspecting various defects of the atypical appearance of the product by deep learning inspection of the image.

In general, a vision inspection apparatus is a device capable of determining and sorting whether a product is defective by photographing or sensing an object to be inspected by a camera or a sensor and inspecting the shape or size of a product, or a foreign object or scratch. As such a vision inspection device can automatically inspect for defects by continuously supplying inspection objects, it is usefully used to thoroughly inspect various products such as mobile phones and automobile parts.

The conventional vision inspection method detects and classifies defects by detecting simple gray changes in 2D images. There is a problem that it cannot be determined that the part where the phase gray difference occurs is defective. For metal products with curved surfaces, defects in appearance (foreign substances, dents, burrs, etc.) cannot be detected through computer vision due to diffuse light reflection on the curved surface or roughness of the surface. In the conventional method, only dimension measurement or hole presence/absence inspection can be performed through computer vision in a state in which the product image is simply saturated.

Product appearance inspection according to the prior art is a situation in which the inspector directly proceeds with a visual inspection through an inspection microscope, and finds defective elements of the appearance by turning one product or changing an angle. However, this method inevitably causes a difference in defect detection due to the limitation of the method that each inspector recognizes by eye, and there is an unreasonable problem in that the inspection speed is divided into a fast person and a slow person. Also, there is a problem that even the same inspector may have different inspection standards for each day or time. As such, in the conventional inspection method, due to the external characteristics of the atypical metal product, the inspection by computer vision is impossible, and thus it has no choice but to rely on the visual inspection. Accordingly, there is a demand for the development of a device capable of more accurately and rapidly detecting defects on the appearance of products having an atypical appearance.

Republic of Korea Patent Registration No. 10-1595708 (published on February 18, 2016)

The present invention has been devised to improve the prior art as described above, and while rotating a product subject to defect inspection according to a specified routine algorithm, acquire a plurality of images through a plurality of cameras, and inspect the acquired images by deep learning. An object of the present invention is to provide an apparatus for inspecting defects of a product having an atypical appearance for more accurately inspecting various defects for the atypical appearance of the product.

In order to achieve the above object and solve the problems of the prior art, an apparatus for inspecting a defect of a product having an atypical appearance according to an embodiment of the present invention is a device in which the atypical product is mounted and the atypical product is selected a rotating mount that rotates according to an algorithm; a plurality of cameras disposed in each area in a form surrounding the atypical product mounted on the rotary mount and photographing the atypical product in each area in response to a rotation algorithm of the rotary mount; a plurality of lights disposed corresponding to the respective cameras in each area in which the plurality of cameras are disposed and irradiating light to the atypical product; a rotation photographing control unit for controlling the rotation of the rotation mount according to the algorithm of the rotation mount and correspondingly controlling the photographing of each of the plurality of cameras; and a defect inspection unit configured to detect defects of the atypical product through the images captured by the plurality of cameras.

In addition, according to the algorithm, the rotation imaging control unit of the device for inspecting defects of a product having an atypical appearance according to an embodiment of the present invention stops the rotation mount for a certain period of time after the rotation mount rotates by a rotation angle of 36 degrees It is characterized in that the rotation of the rotary mount is controlled to perform a routine 10 times consecutively from 0 degree to 360 degree rotation angle.

In addition, the rotation imaging control unit of the apparatus for inspecting a defect of a product having an atypical appearance according to an embodiment of the present invention is configured to allow the plurality of cameras to display the atypical images at each position during the time when the rotation is stopped for each routine. It is characterized in that each product is photographed and each camera is controlled to acquire 10 images during the 10 routines.

In addition, the defect inspection unit of the apparatus for inspecting a defect of a product having an atypical appearance according to an embodiment of the present invention is a computing device equipped with artificial intelligence having a CNN (Convolution Neural Network) neural network, and using the CNN neural network It is characterized in that the defect of the atypical product is inspected through model generation through deep learning image learning.

In addition, in the apparatus for inspecting a defect of a product having an atypical appearance according to an embodiment of the present invention, the deep learning image learning using the CNN neural network is a sample image input step, a learning data labeling step, a deep learning learning progress step , is implemented through the model generation step, and the defect inspection unit compares the images received from the plurality of cameras with the model generated through the deep learning to inspect the appearance defect of the atypical product.

According to the device for inspecting defects of products having an atypical appearance of the present invention, defects such as foreign substances, dents, burrs, etc. in the appearance are detected regardless of the light diffuse reflection of the curved surface or the roughness of the surface for a product having a curved surface in the exterior. A more accurate detection effect can be obtained.

In addition, according to the device for inspecting defects of products having an atypical appearance of the present invention, various types of defects are detected in a shorter time by deep learning inspection of images with multiple inspection models and multiple threads according to various defect types. The effect of more accurate inspection can be obtained.

1 is a diagram showing the configuration of an apparatus for inspecting defects of a product having an atypical appearance according to an embodiment of the present invention.
2 is a diagram illustrating a configuration in which a plurality of cameras are disposed according to an embodiment of the present invention.
3 is a view showing the configuration of lighting according to an embodiment of the present invention.

In order to clarify the characteristics and advantages of the problem solving means of the present invention, the present invention will be described in more detail with reference to specific embodiments of the present invention shown in the accompanying drawings. However, detailed descriptions of well-known functions or configurations that may obscure the gist of the present invention in the following description and accompanying drawings will be omitted. Also, it should be noted that, throughout the drawings, the same components are denoted by the same reference numerals as much as possible.

The terms or words used in the following description and drawings should not be construed as being limited to conventional or dictionary meanings, and the inventor may appropriately define the concept of terms for describing his invention in the best way. Based on the principle that there is, it should be interpreted as meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in this specification and the configurations shown in the drawings are only the most preferred embodiment of the present invention, and do not represent all of the technical spirit of the present invention. It should be understood that there may be equivalents and variations.

In addition, terms including ordinal numbers such as 1st, 2nd, etc. are used to describe various components, and are used only for the purpose of distinguishing one component from other components, and to limit the components. not used For example, without departing from the scope of the present invention, the second component may be referred to as the first component, and similarly, the first component may also be referred to as the second component. In addition, when an element is referred to as being “connected” or “connected” to another element, it means that it is logically or physically connected or can be connected.

In other words, it should be understood that a component may be directly connected or connected to another component, but another component may exist in between, and may be indirectly connected or connected. In addition, terms such as "comprises" or "have" as used herein are intended to designate that the features, numbers, steps, operations, components, parts, or combinations thereof described in the specification exist, but one or the It should be understood that the above does not preclude the possibility of the existence or addition of other features or numbers, steps, operations, components, parts, or combinations thereof.

In addition, terms such as “…unit”, “…group”, and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented as hardware or software or a combination of hardware and software. there is. Also, "a or an", "one", "the" and similar terms are otherwise indicated herein in the context of describing the invention (especially in the context of the following claims). or may be used in a sense including both the singular and the plural unless clearly contradicted by context.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a view showing the configuration of an apparatus for inspecting defects of a product having an atypical appearance according to an embodiment of the present invention, and FIG. 2 is a configuration in which a plurality of cameras are disposed according to an embodiment of the present invention. 3 is a diagram showing the configuration of lighting according to an embodiment of the present invention.

The apparatus 100 for inspecting a defect of a product having an atypical appearance according to an embodiment of the present invention includes a rotation mount 110 , a plurality of cameras 120 , a plurality of lights 130 , and a rotation photography control unit 140 . ), and a defect inspection unit 150 as a configuration.

An atypical product is mounted on the rotating mount 110 and the atypical product is rotated according to a selected algorithm. The atypical product means a product in which the appearance is not formally configured and irregular curves are formed on the exterior surface, so that diffuse reflection of light due to the curved surface or the roughness of the surface is not constant. The rotating mount 110 may be implemented in a rotating form.

A plurality of cameras 120 are arranged in each area in a form surrounding the atypical product mounted on the rotating mount 110 . For example, as shown in FIG. 2, three cameras are installed at intervals of 120 degrees in a form surrounding the side of the atypical product, and one or more cameras may be further installed on the upper side of the three, and orthogonal to the three cameras One or more cameras may be further installed to face the top of the product in the form of The number or direction of installation of the cameras may be implemented so that the administrator can set it in various ways in consideration of the appearance of the atypical product to be inspected. A plurality of cameras 120 photograph the atypical product in each of the regions in response to the rotation algorithm of the rotation mount 110 . The rotation algorithm is implemented by the rotation imaging control unit 140 .

The plurality of lights 130 are disposed corresponding to the respective cameras in each area in which the plurality of cameras 120 are disposed to irradiate light to the atypical product. As shown in FIG. 3 , the plurality of lights 130 may be implemented as diffuse C-shaped lights. In the case of general direct light type lighting, a saturation condition occurs due to diffuse reflection of light irradiated to the curved surface of atypical products, resulting in appearance inspection dead area. Therefore, in the present invention, it is possible to create a certain condition to the extent that the appearance inspection is possible even on a curved surface by applying special diffuse lighting.

The rotation imaging control unit 140 controls the rotation of the rotation mount 110 according to the algorithm of the rotation mount 110 , and correspondingly controls the photographing of each of the plurality of cameras 120 . For example, according to the algorithm, the rotation imaging control unit 140 performs a routine in which the rotation mount 110 rotates by a rotation angle of 36 degrees and then stops for a certain period of time from 0 degrees to a rotation angle of 360 degrees. It is possible to control the rotation of the rotation mount 110 to be performed continuously. In this case, the rotation photography control unit 140 captures the atypical product at each position by a plurality of cameras 120 during the time the rotation is stopped for each routine, and performs 10 times for each of the 10 routines for each camera. It can be controlled to acquire an image of the intestine. The rotation routine method of the rotation mount 110 and the photographing routine method of the camera may be set to various values according to the judgment of the administrator.

The defect inspection unit 150 detects a defect of the atypical product through images captured by a plurality of cameras 120 . The bad inspection unit 150 may be implemented as a computing device equipped with artificial intelligence having a CNN (Convolution Neural Network) neural network. The defect inspection unit 150 inspects the defect of the atypical product through model generation through deep learning image learning using the CNN neural network. Deep learning image learning using the CNN neural network may be implemented through a sample image input step, a training data labeling step, a deep learning learning progress step, and a model generation step. The defect inspection unit 150 may compare the images received from the plurality of cameras 120 with the model generated through the deep learning to inspect the appearance defect of the atypical product.

It is difficult to perform external inspection of atypical products with computer vision (Rule-based), which is the existing inspection method. Existing computer vision is suitable for inspection of the appearance of flat and smooth products. However, in the texture of the material of the inspection surface, which has an atypical appearance of a curved surface, it is difficult to distinguish between the defective part and the inspection surface, so it is not possible to inspect for poor appearance. For the remaining defects, the inspector had no choice but to visually inspect them.

The image applied to the external inspection of atypical metal products is an image that is difficult to detect defects through simple neural networks or shallow layer learning. In the present invention, learning about image defects is performed using a CNN neural network using Tensorflow. In this case, since it is difficult to detect a defect in an image, the depth of the neural network is applied as 101 layers to increase the amount of learning and optimize the defect detection.

Since the algorithm and model were developed based on Python, it may take a long time to configure the inspection device because the image acquisition application and the Python deep learning inspection model must communicate with each other to exchange images and statuses when applying the actual inspection. In the present invention, the required time can be shortened and the configuration can be simplified by making the 8 inspection models dll and making the function call directly from the image acquisition application so that the Python application does not run. To increase inspection reliability, each defect (scratches, dents, burrs, pressed, etc.) is labeled in more than 1,000 products with defective metal, and tens of thousands of data sets are generated and inspected by amplifying them by angle and size. accuracy can be increased.

Because it is necessary to inspect the entire surface of an atypical product, learning is carried out with 8 inspection models, and high-speed inspection speed is required to set the tact time of the inspection device, so single-thread sequential inspection It can be implemented by avoiding the method, applying the inspection Boost Algorithm, and performing the inspection with 8 Multi Threads to complete the inspection in 6 seconds for 88 images. Inspection by deep learning learning is inspected through the shape of the defect. there is.

Noise can be removed from the original image, followed by filtering (LOG) and binarization algorithm processing (Adaptive). After finding an approximate circle with the Circle Fitting algorithm, separating the region of interest based on the circle, and dividing the separated images by 1/4, deep learning learning can be performed, which is to test with one model. Finally, by checking the reliability of the inspection, it is possible to inspect the defects that occur in the appearance of atypical products more accurately and quickly.

As described above, although the present invention has been described with reference to limited embodiments and drawings, the present invention is not limited to the above embodiments, and various modifications and variations from these descriptions are provided by those skilled in the art to which the present invention pertains. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined by the following claims as well as the claims and equivalents.

100: inspection device
110: swivel mount
120: camera
130: lighting
140: rotation shooting control unit
150: defective inspection unit

Claims (5)

In the device for inspecting the defect of the product having an atypical appearance,
a rotating mount on which the atypical product is mounted and rotating the atypical product according to a selected algorithm;
a plurality of cameras disposed in each area in a form surrounding the atypical product mounted on the rotary mount and photographing the atypical product in each area in response to a rotation algorithm of the rotary mount;
a plurality of lights disposed corresponding to the respective cameras in each area in which the plurality of cameras are disposed and irradiating light to the atypical product;
a rotation photographing control unit for controlling the rotation of the rotation mount according to the algorithm of the rotation mount and correspondingly controlling the photographing of each of the plurality of cameras; and
A defect inspection unit that detects defects in the atypical product through images taken by the plurality of cameras
Device for inspecting defects of products having an atypical appearance, characterized in that it comprises a. According to claim 1,
The rotation imaging control unit is configured to perform a routine in which the rotation mount is stopped for a certain period of time after rotating by a rotation angle of 36 degrees according to the algorithm 10 times continuously from 0 degrees to a rotation angle of 360 degrees according to the algorithm. A device for inspecting defects of products having an atypical appearance, characterized in that the rotation is controlled. 3. The method of claim 2,
The rotation photographing control unit controls the plurality of cameras to photograph the atypical product at each location during the time when the rotation is stopped for each routine, respectively, and to obtain 10 images for each of the cameras during the 10 routines. A device for inspecting defects of products having an atypical appearance, characterized in that. According to claim 1,
The defect inspection unit is a computing device equipped with artificial intelligence having a CNN (Convolution Neural Network) neural network, and inspects the defects of the atypical product through model generation through deep learning image learning using the CNN neural network. A device that inspects defects of products with atypical appearance. 5. The method of claim 4,
Deep learning image learning using the CNN neural network is implemented through a sample image input step, a learning data labeling step, a deep learning learning progress step, and a model generation step, and the defect inspection unit receives the images received from the plurality of cameras in the deep learning An apparatus for inspecting defects of a product having an atypical appearance, characterized in that the inspection of the appearance defect of the atypical product is compared with a model generated through the .

Images