KR20220086851A - An automatic detection method of defective tires using internal air measurement part and deep running part - Google Patents

Abstract

The present invention is a step of (a) putting the tire into the transfer conveyor unit, (b) generating a measurement image in which the internal air measuring equipment measures the internal air of the tire transferred from the transfer conveyor unit, (c) the deep learning unit is measured Applying the image to the pre-trained deep learning main model to automatically determine whether it is suitable or unsuitable for the measurement image, (d) If it is less than the set judgment standard, determining that the measurement image is suitable, and (e) the discharge roller discharging the tire. When the determination result is satisfactory, the deep learning unit determines that the tire is suitable and provides an automatic detection method of a bad tire using an internal air measuring facility and a deep learning unit, characterized in that the step (e) is performed.

Description

An automatic detection method of defective tires using internal air measurement part and deep running part}

The present invention relates to an automatic detection method of a defective tire using an internal air measurement facility and a deep learning unit, and more particularly, to determine whether a tire is unsuitable by automatically detecting the presence of an air pocket inside the tire It relates to an automatic detection method of defective tires using internal air measuring equipment and deep learning unit.

In general, in order to determine whether the tire is suitable, the internal air of the tire is measured. Here, the measurement of the internal air of a tire is a manual system in which an operator directly inspects the tire while visually inspecting it, and the measurement method is shown in FIG. 1 .

Referring to FIG. 1 , first, a tire is put in, an internal air measuring device measures the internal air of the tire, and then an image as a measurement result is acquired. At this time, the operator reads based on the acquired image displayed on the UI to determine whether it is appropriate or not, and manually stores the judgment result in the computer.

At this time, the operator visually checks the non-conformity part with respect to the acquired image, and then determines whether the tire is unsuitable. Thereafter, the tire that has undergone the above process is discharged to the outside.

However, the system for measuring the internal air of a tire has a problem in that the reliability of the reading result is lowered as there is a human deviation in the reading of the tire depending on the know-how and condition of the operator.

In addition, the above-described internal air measurement system has a problem in that, when the number of tires to be inspected is large, the reliability of the reading result is further deteriorated along with an increase in work fatigue.

(Patent Document 1) Registered Patent Publication No. 10-0444514 (2009.05.08.)

(Patent Document 2) Unexamined Patent Publication No. 10-2002-0074672 (October 4, 2002)

It is an object of the present invention to solve the above problems by applying a measurement image that measures whether an air pocket exists inside a tire to a pre-trained deep learning main model to automatically determine whether or not it is suitable or unsuitable for internal air It is to provide an automatic detection method of defective tires using measuring equipment and deep learning unit.

In addition, an object of the present invention to solve the above problem is to apply the measurement image that has been determined to be unsuitable to the pre-trained deep learning sub-model to measure the coordinates and area of the unsuitable area of the unsuitable measurement image, and then to the display unit. It is to provide an automatic detection method for defective tires using an internal air measurement facility and deep learning unit that supports workers to efficiently determine the suitability and non-conformity of tires by transmitting them.

In addition, an object of the present invention to solve the above problem is to improve the reliability of the tire reading result by cross-checking the tire by visually re-inspecting the tire by an operator when the threshold of the measurement image is greater than or equal to a predetermined criterion value. It is to provide an automatic detection method for defective tires using internal air measuring equipment and deep learning unit.

The technical problems to be achieved by the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned can be clearly understood by those of ordinary skill in the art to which the present invention belongs from the description below. There will be.

The configuration of the present invention for achieving the above object comprises the steps of: (a) putting a tire into a transfer conveyor; (b) generating, by an internal air measuring device, a measurement image measuring the internal air of the tire transported from the transport conveyor; (c) automatically determining whether the measurement image is suitable or unsuitable for the measurement image by applying the measurement image to the deep learning main model by the deep learning unit; (d) determining that the measurement image is suitable when the result value for the measurement image applied to the deep learning main model learned in advance by the deep learning unit is less than a predetermined reference value; and (e) discharging the tire by the discharging roller unit; and, in step (d), when the determination result for the measurement image applied to the pre-trained deep learning main model is suitable, the deep learning It provides an automatic detection method of a defective tire using an internal air measurement facility and a deep learning unit, characterized in that the additional determines that the tire is suitable and performs the step (e).

In an embodiment of the present invention, the step (c) comprises the steps of: (c1) receiving the measurement image transmitted from the internal air measurement facility by the deep learning unit and extracting non-conformity information; and (c2) pre-processing the measurement image by the deep learning unit.

In an embodiment of the present invention, the step (c1) comprises: receiving the measurement image transmitted from the deep learning unit the internal air measurement facility; and extracting, by the deep learning unit, the inappropriate information from the measurement image.

In an embodiment of the present invention, the step (c2) includes: handling, by the deep learning unit, nonconforming information that is an unstructured data set as a structured data set; detecting the tire area by the deep learning unit; and emphasizing the handled inappropriate information by the deep learning unit.

In an embodiment of the present invention, the step (c) comprises the steps of: (c3) generating the pre-trained deep learning main model that the deep learning unit has learned deep learning by an operator based on the highlighted non-conformity information; and (c4) automatically detecting whether the tire is non-conforming by applying the measurement image to the pre-trained deep learning main model by the deep learning unit and automatically determining whether the measurement image is suitable or not; It may be characterized in that it further comprises.

In an embodiment of the present invention, the step (c) includes: (c5) determining that the tire is unsuitable when the determination result for the measurement image applied to the pre-trained deep learning main model is unsuitable; and (c6) applying the inappropriate measurement image to the deep learning sub-model trained in advance by the deep learning unit to measure the coordinates and area of the inappropriate area of the inappropriate measurement image; It may further include .

In an embodiment of the present invention, the step (c) comprises: (c7) outputting, by the display unit, the coordinates and area of the inappropriate measurement image and the inappropriate area of the inappropriate measurement image transmitted from the deep learning unit; and (c8) re-inspecting the tire by the operator.

The effect of the present invention according to the configuration as described above is to apply the measurement image that measures whether or not there is an air pocket inside the tire to the pre-trained deep learning main model to automatically determine whether it is suitable or not, so that the operator's work It can reduce fatigue.

In addition, the effect of the present invention according to the configuration as described above is to apply the measurement image, which has been determined to be inappropriate, to the pre-trained deep learning sub-model, measure the coordinates and area of the inappropriate area of the inappropriate measurement image, and then transmit it to the display unit This allows the operator to efficiently determine whether the tire is suitable or not.

In addition, the effect of the present invention according to the above configuration is that, when the threshold of the measurement image is greater than or equal to the preset judgment standard, the operator visually re-inspects the tire and cross-checks it, thereby improving the reliability of the tire reading result. have.

It should be understood that the effects of the present invention are not limited to the above-described effects, and include all effects that can be inferred from the configuration of the invention described in the detailed description or claims of the present invention.

1 is a flowchart illustrating a method for manually measuring internal air of a tire according to the related art.
2 is a block diagram illustrating an automatic detection system for defective tires using an internal air measurement facility and a deep learning unit according to an embodiment of the present invention.
3 is a block diagram illustrating an operation flow of an automatic detection system for defective tires using an internal air measurement facility and a deep learning unit according to an embodiment of the present invention.
4 is a flowchart illustrating an automatic detection method of a defective tire using an internal air measuring device and a deep learning unit according to an embodiment of the present invention.
5 is a schematic diagram showing that the deep learning unit applies the measurement image to the deep learning model in the automatic detection method of a bad tire using the internal air measuring equipment and the deep learning unit according to an embodiment of the present invention.
6 is a view showing a process of pre-processing a measurement image in an automatic detection method of a bad tire using an internal air measuring facility and a deep learning unit according to an embodiment of the present invention.
7 is a view showing a detailed algorithm for a deep learning model performed by the deep learning unit in the automatic detection method of a bad tire using the internal air measuring facility and the deep learning unit according to an embodiment of the present invention.
8 is a threshold value that is a result of applying the measurement image to the deep learning model by the deep learning unit in the automatic detection method of a bad tire using the internal air measuring facility and the deep learning unit according to an embodiment of the present invention. It is a drawing.
9 is a graph showing a tire non-conformity detection rate by an automatic detection method of a defective tire using an internal air measurement facility and a deep learning unit according to an embodiment of the present invention.
10 is a diagram illustrating an algorithm for applying a measurement image determined to be unsuitable in an automatic detection method of a bad tire using an internal air measurement facility and a deep learning unit to a sub-model according to an embodiment of the present invention.
11 is a view showing the results of tire fit and non-conformity according to the automatic detection method of defective tires using the internal air measuring equipment and the deep learning unit according to an embodiment of the present invention.

Hereinafter, the present invention will be described with reference to the accompanying drawings. However, the present invention may be embodied in several different forms, and thus is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a part is said to be “connected (connected, contacted, coupled)” with another part, it is not only “directly connected” but also “indirectly connected” with another member interposed therebetween. "Including cases where In addition, when a part "includes" a certain component, this means that other components may be further provided without excluding other components unless otherwise stated.

The terminology used herein is used only to describe specific embodiments, and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present specification, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

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

1. Automatic detection system for defective tires using internal air measuring equipment and deep learning unit

Hereinafter, with reference to FIGS. 2 and 3, the internal air measurement facility and the automatic detection of defective tires using the deep learning unit for explaining the automatic detection method of the defective tire using the internal air measuring facility and the deep learning unit according to an embodiment of the present invention The detection system will be described first.

2 is a block diagram illustrating an automatic detection system for defective tires using an internal air measurement facility and a deep learning unit according to an embodiment of the present invention. 3 is a block diagram illustrating an operation flow of an automatic detection system for defective tires using an internal air measurement facility and a deep learning unit according to an embodiment of the present invention.

The automatic detection system for defective tires using the internal air measuring equipment and the deep learning unit according to an embodiment of the present invention includes the conveying conveyor unit 110 , the internal air measuring unit 120 , the deep learning unit 130 , and the display unit 140 . ) and a discharge roller unit 150 .

The transfer conveyor 110 supports the input tire. In addition, the transfer conveyor 110 transfers the tire to the internal air measuring device 120 while rotating.

The internal air measuring device 120 measures whether an air pocket exists inside the tire transported by the transport conveyor 110 . For this purpose, the internal air measuring device 120 may measure the presence of air in the tire with a laser sensor while maintaining a vacuum state in the head part.

The deep learning unit 130 applies the measurement image measured for the presence or absence of the internal air of the tire measured by the internal air measurement facility 120 to the pre-learned deep learning main model to determine whether the measurement image is suitable or not.

Here, the pre-trained deep learning main model is established by the worker applying images that become a number of samples, and then when the measurement image is applied to the pre-trained deep learning main model, the pre-trained deep learning main model is automatically measured It automatically judges whether the image is suitable or not.

At this time, the pre-trained deep learning main model has a measurement image (pixel), inspection date and time, tire pattern name, tire width, tire aspect ratio, tire outer diameter, tire class, and calibration magnification (pixels) as input factors. ) and mm conversion ratio) are applied.

In addition, the deep learning unit 130 may set a learning set value (hyperparameter) for adjusting the pre-trained deep learning main model as needed, in the present invention stride = 5, threshold = 0.5, max_epoch = 100, batch_size =8, max patience for early stopping=15, inner side trim = -3px, outer side trim = -0.5px.

The deep learning unit 130 applies the measurement image to the pre-learned deep learning main model that reflects the learning setting value, and if the measurement image is greater than or equal to a predetermined reference value, it is determined that the measurement image is inappropriate, and the measurement image is set If it is less than the judgment standard, the measured image is judged to be suitable.

Here, the preset determination reference value is a threshold boundary value.

If the deep learning unit 130 determines that the measurement image is inappropriate for the pre-trained deep learning main model, the unsuitable measurement image is applied to the pre-trained deep learning sub-model to match the coordinates and area for the inappropriate area of the unsuitable measurement image. It can be marked on the measurement image.

The deep learning unit 130 transmits a result of whether the measurement image is suitable or not suitable to the display unit 140 .

The display unit 140 outputs a measurement image related to the internal air of the tire measured by the internal air measuring device 120 .

The discharge roller unit 150 is a device for discharging the measured tire to the outside, and a plurality of rollers are positioned to be spaced apart from each other and rotate while discharging the measured tire to the outside.

2. Automatic detection of defective tires using internal air measuring equipment and deep learning unit

Hereinafter, a method for automatically detecting a defective tire using an internal air measuring device and a deep learning unit according to an embodiment of the present invention will be described with reference to FIGS. 4 to 11 .

4 is a flowchart illustrating an automatic detection method of a bad tire using an internal air measuring facility and a deep learning unit according to an embodiment of the present invention.

Referring to FIG. 4 , the method for automatically detecting defective tires using an internal air measuring device and a deep learning unit according to an embodiment of the present invention includes (a) putting a tire into the transfer conveyor unit 110 (S100), ( b) generating a measurement image in which the internal air measurement facility 120 measures the internal air of the tire transferred from the transfer conveyor unit 110 (S200), (c) the deep learning unit 130 pre-previews the measurement image Step (S300) of automatically determining whether the measurement image is suitable or unsuitable by applying it to the learned deep learning main model (S300), (d) the deep learning unit 130 for the measurement image applied to the pre-trained deep learning main model When the result value is less than the preset judgment standard, determining that the measurement image is suitable (S400) and (e) the discharge roller unit 150 discharging the tire (S500), and in the step (d) , when the determination result for the measurement image applied to the pre-trained deep learning main model is suitable, the deep learning unit 130 determines that the tire is suitable and performs step (e).

First, in step (a), when the tire is transferred to the transfer conveyor unit 110 , the transfer conveyor unit 110 rotates and transfers the tire to the internal air measuring device 120 .

Next, in step (b), the internal air measuring device 120 measures whether an air pocket exists inside the tire transported by the transport conveyor 110 . For this purpose, the internal air measuring device 120 may measure the presence of air in the tire with a laser sensor while maintaining a vacuum state in the head part.

5 is a schematic diagram showing that the deep learning unit applies the measurement image to the deep learning model in the automatic detection method of a bad tire using the internal air measuring equipment and the deep learning unit according to an embodiment of the present invention.

Next, the step (c) is, (c1) the deep learning unit 130 receives the measurement image transmitted from the internal air measurement facility 120 and extracts non-conforming information, and (c2) the deep learning unit 130 includes the step of pre-processing the measurement image.

Specifically, the step (c1) includes the step of the deep learning unit 130 receiving the measurement image transmitted from the internal air measurement facility 120 and the deep learning unit 130 extracting non-conforming information from the measurement image. include

Referring to FIG. 5 , as shown in the upper part of FIG. 5 , the deep learning unit 130 receives the measured image and then converts the measured image into pixel values. For example, the numbers shown in FIG. 5 may represent business cards for each pixel position of the measurement image.

Accordingly, the deep learning unit 130 extracts non-conformity information from the measurement image measured by the internal air measurement facility 120, extracts features from the non-conformity information (S420), and determines conformity/nonconformity by type based on the extracted characteristics. Classify and define For example, the deep learning unit 130 may define the non-conforming image in the form of C-type, H-type, and F-type according to the shape.

6 is a view showing a process of pre-processing a measurement image in an automatic detection method of a bad tire using an internal air measuring facility and a deep learning unit according to an embodiment of the present invention.

Next, referring to FIG. 6 , in step (c2), the deep learning unit 130 handles the unsuitable information that is an unstructured dataset as a structured dataset, and the deep learning unit 130 detects the tire area. and emphasizing the handled inappropriate information by the deep learning unit 130 .

Step (c2) describes the process of pre-processing the measurement image in order to apply the measurement image to the pre-trained main model.

7 is a view showing a detailed algorithm for a deep learning model performed by the deep learning unit in the automatic detection method of a bad tire using the internal air measuring equipment and the deep learning unit according to an embodiment of the present invention.

Next, after step (c2), step (c) is, (c3) deep learning unit 130 based on the highlighted non-conformity information to generate a pre-trained deep learning main model learned by deep learning by an operator Step and (c4) the deep learning unit 130 applies the measurement image to the pre-trained deep learning main model to automatically determine whether the measurement image is suitable or unsuitable to automatically detect whether the tire is non-conforming. include more

Specifically, in step (c3), an operator creates a pre-trained deep learning main model using a plurality of images as samples. Additionally, the deep learning unit 130 may define the non-conforming image in the form of C-type, H-type, and F-type according to the shape.

Subsequently, the deep learning unit 130 applies the measurement image to the pre-trained deep learning model to determine whether the measurement image is suitable or unsuitable for the measurement image according to the synchronization rate or matching rate with the non-conforming image, and the explanation thereof will be described later in detail together with the threshold.

Next, in the step (c4), a detailed algorithm of the deep learning main model trained in advance is applied as shown in FIG. 7 , and the algorithm shown in FIG. 7 exemplarily may be the Densenet-121 algorithm.

8 is a threshold value, which is a result value of the deep learning unit applying the measurement image to the deep learning model in the automatic detection method of a bad tire using the internal air measuring equipment and the deep learning unit according to an embodiment of the present invention. It is a drawing.

In addition, the present invention further comprises the step of calculating a threshold, which is a result value of applying the measurement image to the deep learning main model, previously learned by the deep learning unit 130, between the steps (c3) and (c4). It may be included, and in this regard, it is illustrated in detail in FIG. 8 .

Here, the deep learning unit 130 compares the calculated threshold with a preset threshold that is a preset judgment reference value to determine whether the measured image is suitable or not.

In this case, the preset criterion value is 0.5, and it may be understood that the synchronization rate or matching rate between the measured image and the non-conforming image is 50%.

9 is a graph showing a tire non-conformity detection rate by an automatic detection method of a defective tire using an internal air measurement facility and a deep learning unit according to an embodiment of the present invention.

However, as shown in FIG. 9 showing the threshold according to the measurement image, the non-conformity detection rate of the measurement image is 1.66%.

The deep learning unit 130 determines that the measurement image is inappropriate when the threshold for the measurement image is 0.5 or more, which is a preset judgment reference value.

On the other hand, the deep learning unit 130 determines that the measurement image is suitable when the threshold for the measurement image is less than 0.5, which is a preset determination reference value.

10 is a diagram illustrating an algorithm for applying to a sub-model a measurement image determined to be unsuitable in an automatic detection method of a bad tire using an internal air measurement facility and a deep learning unit according to an embodiment of the present invention.

Next, after step (c4), step (c) is performed, (c5) if the determination result for the measurement image applied to the pre-trained deep learning main model is inappropriate, the deep learning unit 130 determines that the tire is unsuitable. The step of determining and (c6) the deep learning unit 130 applies an inappropriate measurement image to the pre-trained deep learning submodel to measure the coordinates and area of the inappropriate area of the inappropriate measurement image.

The deep learning unit 130 may apply the algorithm shown in FIG. 10 in order to apply the inappropriate measurement image to the pre-trained deep learning submodel, and the algorithm shown in FIG. 10 may be exemplarily the Yolo algorithm.

Accordingly, in the pre-trained deep learning submodel, the coordinates and area of the inappropriate area of the inappropriate measurement image are marked on the measurement image and stored.

11 is a view showing the results of tire fit and non-conformity according to the automatic detection method of defective tires using the internal air measuring equipment and the deep learning unit according to an embodiment of the present invention.

Next, after the (c6) step, the (c) step is, (c7) the coordinates and the area of the inappropriate measurement image transmitted from the deep learning unit 130 by the display unit 140 and the inappropriate area of the inappropriate measurement image The step of outputting (S410) and (c8) further includes a step (S420) of re-inspecting the tire by the operator.

In the step (c7), the display unit 140 outputs a measurement image related to the internal air of the tire measured by the internal air measuring device 120, and the related information is shown in FIG.

In addition, the display unit 140 also displays the result according to the suitability/nonconformity determination of the tire in the deep learning unit again. Accordingly, the operator checks the condition of the tire through the display unit 140 and re-inspects the condition of the tire by performing a visual inspection again.

The above description of the present invention is for illustration, and those of ordinary skill in the art to which the present invention pertains can understand that it can be easily modified into other specific forms without changing the technical spirit or essential features of the present invention. will be. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a dispersed form, and likewise components described as distributed may be implemented in a combined form.

The scope of the present invention is indicated by the following claims, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included in the scope of the present invention.

110: transfer conveyor unit
120: internal air measuring equipment
130: deep learning unit
140: display unit
150: discharge roller unit

Claims (7)

(a) inputting the tire into the transfer conveyor unit;
(b) generating, by an internal air measuring device, a measurement image measuring the internal air of the tire transported from the transport conveyor;
(c) applying the measurement image to the deep learning main model trained in advance by the deep learning unit to automatically determine whether the measurement image is suitable or not suitable;
(d) determining that the measurement image is suitable when the result value for the measurement image applied to the deep learning main model learned in advance by the deep learning unit is less than a predetermined reference value; and
(e) discharging the tire by the discharge roller unit;
In the step (d), if the determination result for the measurement image applied to the pre-trained deep learning main model is suitable, the deep learning unit determines that the tire is suitable and performs the step (e), characterized in that A method for automatic detection of defective tires using internal air measuring equipment and deep learning unit.
According to claim 1,
Step (c) is,
(c1) the deep learning unit receiving the measurement image transmitted from the internal air measurement equipment and extracting non-conformity information; and
(c2) the deep learning unit pre-processing the measurement image; an automatic detection method of bad tires using an internal air measuring facility and a deep learning unit, characterized in that it comprises.
3. The method of claim 2,
The step (c1) is,
Receiving the measurement image transmitted from the deep learning unit the internal air measurement equipment; and
The deep learning unit extracting the non-conformity information from the measurement image; Automatic detection method of defective tires using the internal air measuring equipment and deep learning unit, characterized in that it comprises a.
3. The method of claim 2,
The step (c2) is,
handling, by the deep learning unit, nonconformity information, which is an unstructured dataset, as a structured dataset;
detecting the tire area by the deep learning unit; and
The deep learning unit emphasizing the handled nonconformity information; automatic detection method of bad tires using the internal air measuring equipment and deep learning unit, characterized in that it comprises.
5. The method of claim 4,
Step (c) is,
(c3) generating, by the deep learning unit, the pre-trained deep learning main model, which is deep learning learned by an operator based on the highlighted non-conformity information; and
(c4) applying the measurement image to the pre-trained deep learning main model by the deep learning unit to automatically determine whether the measurement image is suitable or unsuitable to automatically detect whether the tire is inappropriate; Automatic detection method of bad tires using the internal air measuring equipment and deep learning unit, characterized in that it further comprises.
6. The method of claim 5,
Step (c) is,
(c5) determining that the tire is unsuitable when the result of the determination on the measurement image applied to the pre-trained deep learning main model is unsuitable; and
(c6) applying the inappropriate measurement image to the deep learning sub-model trained in advance by the deep learning unit to measure the coordinates and area of the inappropriate area of the inappropriate measurement image; A method for automatic detection of defective tires using equipment and deep learning unit.
7. The method of claim 6,
Step (c) is,
(c7) outputting, by the display unit, the coordinates and areas of the inappropriate measurement image and the inappropriate area of the inappropriate measurement image transmitted from the deep learning unit; and
(c8) the step of re-inspecting the tire by the operator; the automatic detection method of a defective tire using an internal air measurement facility and a deep learning unit, characterized in that it further comprises.

Images