KR102578920B1 - Apparatus for PET sorting based on artificial intelligence - Google Patents

Abstract

The present invention relates to a PET screening device based on artificial intelligence.
The PET sorting device according to the present invention includes a learning unit that trains a pre-built learning model that learns to select the color of plastic waste included in a waste image of plastic waste, and a vision sensor to capture a plurality of plastic wastes input to the conveyor. An image acquisition unit that acquires a waste image, an analysis unit that inputs the acquired waste image into a learned model to extract location information on plastic waste of a preset specific color, and an analysis unit that extracts location information about plastic waste of a preset specific color. A spectrum acquisition unit that sets a light irradiation area using location information, transmits the set light irradiation area to a spectral sensor, and then receives the spectrum of plastic waste located in the light irradiation area from the spectral sensor, the acquired spectrum a determination unit that determines whether the plastic waste located in the light irradiation area corresponds to PET, and a control unit that sets the location information on the plastic waste determined to be PET as the final sorting coordinates and transmits it to the pickup robot. Includes.
According to the present invention, by providing a total platform that can automatically sort and separate recycled PET, the sorting process in the recycling field can be automated and the sorting accuracy and sorting speed can be improved.

Description

PET sorting device based on artificial intelligence {Apparatus for PET sorting based on artificial intelligence}

The present invention relates to a PET sorting device based on artificial intelligence. To be described in more detail, a PET sorting device that uses an artificial intelligence algorithm to select the material of PET (polyethylene terephthalate) to be recycled and collects only PET of a specified color. It's about devices.

As the COVID-19 crisis continues and non-face-to-face consumption such as courier and food delivery increases, a lot of recyclable waste such as plastic is being generated. Typically, 30-40% of the amount of recycled products is not sorted and is landfilled or incinerated. When plastic waste is landfilled, it accumulates in the human body in the form of microplastics through various routes, which can have a negative impact on health.

Therefore, the importance of the selection process for plastic waste is emerging. However, since plastic waste is used in a variety of materials depending on its purpose and form, there is a problem in that it is difficult to distinguish between materials and cannot be sorted in the manual process.

In addition, in the case of PET with added color, the quality of the recycled raw material may deteriorate, so consumers must separate and dispose of it, but in most cases, they are not aware of this accurately. Therefore, since plastic waste is manually sorted according to type or color, most sorting companies have the problem of increasing time and labor costs due to sorting, and they only select items that are certain to be sorted, so they are not sorted. Plastic waste is landfilled and cannot improve the environmental pollution problem.

The technology behind the present invention is disclosed in Korean Patent No. 10-1270354 (announced on May 27, 2013).

According to the present invention, the purpose of the present invention is to provide a PET sorting device that selects the material of PET (polyethylene terephthalate) to be recycled using an artificial intelligence algorithm and collects only PET of a specified color.

In order to achieve this technical task, the PET sorting device based on artificial intelligence according to an embodiment of the present invention selects the color of plastic waste included in a waste image taken of plastic waste and learns to select atypical plastic. A learning unit that trains a pre-built learning model, an image acquisition unit that acquires waste images of a plurality of plastic wastes put into the conveyor from the vision sensor, and inputs the acquired waste images into a learning model for which learning has been completed. An analysis unit that extracts location information about plastic waste of a set specific color, sets a light irradiation area using the extracted location information of plastic of a specific color, and transmits the set light irradiation area to the spectral sensor, A spectrum acquisition unit that receives the spectrum of the plastic waste located in the light irradiation area from a spectral sensor, a determination unit that uses the acquired spectrum to determine whether the plastic waste located in the light irradiation area corresponds to PET, and It may include a control unit that sets the location information on the plastic waste determined to be PET as the final sorting coordinates and transmits it to the pickup robot.

The learning unit builds a learning model based on the HSV Color algorithm and Dection, and calculates each HSV (Hue, Saturation, and Value) value measured for each of the plurality of plastic wastes included in the waste image. A pre-built learning model can be trained using the position value corresponding to the HSV value.

The analysis unit may input the waste image into a learning model that has completed learning, and extract a region of interest for plastic waste with a specific color and a position value for the center point of the region of interest through the learning model.

The analysis unit sets the starting point of the conveyor as a reference coordinate, sets the position value of the center point of the area of interest as a relative coordinate value using the conveyor width, moving speed, and the reference coordinate value, and uses the set relative coordinate value as a relative coordinate value. Coordinates on the conveyor can be obtained by applying geometry transformation.

The determination unit sets each spectrum range according to the type of the plastic waste, and if each spectrum for the plastic waste located within the coordinate value on the conveyor is located in the spectrum range corresponding to PET, the corresponding plastic waste is selected. It can be judged by PET.

The control unit may obtain a robot pickup coordinate value by applying a transformation matrix to the coordinate value on the conveyor, and extract the obtained robot pickup coordinate value as the final selection coordinate.

The control unit can use the final sorting coordinate value of the plastic waste determined to be PET to have the robot pick up plastic waste made of a material other than PET.

According to the present invention, by providing a total platform that can automatically sort and separate recycled PET, the sorting process for recycling can be automated and the sorting accuracy and sorting speed can be improved.

Additionally, according to the present invention, only plastic waste with a specific color can be quickly selected from a plurality of plastic wastes using the HSV Color algorithm and a learning model based on Dection.

1 is a diagram for explaining an automatic selection and separation system according to an embodiment of the present invention.
FIG. 2 is a configuration diagram for explaining the PET sorting device shown in FIG. 1.
Figure 3 is a flowchart illustrating a method of sorting PET using an automatic sorting and separation system for recycled PET according to an embodiment of the present invention.
Figure 4 is a drawing of plastic waste.
Figure 5 is an example diagram for explaining step S340 shown in Figure 2.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings. In this process, the thickness of lines or sizes of components shown in the drawing may be exaggerated for clarity and convenience of explanation.

Additionally, the terms described below are terms defined in consideration of functions in the present invention, and may vary depending on the intention or custom of the user or operator. Therefore, definitions of these terms should be made based on the content throughout this specification.

Hereinafter, an automatic sorting and separation system for recycled PET according to an embodiment of the present invention will be described in more detail using FIGS. 1 and 2.

1 is a diagram for explaining an automatic selection and separation system according to an embodiment of the present invention.

As shown in Figure 1, the automatic sorting and separation system according to an embodiment of the present invention includes a conveyor 100, a vision sensor 200, a spectral sensor 300, a PET sorting device 400, and a sorting robot 500. Includes.

First, the conveyor 100 serves to transport the input plastic waste from one point to another.

The vision sensor 200 is located on one side of the conveyor 100 and photographs the transported plastic waste. Then, the vision sensor 200 transmits the captured waste image to the PET sorting device 400, which will be described later.

The spectral sensor 300 radiates light into a set light irradiation area and measures the spectrum reflected from plastic waste located within the light irradiation area.

To explain this again, the spectral sensor 300 receives the light irradiation area from the PET sorting device 400. Then, the spectral sensor 300 irradiates light within the set light irradiation area. For example, assuming that the light irradiation area is approximately 950 nm to 1650 nm, the spectral sensor 300 measures the spectrum within approximately 950 nm to 1650 nm. Then, the measured spectrum is transmitted to the PET screening device 400.

The PET sorting device 400 inputs the waste image received from the vision sensor 200 into a fully trained learning model to obtain location information on plastic waste that has a specific color and is presumed to be PET. Then, the PET sorting device 400 sets a light irradiation area using the acquired location information and transmits information about the set light irradiation area to the spectral sensor 300.

Next, the PET sorting device 400 receives the spectrum of the plastic waste located in the light irradiation area from the spectral sensor 300, and uses the received spectrum to determine whether the corresponding plastic waste is PET.

Lastly, the sorting robot 500 is located on one side of the conveyor 100 and picks up plastic waste excluding PET. To elaborate, the sorting robot 500 receives location information about PET from the PET sorting device 400. Then, the sorting robot 500 uses the received location information to pick up and separate only plastic waste other than PET.

FIG. 2 is a configuration diagram for explaining the PET sorting device 400 shown in FIG. 1.

As shown in Figure 2, the PET screening device 400 according to an embodiment of the present invention includes an image acquisition unit 410, a learning unit 420, an analysis unit 430, a spectrum acquisition unit 440, and a determination unit. It includes 450 and a control unit 460.

The image acquisition unit 410 acquires a waste image of plastic waste from the vision sensor 200.

The learning unit 420 builds a learning model based on the HSV Color algorithm. Then, the learning unit 420 inputs the waste image into the constructed learning model and trains the learning model to extract an area for plastic waste with a specific color and output a position value for the extracted area.

Here, the learning model consists of at least one algorithm among the CNN (Convolutional Neural Network) algorithm, SSD (Single Shot Detector) algorithm, and YOLO (You Only Look Once) algorithm.

Next, the analysis unit 430 inputs the waste image containing the selection target into the learning model on which learning has been completed. Then, the learning model selects plastic waste with a preset color from the waste image and outputs location information of the selected plastic waste. Here, the location information represents pixel coordinate values within the image.

The analysis unit 430 converts the output location information of the plastic waste into two-dimensional coordinate values on the conveyor 100.

The spectrum acquisition unit 440 sets the light irradiation area using two-dimensional coordinate values on the conveyor 100. And the set light irradiation area is transmitted to the spectral sensor 300. Then, the spectral sensor 300 radiates light into the received light irradiation area and measures the spectrum of light reflected from the plastic waste located within the light irradiation area. When the spectrum measurement is completed, the spectrum acquisition unit 440 receives the measured spectrum from the spectral sensor 300.

The determination unit 450 analyzes the material of the plastic waste using the received spectrum, and determines whether the plastic waste located in the light irradiation area corresponds to PET according to the analysis result.

Finally, the control unit 460 extracts location information on plastic waste not determined to be PET and transmits the extracted location information to the sorting robot 500.

Hereinafter, a method of sorting PET using an automatic sorting and separation system for recycled PET according to an embodiment of the present invention will be described in more detail using FIGS. 3 to 5.

Figure 3 is a flowchart illustrating a method of sorting PET using an automatic sorting and separation system for recycled PET according to an embodiment of the present invention.

As shown in FIG. 3, the PET screening method using the automatic selection and separation system according to an embodiment of the present invention is divided into a step of training a learning model and a step of selecting PET using the learned learning model.

In the step of training the learning model, first, the PET sorting device 400 collects waste images (S310).

Figure 4 is a drawing of plastic waste.

As shown in Figure 4, plastic waste is formed in various shapes, colors, and sizes depending on the material and purpose, and the plastic waste input to the conveyor 100 is formed in an irregular shape depending on the degree of compression. In addition, looking at the color of plastic waste when it is first received at the sorting workshop and after cleaning, plastic waste is discolored by the chemicals used during cleaning.

Therefore, the user takes pictures of the plastic waste that has been put into the sorting workshop, or the plastic waste that has been cleaned using chemicals, or the plastic waste that has been compressed, and the captured waste image is sent to the PET sorting device 400. Enter.

Next, the learning unit 420 inputs waste images into the learning model and trains it to select plastic waste having a specific color (S320).

To explain this again, the learning model distinguishes between the background color of the conveyor 100 and the color of plastic waste based on the HSV Color algorithm and Dection.

Therefore, the learning unit 420 learns the position value corresponding to the selected candidate area through each HSV (Hue, Saturation, Value) value measured for each of the plurality of plastic wastes included in the waste image. Input it into the model and conduct supervised learning.

Then, the learning model outputs the position value of the object extracted from the input waste image.

When learning of the learning model is completed through steps S310 and S320, the PET sorting device 400 according to an embodiment of the present invention selects PET of a specific color using the learned learning model.

First, the PET sorting device 400 receives a waste image of the plastic waste to be sorted from the vision sensor 200 (S330).

When plastic waste to be sorted is input into the conveyor 100, the vision sensor 200 performs photography according to a preset frame rate.

At this time, the vision sensor 200 is set to a frame rate calculated through Equation 1 below in order to obtain an image synchronized with the conveyor 100.

Here, Conveyer belt velocity represents the moving speed of the conveyor 100, and Field of view of camera represents the observable area of the vision sensor 200.

For example, assume that the maximum speed of the conveyor 100 is 1,000 mm/sec and the observable area of the vision sensor is 520x370 mm. And, by substituting the moving speed and observable area of the conveyor 100 into Equation 1, 4.8 [frames / sec] is calculated.

Then, the vision sensor photographs the plastic waste input into the conveyor 100 approximately every 5 [frames/sec] and transmits the captured waste image to the PET sorting device 400.

Then, the analysis unit 430 inputs the received waste image into the learned model to determine whether plastic waste exists in the waste image (S340)

To elaborate, the learning model determines whether plastic waste exists in the waste image by distinguishing between the color of the conveyor 100 and the color of the plastic waste.

Figure 5 is an example diagram for explaining step S340 shown in Figure 2.

As shown in Figure 5, when it is determined that plastic waste exists, the learning model extracts the area corresponding to the plastic waste as the area of interest.

Next, the analysis unit 430 obtains location information of plastic waste having a specific color from the learning model (S350).

The learning model determines whether the plastic waste included in the extracted region of interest corresponds to a specific color. To explain this again, the learning model determines whether the HSV value of the region of interest corresponds to a specific HSV value. Then, the learning model extracts a region of interest corresponding to a specific HSV value and outputs the coordinates of the extracted region of interest.

For example, assuming that the HSV values of transparent PET after cleaning are 44(H), 29(S), and 98(V), the analysis unit 430 measures 44(H), 29(S), and 98(V). A region of interest with a value of is extracted, and coordinate values of the extracted region of interest are extracted.

When step S350 is completed, the spectrum acquisition unit 440 acquires a spectrum for plastic waste having a specific color from the spectral sensor 300 (S360).

In detail, the analysis unit 440 sets the starting point of the conveyor 100 as a reference coordinate and obtains a relative coordinate value for the center point of the area of interest using the conveyor width, moving speed, and reference coordinate value. Here, the center point of the region of interest represents the coordinate value of the region of interest obtained in step S350.

Next, the analysis unit 440 obtains coordinate values on the conveyor by applying geometry transformation to the relative coordinate values.

The spectrum acquisition unit 440 sets the light irradiation area using the obtained coordinate values on the conveyor. Then, the spectrum acquisition unit 440 transmits the set light irradiation area to the spectral sensor 300.

Then, the spectral sensor 300 radiates light into the received light irradiation area and acquires a spectrum corresponding to the plastic waste.

When the spectrum acquired through step S360 is transmitted to the determination unit 450, the determination unit 450 compares the received spectrum with a preset spectrum range to determine whether the corresponding plastic waste corresponds to PET (S370) .

First, the determination unit 450 collects each spectrum range according to the type of plastic waste.

Then, the determination unit 450 compares the spectrum received from the spectrum acquisition unit 440 with the previously stored spectrum range. For example, when the PET sorting device 400 wants to sort transparent PET, the determination unit 450 compares the received spectrum with the spectrum range corresponding to transparent PET. And, if the received spectrum falls within the spectrum range, the determination unit 450 determines that the corresponding plastic waste is PET.

Next, the control unit 460 transmits the coordinate values for the plastic waste not determined to be PET to the sorting robot 500, and causes the sorting robot 500 to pick up the remaining plastic waste excluding PET of a specific color ( S380).

To elaborate, the control unit 460 converts the coordinate values on the conveyor where the plastic waste determined to be PET is located into pickup coordinate values.

Then, the control unit 460 sets the converted pickup coordinate value as the final sorting coordinate value and transmits the set final sorting coordinate value to the sorting robot 500. Then, the sorting robot 500 uses the received final sorting coordinates to pick up plastic waste of a material other than PET with a specific color. That is, only transparent PET is left on the conveyor 100 so that it can be transported to a specific location.

As such, the automatic sorting and separation system according to the present invention provides a total platform that can automatically sort and separate recycled PET, thereby automating the sorting process in the recycling field and improving sorting accuracy and sorting speed.

In addition, the automatic sorting and separation system according to the present invention can quickly select only plastic waste with a specific color from a plurality of plastic wastes using a learning model based on the HSV Color algorithm and Dection.

The present invention has been described with reference to the embodiments shown in the drawings, but these are merely illustrative, and those skilled in the art will understand that various modifications and other equivalent embodiments are possible therefrom. will be. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the patent claims below.

100: Conveyor
200: Vision sensor
300: Spectral sensor
400: PET sorting device
410: Image acquisition unit
420: Learning Department
430: analysis unit
440: Spectrum acquisition unit
450: Judgment unit
460: control unit
500: Pick-up robot

Claims (7)

In a PET screening device based on artificial intelligence,
A learning unit that selects the color of plastic waste included in a waste video of plastic waste and trains a pre-built learning model to learn to select atypical plastic;
An image acquisition unit that acquires waste images taken from a vision sensor of a plurality of plastic wastes put into the conveyor,
An analysis unit that inputs the obtained waste image into a fully trained learning model to extract location information about plastic waste of a preset specific color;
A light irradiation area is set using the extracted location information of the plastic of a specific color, the set light irradiation area is transmitted to the spectral sensor, and then the spectrum of plastic waste located in the light irradiation area is transmitted from the spectral sensor. receiving spectrum acquisition unit,
A determination unit that determines whether the plastic waste located in the light irradiation area corresponds to PET using the acquired spectrum, and
It includes a control unit that sets the location information on the plastic waste determined to be PET as final sorting coordinates and transmits it to the pickup robot,
The analysis unit,
An automatic sorting and separation system that inputs the waste image into a fully trained learning model and extracts a region of interest for plastic waste with a specific color and a position value for the center point of the region of interest through the learning model. According to paragraph 1,
The learning department,
Build a learning model based on the HSV Color algorithm and Decision,
Automatically learns a pre-built learning model using each HSV (Hue, Saturation, Value) value measured for each of the multiple plastic wastes included in the waste image and the position value corresponding to the HSV value. Selective separation system. delete According to paragraph 1,
The analysis unit,
Set the starting point of the conveyor as a reference coordinate value, and set the position value for the center point of the area of interest as a relative coordinate value using the conveyor width, moving speed, and the reference coordinate value,
An automatic sorting and separation system that obtains coordinate values on a conveyor by applying geometry transformation to the set relative coordinate values. According to paragraph 4,
The judgment department,
Set each spectral range according to the type of plastic waste,
An automatic sorting and separation system that determines that the corresponding plastic waste is PET when each spectrum of the plastic waste located within the coordinate values on the conveyor is located in the spectrum range corresponding to PET. According to clause 5,
The control unit,
An automatic sorting and separation system that obtains robot pickup coordinates by applying a transformation matrix to the coordinate values on the conveyor and extracts the obtained robot pickup coordinates as final sorting coordinates. According to clause 6,
The control unit,
An automatic sorting and separation system that uses the final sorting coordinates of plastic waste determined to be PET to have a robot pick up plastic waste of materials other than PET.

Images