RU188755U1 - AUTOMATIC RECEPTION TARE - Google Patents

Abstract

The utility model relates to machine vision systems, in particular, to robotic units that receive used packaging and are driven by its presence in the receiving device with payment of its collateral value. The tasks of the utility model are to improve the recognition accuracy of the received packaging by eliminating barcode read errors during processing tare and increase the speed of decision making. A tamper receiving machine includes a housing 1 with a receiving opening 2, behind which recognition compartment 3 is located. In the recognition compartment 3 a device is located that collects information about a recognized container — video cameras 4, a container recognition device — a microcontroller 5, which contains in memory a neural network algorithm and performs logical operations with the information received, locking devices 6 (electromagnetic pushers) and sorting flaps 7. there is a storage compartment 8 of the received container, divided by a bulkhead into two parts 9 and 10. On the front panel, the machine is equipped with a display 11 and a means of issuing a reward 12 (a ribbon printer).

Description

The utility model relates to machine vision systems, namely, to robotic units that receive used packaging and are driven by its presence in the receiving device with payment of its collateral value (Reverse vending machine).

A known machine for receiving used bottle containers, comprising a housing in the form of a rectangular container with a door, a receiving device with an inlet opening for input and output of the container, a recognition device, a means of accumulating the container and a means of issuing a reward. The receiving device is made in the form of a funnel with a recognition device at the end containing movable elements providing analysis of the size of the bottle neck corolla and interacting with series-connected contacts located in the electromagnet electric supply circuit, the anchor of which is connected to the gripping element of the neck, and the receiving device is placed on movable carriage moving along the guide rail by means of a screw pair driven by an electric motor, and the change in direction rotation of the electric motor is carried out through limit switches (RF patent No. 2639065, IPC G07F 11/00, G07F 7/06).

The reward tool is made in the form of two parallel-arranged elements with a gap in which a plate with a hole is placed, which is set in motion by an electromagnet, and the upper element contains a coin-shaped magazine with coins, as well as a sensor for monitoring the presence of coins.

The machine is designed as a stand-alone device containing a solar battery connected by means of a charge controller with a battery. The machine further comprises a container placed on the outside of the door. The disadvantages of the device are:

- the possibility of receiving only the bottle;

- the inability to sort accepted containers.

Closest to the claimed device is a hardware-software complex (hereinafter referred to as the AIC) "Fandomat" by Loetec Elektronische Fertigungssystem GmbH. AIC "Fandomat" is used for automatic acceptance of used bottles of polyethylene terephthalate (hereinafter referred to as PET) and aluminum cans with a capacity of 0.2 to 1 liter. The total volume of the receiving compartment of the machine - 430 liters. In exchange, money, checks or plastic cards can be issued (http: //www.loetec.coni/dateien/de/produkte/pfaridomat.html; http://www.waste.ru/modules/equipment/item.php?itemid= 326).

The APC contains a housing in the form of a rectangular container in which a receiving device with a window for inputting and withdrawing a container (recognition compartment) with a device located in it that collects information about a recognizable object (barcode scanner) and a container recognition device (microcontroller producing logical operations with the information received), the container storage compartment and the means of issuing remuneration. Tare identification is performed by scanning the barcode on the tare label and comparing it with the base of accepted containers (in the base about 30 thousand barcodes). Recognized container gets into the storage compartment, passing through the open restrictive sash. Unidentified objects, as well as bottles of liquid residues are returned back. All information is displayed on the LCD screen.

The disadvantages of the device are:

- low decision making speed;

- the inability to accept the damaged packaging (dented, deformed, without a label, etc.) due to the impossibility of high-quality reading of the barcode.

Signs of the prototype, which coincide with the essential features of the invention: the case, the container recognition compartment with the device located therein, collecting information about the object to be recognized, and the container recognition device (microcontroller producing logical operations with the information received), the container storage compartment and the means of reward.

The objectives of the utility model are to increase the accuracy of recognition of the accepted packaging by eliminating errors in reading bar codes when processing containers and increasing the speed of decision making.

The task of the utility model with the achievement of the technical result was solved due to the fact that in a known machine for receiving containers, comprising a housing, a receiving device with a device that collects information about the container being recognized, and a container recognition device - a microcontroller, a means of accumulating containers and means reward, as a device that collects information about a recognizable object, it contains at least one video camera, and the microcontroller contains in memory a neural network algorithm that allows To recognize tara.

FIG. 1 shows the design of the proposed tare receiving machine; FIG. 2 shows the recognition compartment and the storage compartment; FIG. 3 is the optimal placement of video cameras.

A tare receiving machine includes a housing 1 with a receiving opening 2, behind which recognition compartment 3 is located. In recognition compartment 3, there is a device that collects information about a recognizable container — video cameras 4; a container recognition device, a microcontroller 5, contains a neural network algorithm in memory logical operations with the received information, locking devices 6 (electromagnetic pushers) and sorting shutters 7. In the lower part of the machine, there is a storage compartment 8 of the accepted container divided by a bulkhead into ve parts 9 and 10. The front panel of the machine has a display 11 and means for issuing reward 12 (tape printer). The microcontroller 5 contains in memory a neural network algorithm, thanks to which the operation of convolving an image of a recognizable tare is performed. The result of this operation are the signs of the image. The obtained signs are compared with the already known neural network signs of tare classes (“PET bottle” and “aluminum can”) stored in the microcontroller's memory, and the probability of their correspondence is calculated. Based on the obtained probabilistic indicators, the microcontroller decides whether the recognized container belongs to one of the classes (“PET bottle” and “aluminum can”) or it does not correspond to any of the classes (non-accepted containers or other items that are not containers).

The machine can be installed at least one video camera. The best option is to install two cameras on the upper surface of the recognition compartment (Fig. 3). When installing one camera, a significant increase in the height of the recognition compartment 3 is required. Installing a larger number of cameras increases the technical complexity of the implementation of the automaton and gives a slight decrease in the height of the recognition compartment 3 due to the focal length limit. When choosing the number of cameras, it is necessary to take into account that the total area of visibility of all cameras, when viewed from above, should not be less than the recognition compartment area 3 at the level of 100 mm (maximum diameter of the container in question) from the bottom point.

The device works as follows. The user places the container through the receiving opening 2 into the recognition compartment 3. The video cameras 4 record the appearance of the container in the recognition compartment 3 and transmit the image to the microcontroller 5, the Microcontroller 5, registering the location of the container in the recognition compartment 3 as a whole, collects images received from all cameras 4 installed on the machine , sticks them together, receiving the full image of the container and recognizes it. Next, the microcontroller 5 compares the identified packaging with the already known neural network signs of the classes of packaging (PET bottle and aluminum can) stored in the microcontroller’s memory, and calculates the probability of their correspondence. Based on the obtained probabilistic indicators, the microcontroller 5 decides whether the recognized container belongs to one of the classes (PET bottles and aluminum cans) or it does not correspond to any of the classes (non-accepted containers or other items that are not containers). If it is decided that the PET bottle is in the recognition compartment 3, a signal is sent to one of the locking devices 6, which opens one of the flaps 7 and the bottle rolls into the storage compartment 8 in compartment 9 (Fig. 3). If it is decided that the aluminum can is located in the recognition compartment 3, a signal is sent to the second locking device 6, which opens another door 7 - the can is rolled into the storage compartment 8 in compartment 10 (Fig. 3). If it is decided that there is unacceptable container (or other non-container) in the recognition compartment 3, a message is sent to the display 11 asking the user to remove the container from the recognition compartment 3. In case of successful acceptance of the packaging, the microcontroller sends to the display 11 message of gratitude for the use and prints Quick Response code (hereinafter QR code) on paper tape. After completion of all operations, all elements go to the starting state and expect a new appearance of the container in the recognition compartment 3.

The QR code is unique for each fact of acceptance of the container (generated by the microcontroller). Later, the user can use this QR code in the partner network providing goods and / or services (for example, receiving discounts at the grocery supermarket upon presentation at the checkout or electronic store).

In the proposed automaton, tare recognition does not depend on the presence and quality of the barcode on it, which eliminates the need for correct positioning of the tare in the recognition compartment 3. Recognition is performed by the characteristic appearance of the tare.

A number of tests were carried out aimed at measuring the time and quality indicators of tare recognition by the machine. According to tests, the recognition of a container placed in the recognition compartment 3 using neural networks is faster than the prototype. The machine is able to make a decision in less than 1 second. During the testing of the machine, a neural network was set up, capable of identifying tare in 3 classes - “PET bottle”, “aluminum can”, “other”. Testing was carried out both on tare samples in perfect condition and on deformed samples (dented, crushed, flattened and other PET bottles and aluminum cans). The automaton has shown good results in recognizing ideal and slightly deformed samples, medium ones - with the deformation of the container, which significantly changes its geometry. Samples that are not tare, the system was able to determine and refuse to accept them.

The prior art does not know the use of neural networks for tare recognition.

Thus, the proposed device automaton for receiving containers using neural networks will significantly improve the recognition accuracy of containers and increase the speed of decision making.

Claims (1)

A tare receiving machine, comprising a housing, a tare recognition compartment with a device located therein that collects information about a recognizable item and a microcontroller tare recognition device, a tare storage compartment and a means of issuing a reward, characterized in that as a device that collects information about a recognizable item , it contains at least one video camera, and the microcontroller contains in memory a neural network algorithm that allows recognition of the container.

Images