Abstract:

The concept of "reconversion" that emerged for the supplementation of materials that are necessary for both the conservation of nature and the continuously increasing human needs and in limited quantities in nature has become one of the most important topics in recent years. As a result of a particular recycling process, the collection of recyclable waste known as "re-use raw materials" has become a problem that top and local governments around the world are also concerned. For this purpose, special boxes for recyclable waste are placed in certain centers and people are trying to be encouraged to recycle. In this study, a material recognition system is developed using the necessary electronic materials and software to detect paper, glass and plastic waste in the recycling boxes in real time for use in recycling projects. The system performs a simulation function that recognizes the materials of recyclable solid waste thrown into the recycling box and charges the user account according to the material. The advanced hardware also includes electronic devices that can be connected to Raspberry Pi, such as camera, LCD screen, LED, IR LED, circuit board and jumper cable. At the development stage of the software necessary for the recognition of materials, 845 paintings containing paper, glass and plastic materials have been prepared in the work framework and 662 of them have been used for training on the Tensorflow object recognition library. In order to detect materials in real time by the developed hardware and to use the obtained object recognition model on the hardware, the Raspberry Pi has a computer-based OpenCV vision library installed. Finally, the developed hardware has been fixed on the boxes dedicated to the relevant materials and the system has been made to work in real time. Some waste was thrown into the box to verify the system works properly and the results were displayed on the LCD screen.

Keywords:

Abstract:

The concept of "recycling", which emerged to reinforce the limited amount of materials in nature, which is necessary for both the protection of nature and the ever-increasing human needs, has been one of the most important issues in recent years. The issue of collecting recyclable wastes known as “the wastes whose raw materials can be reused” as a result of a certain recycling process has been a problem that the top and local governments interest around the world. For this, special boxes for recyclable wastes are placed in certain centers and people are encouraged to recycle. In this study, a material recognition system is developed by using the necessary electronic materials and software to detect paper, glass and plastic wastes in recycling bins in real-time to be used in recycling projects. The system functions as a simulation that recognizes the materials of recyclable solid wastes thrown into the recycling bin and charges the user account according to the material. The developed hardware includes electronic devices that can be connected to the Raspberry Pi such as camera, LCD screen, LED, IR LED, breadboard and jumper cable. During the development of the software required for the recognition of the materials, 845 pictures including paper, glass and plastic materials were prepared within the scope of the study and 662 of them were used for training on the Tensorflow object recognition library. The OpenCV computer vision library has been loaded into the Raspberry Pi so that the materials can be detected in real-time by the developed hardware and the obtained object recognition model can be used on the hardware. Finally, the system has become works in real-time by fixing the developed hardware on boxes dedicated to the relevant materials. To verify that the system is working properly, some waste has been thrown into the boxes and the results are displayed on the LCD screen.