Automatic Intelligent Garbage Classification Bin

Project Source

Xingke Yang and his teammates participated in the biennial "Engineering Practice and Innovation Ability Competition of Chinese College Students" in 2020. Based on the functional requirements of the "Intelligent Garbage Classification Project", they invented an intelligent garbage classification bin.

Project Description

Utilizing intelligent image recognition and control algorithms, along with an infinite-degree-of-freedom flap structural design, this intelligent garbage classification bin can intelligently identify, compress, classify, and store common household garbage. In addition, it has a full-load warning function and is equipped with a high-brightness display that supports video and picture playback in multiple formats and displays various data in the garbage bin. Garbage is categorized into "recyclable waste, kitchen waste, hazardous waste and other waste".

Project Work Structure

■External Device Design

  • Input Device:Shot 30 kinds of garbage from multiple angles to form a neural network data set, imported the learned network into the camera, and exchanged data with the Arduino motherboard through the I2C bus.
  • Display Device: Utilized the Raspberry Pi supporting display screen to display the identified garbage name, the number of this time, whether it was completed or not, the full load status, and the current total number of various types of garbage.
  • Detection Device: Utilized four ultrasonic distance sensors to monitor a full load of each trash can in real-time and fed the measured distance of the garbage in the trash can from the top of the bucket to the main control board in time to realize the full-load function monitoring.
  • Actuator Device: Employed the 180° steering gear, lift push rod and compression push rod to form the execution device to control the inclination of the garbage receiving platform to any angle, compress cans and other items.
    •
    ■Control Algorithm

  • Garbage Type Identification Algorithm Design
    1. Initialized the configuration for the above peripherals, request the camera to obtain the recognition result, utilized two recognitions to de-jitter, and selected and judged the returned ID value to determine the type of garbage to execute the corresponding function of the throwing function.
    2. Stored the key information in the EEPROM memory for recording and storage and sent different garbage corresponding character messages to the Raspberry Pi for display through the serial port.
  • Results Display Algorithm Design
    1. Utilized the relevant library functions to call the video file, played the garbage classification propaganda video in a loop, and monitored the serial port data.
    2. Called the prepared audio files in different garbage execution functions for voice broadcast prompts.
  • Garbage Placement and Full Load Judgment
    1. Completed the garbage dumping by controlling the lowering of the telescopic push rod and controlling the angle of the steering gear.
    2. Sent information to the upper computer display, stored the information in the EEPROM memory and judged whether the trash can was fully loaded by querying the distance value returned by the ultrasonic sensor after delivery.

    Project Result

    In the actual application scenario, the accuracy of recognizing and classifying 30 kinds of common household garbage can reach 95%, and a piece of garbage can be classified and put away in about 2.5 seconds. At the same time, when the capacity of the garbage bin exceeds 80%, it will trigger a full-load alert.

    Project Achievements

  • National First Prize:After the selection of the school competition, the provincial competition, and the national competition. Finally achieved the Gold Medal (National First Prize) in this national Class A subject competition. According to the speed and accuracy of the tasks completed in the competition, the scoring of the equipment function is ranked. Among the 900+ teams participating in the same program nationwide, our device ranked 11th in terms of task completion and speed.
  • National Utility Model Patent: We applied for a utility model patent for the initial first version of the intelligent garbage classification bin after the competition, which has now been granted.
    •