Accurate and High Speed Garbage Detection and Collection Technique using Neural Network and Machine Learning

Abstract

Due to the increasing development of cities’ populations, which has resulted in massive garbage output, waste management systems in urban areas are confronting issues. The ravage of possessions can be employed powerfully with the incorporation of the internet of things (IoT), TensorFlow based deep learning model, as conventional ravage managing system are extremely uneconomical. The major goal of this study is to create a smart waste management system based on a deep learning model that optimizes trash isolation and allows for bin status monitoring in an IoT context. Yolo real time object detection algorithm is employed and educated with a dataset that includes paper, cardboard, glass, metal, and plastic for garbage sorting and grouping. Yolo algorithm enhances the detection speed and yields precise findings with low background noise. Yolo uses convolutional neural network to detect the object. The camera module detects garbage and the servomotor linked to a plastic board, categorizes the waste into the appropriate waste cubicle using the educated model on TensorFlow Lite and Raspberry Pi 4. The garbage fill is monitored by an ultrasonic sensor, and the latitude and longitude are obtained in real time by a GPS module. The smart bin’s LoRa module transmits the bin’s status to the LoRa receiver at 915 MHz. The smart bin’s electronic mechanisms are safeguarded by an RFID-based locker that can only be opened with a registered RFID badge for maintenance or upgrades. This work is framed out of the technologies such as Robotics, neural network, Internet of Things and deep learning algorithm. The garbage detection system is more precise and faster than the other existing methods. The YOLO algorithm can predict objects in real time, which speeds up detection. It’s a prediction method that produces exact results with little background noise. The algorithm has outstanding learning capabilities, allowing it to learn and apply object representations to object detection.