Implementasi Metode Deep Neural Network dalam Sistem Pendeteksi Logam Berbasis Mikrokontroler

Abstract

This research aims to develop a microcontroller-based metal detection system with the integration of the Deep Neural Network (DNN) method to improve the accuracy of metal type classification. The DNN method here uses infrared sensors as auxiliary sensors because the inductive proximity sensor used can only detect the presence or absence of metals. The system is designed with a microcontroller capable of running a lightweight DNN model. Metal sensors such as inductive proximity and infrared sensors are used to generate input data. The data is then processed to be classified by the DNN model and find out how much metal passes through the sensor and what the frequency. And the manufacture of the tool was successfully carried out with the test results of 5 metals detected with frequencies of 0.000 Hz, 0.133 Hz, 0.099 Hz, 0.119 Hz, 0.060 Hz. Using the tool made, the test results are still not perfect but can be refined if more suitable components are used.