Multi-Layer Perceptron Neural Network Implementation as Train Type Classification

The purpose of train detection systems is to check that related track section clear vehicles before a may be authorized pass through railroad. the important task for ensuring safety traffic. Multi-layer Perceptron classifier, which consists feedforward neural networks constructed multiple layers interconnected artificial neurons, proved effective trainset class classification in this study. Using Raspberry Pi and IMU sensor BNO055, dynamic response any type interaction can handled by windowing Real Fast Fourier Transform (RFFT). Dense layer with 5 using ReLu activation function, specifying input shape as (6= 3-axis accelerometer X, Y, Z directions, 3 axis directions from gyroscope). process implementation, consist three classes types, has been completed accuracy above 92,7%.