Title:

Machine-Learning Classification of Two-Dimensional Objects Using Reflected Waveforms

The ability to classify an object using a reflected wave could prove extremely useful in aerospace, defense, autonomous vehicles, or other applications. Time Series Classification (TSC) tasks are frequently tackled using machine learning techniques such as Support Vector Machines (SVM) and various Neural Network methods. We implement SVM for classification on a numerically simulated dataset of fifteen two-dimensional objects, each sampled at nine angles of reflection. Multiple types of SVM models were implemented from scratch in MATLAB, with a Directed Acyclic Graph-based model (DAG-SVM) achieving the best accuracy. Due to its high accuracy and fast computational speed, we recommend a linear-kernel DAG-SVM model fit on the dataset's principal components for this task. Further research could involve application of the DAG-SVM model to wave-based classification of three-dimensional objects, as well as an investigation of other machine learning methods such as Convolutional, Artificial, or Boosted Neural Networks.

Name

Comment*

Comments are viewable only by submitter

Captcha*