Investigating Application of Machine Learning in Identification of Polygon Shapes for Recognition of Mechanical Engineering Drawings

Abstract

This paper investigates the applications of Machine Learning in recognition of 2D drawings of machine components. Recognition of primitive geometric shapes such as polygons within engineering drawings forms basic element of such approach. Machine learning algorithms are used to identify 3 to 7 sided polygons with random shapes and segmented edges. The uncertainty induced by segmented edges poses a challenge for predicting number of sides using statistical method such as Machine Learning. Different types of datasets with varying amount of uncertainty are used. The recognition of shapes is attempted with different sets of features such as coordinates of points, slopes of lines and geometric parameters such as area, perimeter and centroid. Three machine learning models namely, Random Forest Classifier, K- Nearest Neighbors Classifier and Support Vector Classifier are adopted. The performance of these models for identification of polygons is discussed.