2D Geometric Object Shapes Detection and Classification

Abstract

In computer vision, object detection is a basic process for advanced procedure forms such as object detecting, analyzing, tracking, etc., for further processes, features extraction play a vital part to identify the objects accurately. Most of the existing frameworks may not be able to distinguish the objects appropriately when different objects have a place to a single frame. In this work, an automatic detection and recognition system of two-dimensional geometric shapes have been proposed. Firstly applied Genetic Algorithm (GA) to fill the shape after performing proper segmentation pre-processing method. The proposed framework is able of identifying numerous objects in the input image, determining the sort of the identified object, and labeled the recognized objects. Statistical method has been applied for each shape to extract the objects corners' points by calculating the largest boundary to form the features vector. Ultimately, the identified objects are classified as geometrical shapes such as square, rectangular, triangular, or circular. The proposed method achieved high accuracy around 98.3%, and an average computational time 0.521 sec. as an effective classification technique.