Detection and Classification of Geometric Shape Objects for Industrial Applications

Abstract

Human being can have a strong capability of detection and recognition of objects to understand the natural environment. But for a machine to understand the nature like human is a challenging phenomenon due to the inconsistency of the environment, irregularity in the properties of surrounding objects etc. However, the proper learning procedure for the machine using the objects’ shape, size, color, texture and other related properties may produce the satisfactory detection and classification results. Most of the existing systems may not be able to detect the objects properly when multiple objects belong to a single frame. The proposed system will be able to detect multiple objects from an image, count the number of detected objects, separate these objects into individual image through greyscaling, thresholding, edge detection, finding the objects corner points, cropping etc. Finally the detected objects are recognized as geometrical shapes such as triangular, rectangular, and circular for simplicity. The classification is performed through ANN (Artificial neural network) and SVM (Support vector machine) using synthesized datasets. This proposed solution obtained 95% classification accuracy using ANN and 95% to approximately 100% using SVM. The learned ANN and SVM will also be able to classify the newly detected objects from an image.