Object Color Identification and Classification using CNN Algorithm and Machine Learning Technique

Abstract

The need for automation in the textile industry is growing rapidly today. Color based object sorting is a highly challenging process to be considered and needs to be addressed. It involves an automated material handling system. It synchronizes the movement of robotic arm to pick up objects moving on a mobile robot. It aims in classifying the colored objects then picking and placing the objects in its respective pre-programmed place. Thereby eliminating the monotonous work done by human, achieving accuracy and speed in the work. The core objective of the project is to propose an intelligent color-based object sorting system using deep learning technique like Convolution Neural Network for extraction of feature embedded with the machine learning algorithm. The two classifiers Random Forest and K-NN algorithm were implemented and studied for better classification. Based on the performance metrics, the Radom Forest algorithm out performs in classification. The project module involves cameras that captures the object’s color through the computer vision Library and sends the signal to the controller. The dataset of the captured images has been uploaded and compared with the trained data set. The ESP 32 Module transmit a signal to relay circuit, which then drives the robotic arm’s multiple motors to grip the object and position it in the given area. Based on the color observed, the robotic arm goes to the given point, releases the object, and returns to its original position.