Detecting, Tracking, and Calculating the Speed of Colored Balls Using Deep Learning

Abstract

Developing tracking in various applications motivates researchers to explore this field.  In this work, an intelligent system is suggested to automatically detect moving colored balls in real time and calculate the speed and direction of these balls based on the deep learning algorithm RCNN. The data is compared with the Alex-net algorithm because it is a standard method. The proposed algorithm is one of the machine learning algorithms based on the principle of training and learning, which relies on the sequential classifier. The proposed system consists of a phone camera, colored balls, and different environmental lighting (changing from one to eight lights). There are two luxmeters used to measure the intensity of light. Four parameters are measured to evaluate the performance of algorithms and system setup: accuracy, average time, detection ratio, and speed. The best class and training were selected and approved for detecting the blue and green balls. This proposed algorithm can be used to detect any moving object. Results showed a high quality of ball detection and tracking with almost 100% accuracy.