AN EXAMPLE OF MACHINE LEARNING APPLICATION IN THE PROBLEM OF THE MOVING OBJECT DETECTION AND TRACKING IN VIDEO SEQUENCE

Abstract

This survey has been carried out for detecting and tracking the moving objects in the video which are of interest under variable lighting conditions. In this survey, deep learning based solution has been proposed which uses a pre-trained model for detecting the objects which are of interest, and a tracking algorithm for tracking the detected objects. The proposed method has a fixed camera requirement.

Another method has also been proposed which uses a pre-trained model for detecting the objects which are of interest like the previous method, but instead of the tracking algorithm it uses a pre-defined radius for identifying the object in two sequential video frames. This method is applicable for detecting and tracking the objects which are of interest without knowing whether they are moving or not. The proposed method is applicable also when the camera moves slowly.

Unlike the “Background Subtraction” based solutions the proposed methods are applicable under variable lighting conditions because a good trained model solves that problem.

A comparison has been performed between the proposed deep learning based methods and the combination of the “Background Subtraction” and “GOTURN” algorithms.

The problem of object detection has been solved using the modern “Tensorflow” framework.