Implementation of Object Detection and Tracking by Using Deep Learning Based Convolutional Neural Networks

Abstract

Video object detection plays a significant role in various applications, including security, remote sensing and hyperspectral. In recent years, deep learning-based algorithms have made significant advances in video object recognition. The conventional machine learning applications have resulted in poor accuracy. In this article, a unified deep learning-based convolutional neural network (DLCNN) is developed for composite multi-object recognition in videos. DLCNN analyses a composite item like a collection of background and adds part information into feature information to enhance hybrid object recognition. Correct component information may help forecast the shape and size of a feature data, which helps solve challenges caused by different forms and sizes of various objects. Finally, the DLCNN draws a bounding box to detect objects using background features. Further, the simulation results show that the proposed method's performance is improved compared to the state of art approaches.