Abstract
Background subtraction is a mainstream algorithm for moving object detection in video surveillance systems. It segments moving objects by using the difference between the background and input images. The key to background subtraction is to establish a reliable initial background. In this study, we propose a background subtraction algorithm based on category entropy analysis that dynamically creates color categories for each pixel in the images. The algorithm uses the concept of a joint category to build background categories that can adapt to the color disturbance of the background. Furthermore, in order to overcome dynamic background environments, this paper proposes the concept of color category entropy to estimate the number of necessary background categories and establish sufficient and representative background categories to adapt to dynamic background environments. In addition, recent mainstream methods for background subtraction were implemented and analyzed in comparison with our algorithm.
Highlights
IntroductionAdvances in network construction and computer hardware have led to a thrivingInternet of Things (IoT) industry
In recent years, advances in network construction and computer hardware have led to a thrivingInternet of Things (IoT) industry
In order to solve these problems encountered in the initial background extraction process, this paper proposes an entropy-based initial background extraction (EBBE) method, which—based on the analysis of the change in color category entropy—can automatically determine the background convergence time required for each pixel
Summary
Advances in network construction and computer hardware have led to a thrivingInternet of Things (IoT) industry. The IoT system has allowed various types of sensors (temperature, pressure, current, image) to be installed in electronic devices, which enable terminal electronics to collect environmental information. Benchmark datasets are critical to the qualitative and quantitative analysis of algorithms. We selected four image segments from two benchmark datasets with ground-truths, Carnegie. Mellon [26] and Change Detection.Net (CDW) [27,28], to verify algorithm performance. Carnegie Mellon: This dataset has only one image sequence containing 500 original images with artificially labeled ground-truths. Calculate the number of necessary candidate background categories NBG = [2E]. The number of samples in the candidate background category > THmatch ╳ NBG.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
