Ice Detection on Edge Device Based on Most Significant Digit First SVM

Abstract- The present notion of smart cities inspires urban planners and academics to provide citizens with modern, secure, and sustainable infrastructure and a decent standard of living. To meet this demand, video surveillance cameras have been installed to improve the safety and well-being of the populace. Despite technological advances in modern science, detecting anomalous events in surveillance video systems is difficult and takes extensive human effort. Due to the context-dependent nature of the anomalous concept, we identify several objects of interest and freely accessible datasets for anomaly detection. Iced surface in highways is one of the serious anomaly which causes a large number of accidents on the highway every year. CCTVs installed on highways can be a good tool to detect iced surface. Detection is a time-sensitive application of computer vision; hence, this paper has concentrated on investigating iced surface detection using edge devices and methods. We have leveraged Most-Significant Digit First arithmetic to improve the performance and resource utilization of the Support Vector Machine. We have applied our proposed method to address the problem of ice detection on highways, where the experimental results indicated a significant improvement in terms of accuracy, speedup, and energy consumption. These metrics are essential for edge computing and real-time intelligent surveillance applications.