Mathematical Preliminaries for Video Anomaly Detection Techniques

The video-based anomaly detection techniques usually construct a normal model by learning and determine those testing samples that deviate from the normal model as abnormalities. Even though the treatment of video anomaly detection in this book is more engineering than mathematical, we provide some mathematical preliminaries for beginners to better understand the anomaly detection techniques we will introduce. More specifically, in this chapter, some of the concepts in probability and statistics and information theory are briefly reviewed. We will also look at some knowledge of neural networks that leads to the cutting-edge deep learning-based techniques.