Beginning Anomaly Detection Using

 Video  Surveillance

Another domain where anomaly detection is becoming extremely important is 

video surveillance. Nowadays, it is very common to see security cameras and video 

surveillance systems no matter where you go: a local school, a local park, Main Street, 

near a neighbor’s house, or in your own house. The point is, video surveillance is here 

to stay. Given all the new technological advancements in smart apps and smartphones, 

this is definitely not going to change any time soon. Rather, we should expect much 

more video surveillance. In the very near future, we will see lot more smart cars and 

self-driving cars. They also depend on continuous processing of video using real-time 

analysis and detecting various objects. At the same time, they can also detect any kind 

of anomaly. In a strictly security video surveillance sense, anomaly detection can be 

used to detect the normal for the specific camera that is looking at your backyard. When 

a specific anomaly is detected because of some kind of motion within the vicinity of 

your house, such as a wild animal or even an intruder walking on your lawn, your home 

security system is able to see that this is not normal. In order for the cameras to do this 

effectively, the manufacturers train very sophisticated machine learning models to 

assess the video signals in real time. The feed coming from the cameras is determined 

as normal or abnormal. For example, if you are driving in a self-driving car on the 

interstate, video of the car will clearly indicate what is normal right now according to 

how the road should look, where the signs should be, where the trees should be, and 

where the next car should be. Using anomaly detection, self-driving cars can avoid any 

abnormalities happening on the path and then take corrective action before anything 

bad can happen.

Figure 

8-15

 is an object-detecting video surveillance system.

Chapter 8   praCtiCal Use Cases of anomaly DeteCtion

---

313