Video Analytics - Video that helps...

Video analytics is an excellent way to improve the quality of information extracted from camera video feeds. With human, vehicle, and pet detection being some of the most valuable types available, it's no surprise that video analytics have become so popular in recent years.

Video analytics provide valuable insight into events and activities which occur in front of one or more cameras, making it possible to know in great detail when and how something happened. This extra information can also be helpful in situations like abnormalities, behavioral patterns, and deviations from expected behaviors.

Video analytics add much more than just numbers to a video feed

Human Detection is used to identify whether a detected object is a person or not. This is a crucial factor in making a sound judgment on the situation and enabling you to search efficiently through detected objects by filtering out non-people activity.

People Detection identifies everyone within a detected group of people, which is handy for video analytics-based search functions and manual searches. In cases where several cameras cover one area, it is possible to view the detected people from all the cameras at once to compare behavior across different angles.

Vehicle Detection allows you to quickly identify vehicles twofold: their presence can be detected and their movement within a video frame or between consecutive frames. This allows, for instance, an accurate calculation of how long a vehicle was present in the field of view or to see how many cars drove by. Vehicle Detection also allows for specific identification of different types of vehicles, making it possible to filter out trucks from vehicles.

Animal (Pet) Detection is used to distinguish between small animals that move quickly through the field of view, such as dogs or cats. Although not yet very popular, pet detection is becoming increasingly important for pet owners that want to know if their pet returned home or not.

Tesla announced in May of 2021 that the Full Self Driving computer and their software will only rely on video camera footage gathered from the 8 cameras mounted around the car (as well as the ultrasound sensors on the bumpers) and not on the expensive Lidar equipment.

One other area for the usage of video analytics is process automation. With the advent of video image processing in the last decade, many more start-ups have emerged to provide their services for retail, FMCG, hospitality, building management, etc. Combined with machine learning (ML) and artificial intelligence (AI), there is an endless amount of data to train the algorithms to automate processes that would take years if done by humans. Doing these processes with humans would also be more error-prone.

One other controversial issue around video analytics is privacy. This is a separate topic that can be discussed in a future article.