Software that helps computers watch videos

Researchers at the University of Missouri, USA are developing a software that enables computers to analyze videos like humans.

With the above application, surveillance camera detects that a person falls without standing up. The software can also be used to search within videos and certain objects for security.

"The goal of the research is how to make computers understand the content of videos and identify objects. This analytical capability brings a lot of applications at a cheaper cost than object detection. or human with sensors, " said Tony Han, professor of electrical and computer engineering at the University of Missouri's Technical College.

The software is recognizing and analyzing objects in the video.

Researchers believe that human detection software can also be applied to driver assistance. For example, if the software detects a pedestrian in the same lane as the vehicle, it will cause the vehicle to stop to avoid an accident.

This software can also help improve the care of the elderly living alone. If there is an indoor monitoring and monitoring system, when the elderly person suddenly falls or has other abnormal phenomena, the system will alert to call the doctor.

Professor Han is working with students.

Professor Han said: "Students and I are working with algorithms that automatically detect objects, but it is difficult to create a perfect algorithm. We are trying to find an algorithm. The detection is more reliable, but it takes a lot of time to test. We prepare more than 3,000 images with locations and objects to test the above algorithms. "

Earlier this year, Han and his students participated in the PASCAL object detection competition. In the first phase of the competition, they won the championship in the competition to discover potted plants, chairs and second prizes in the competition to detect people, cars, horses and bicycles.