A great number of computer vision software, especially computer vision cameras, is placed in public facilities such as banks, railway stations, airports, subways, etc. globally. The abandoned object detector is a widespread machine vision technology tool applied to deliver secure digital solutions helping to prevent and detect criminal activities and monitor other non-criminal ones around facilities.
It deals with recognizing moving objects and ones that stopped moving over a period of time. Nowadays, it is a technologically challenging issue in the computer vision and machine vision. In this article, we will consider some algorithms that our team applies to solve the mentioned above issue. We use two types of computer vision algorithms: based on the tracking and on the search. Let’s delve into details:
Tracking moving objects
This type is applied to detect the objects that an object/person left and a person who left it as well with computer vision camera. How it works:
The background should be separated.
Objects’ moving directions on the screen should be built.
The computer vision camera captures the "split" of the object that moves into two, one of which continues to move, while the other remains static.
Objects’ speed, moving direction, color characteristics are analyzed.
Obtained data is used to compare objects and divide them into two properties.
Foreground objects search
All the objects in motion on the screen refer to the foreground objects. This method implies two background models usage: short-term and long-term. This is because a static object merges with the background very fast, and there are some difficulties to track the time period since the object hasn’t moved.
Short-term model. Its update time is very short and all foreground objects changes should be captured quickly. If a new moving object appears on the screen, it will not merge the background. Otherwise, if an object has stopped moving, it will be merged. This model determines whether there was a movement on the screen or a moving object, but then an object has become static and merged with the background.
Long-term model. Its update time is much more slowly. New objects appearing on the screen stay in the foreground, but then become the background part after a while. This model is used for abandoned objects detection. It allows one to capture the fact that it appeared on the screen but was not there from the very beginning as part of the original background.
ProVision Lab has a great technical background and a significant presence in the computer vision industry. There are a number of business processes that today can be replaced by artificial intelligence-based applications and eliminate mistakes. We have off-the-shelf solutions for your business in medicine, engineering, agriculture, and motor industries.
If you have any questions, let us know and we will offer you a computer vision consulting services or help you build and implement any business idea into computer vision software that will enhance clients retention and boost revenue.