Aug. 5, 2021
Computer vision is a tremendously broad field, with dozens of domains and potential applications. Motion tracking is a subfield of computer vision that seeks to follow the motion of a person or object across multiple frames in a video.
Motion tracking is an extension of object recognition, another highly popular activity in computer vision. In object recognition, an AI model is tasked with identifying the different objects in an image. Motion tracking depends on object recognition to locate the objects and their initial positions, and then must also follow their movement throughout the video, recognizing them as the same objects.
Successful motion tracking presents several challenges:
There are several algorithms for motion tracking, depending on the exact situation, such as whether the objects are 2D or 3D, or whether the camera is also moving. “Background subtraction” methods are often suitable for the most simple use cases, such as when objects are not moving too quickly. In these methods, the pixel values of one video frame are subtracted from the values of the next consecutive frame. If two objects in two different frames have significant overlap, then the model considers them to be the same object.
Other, more advanced motion tracking algorithms attempt to calculate a person’s or object’s trajectory, or account for prior information about the person or object, in order to improve the model’s accuracy. For example, if the model knows it is tracking a person, it can limit itself to searching for “human-like” shapes in the next frame.
Below are just a few use cases for motion tracking in different industries:
Looking for more computer vision definitions? Check out our computer vision glossary. If you want to implement motion tracking for your own business needs, you can get in touch with our team of AI experts for a chat about your situation, or sign up today with a free account on the Chooch computer vision platform.