Depth Spatiotemporal Information Fusion Moving Target Tracking Algorithm Based on CUDA Library
Mobile target tracking technology plays an important role in many areas such as intelligent
transportation, video surveillance, virtual reality, and public safety. Although the current target
tracking technology is widely used and related application technologies have gradually
matured, there are still some performance defects in complex environments. In order to make
the target tracking algorithm adapt to the changes of the complex environment, and get better
target tracking effect in the practical application scene, we propose a depth spatiotemporal
information fusion moving target tracking algorithm (DSIF) based on CUDA library. The
algorithm is mainly divided into two technologies: moving object detection and target tracking.
A method of combining clustered differential threshold and matching model is used to detect
moving objects. By dividing the detected mobile events into clusters, we use distance mean
and matching the template to redefine mobile events and filter noise and other interference
elements. In the target tracking, effective tracking is performed on the same target based on
target feature information and spatiotemporal information change between frames.
Experimental results show that the algorithm not only can accurately track the locked target but
also can effectively distinguish and track multiple different targets at the same time