Multiple Objects Real-Time Tracking And Detection Technology In Satellite Video Based On Multi-Core DSP

With the continuous development of video satellite,the satellite video is playing a more important role in the field of military and civilian.Multiple objects real-time tracking and detection etchnology in satellite is more popu;ar.While the performance of digital signal processors also continue to increase5realizeing multiple objects real-time tracking and detection on the hardware platform is becoming possible.It also will be an basis in on-orbit information processing.This paper include three parts multiple objects detection on satellite video,multiple objects real-time tracking on satellite,complete these algorithm on multi-core DSP.In order to tackle the problem of real-time multiple objects tracking in satellite video images,we adopted background subtraction method to detect moving objects.To solve the background subtraction's error detection,we proposed two-level hierarchical association:first level was the moving objects,linking between adjacent frame;second level was the associated lines matched each other to form new tracklets,to track multiple targets,used greedy algorithm and a set of actual physical constraints to keep complexity,gained the complex background's information of large number of moving objects which have a small number of pixels,reduced the noise due to pixel response non-uniformity.Multiple objects tracking algorithm's real-time was verified.Finally,we realize these algorithm on multi-core DSP hardware through master slave mode scheduling.In this paper,the data source was the Jilin-1 smart video satellite.