A novel technique for recognition and tracking of moving objects based on E-MACH and proximate gradient (PG) filters
Haris Masood, Saad Ur Rehman, Muazzam Ali Khan, Qaiser Javed, Muhammad Abbas, M. S. Alam, Rupert C. D. Young · 2017
Recognition and Tracking of Images is still one of the most sought after areas in the field of Image Processing mainly because of applications and presentations associated with it. Recognition of objects in a crowded and occluded environment is very challenging task as all the other objects besides the object of interest acts as noise. Other challenges associated with recognition and Tracking is the ever changing coordinates of the object during the tracking procedure. In this paper both the above mentioned challenges have been addressed using a novel technique which merges a very efficient recognition and tracking procedures. First recognition of image is done using modified Maximum Average Correlation Filter (MACH) which will identify the two dimensional coordinates of the object if interest in a secluded environment. The coordinates are then updated using a Proximal Gradient (PG) filter which uses Particle Filter as for systematic and periodic updation of images recursively. Proximal Gradient Filter uses the random variables and their posterior distribution for efficient prediction of coordinates of object of interest. End result is very efficient and fast tracking of object in subsequent frames.