For Project Enquiry +91 9791 044 044

EXISTING SYSTEM

PROPOSED SYSTEM

EXISTING CONCEPT:

          These distance measures explicitly utilize the fact that for natural images there is a high correlation between spatially close pixels.
          Image matching is used in various computer vision tasks, and the requirements to the distance measure are application dependent.

PROPOSED CONCEPT:

          In the proposed system three overlap-based measures for multiple extended-target video trackers that evaluate performance at frame level taking into account (i) the accuracy and the cardinality errors (ii) long-term tracking accuracy using lost-track-ratio information; and (iii) ID changes in a parameter-independent manner.

EXISTING TECHNIQUE:

          Distance-based measures

          Overlap -based measures

PROPOSED  TECHNIQUE:

          Moving Target Detection

          Segmentation

TECHNIQUE DEFINITION:

          Distance-based measures are defined as a zero-overlap between estimated and corresponding ground-truth states.

          Overlap-based measures generally consider the estimated target-size variations and can detect instances of tracking failure.

TECHNIQUE DEFINITION:

          Moving Target Detection and Segmentation can be achieved using change detection, i.e., by comparing the current frame with the previous frame or multiple frames to locate regions where change is occurring due to foreground target motion.

DRAWBACKS:

          Generally consider the estimated target-size variations and can detect instances of tracking failure.

          Some existing measures are numerically unbounded and not well defined for the worst tracking case.

ADVANTAGES:

          Three measures (METE, MELT, NIDC) that quantify key factors in extended multi-target tracking: accuracy, cardinality and ID changes.

          Overlap threshold limits its ability to clearly distinguish different tracking results.