Trajectory-oriented calculation time reduction method based on multi-hypothesis tracking algorithm

Abstract

The invention discloses a trajectory-oriented calculation time reduction method based on a multi-hypothesis tracking algorithm. The multi-hypothesis tracking algorithm keeps multiple hypotheses about the correlation between a trajectory and a measured value in the tracking process. What is well known is that when multiple targets are tracked or tracking is conducted in a strong clutter environment, time and memory resources need to be calculated due to hypothetical exponential growth. When a trajectory-oriented multi-hypothesis tracking algorithm is applied to the start of a trajectory, each measurement point scanned each time must consider a measurement hypothesis from a new target, so that storage resources are optimized. For the problem of large calculation amount of a multi-hypothesis tracking algorithm, an N-optimal algorithm is applied to optimize storage resources, and calculation time is shortened. A best hypothesis is generated from each likelihood matrix using an N-best algorithm. A second best hypothesis is generated by the following procedures. A second hypothesis is generated from a likelihood matrix, and the likelihood matrix generates a generally best hypothesis; and an optimal hypothesis is searched from the second hypothesis and other likelihood matrixes.

Description

technical field [0001] The invention relates to the technical field of multi-hypothesis tracking. It mainly involves a trajectory-oriented multi-hypothesis tracking method to reduce computing time. Background technique [0002] Data association is an important part of multi-object tracking. In some naval applications, correlation is paramount, so trackers are called correlators. The need to exploit multiple frames or data scans to track multiple targets was recognized long ago, but early work focused on single target tracking. The use of multiple data association assumptions to explain the origin of all measurements first appeared in the late 1970s, where the best assumptions were solved in batches by 0-1 integer programming, and multiple association assumptions were recursively evaluated. When data association is difficult due to high object density, dense clutter, and low detection probability, multi-hypothesis tracking almost immediately becomes the standard method for...

AI Technical Summary

Problems solved by technology

However, when tracking multiple targets or in a highly cluttered environment, it requires a lot of computing time and memory resources due to the exponential growth of assumptions

Images