Optimized Multi-Sensor Multi-Target Tracking algorithm for air surveillance system

Abstract

Multiple Target Tracking is an essential requirement for modern air surveillance system employing with multiple sensors to interpret the environment. The fundamental problem in a multiple target tracking system is that of data association - with the goal of choosing most probable association between a particular observation and target to be tracked. This paper presents an Optimized Multi-Sensor Multi-Target Tracking algorithm (OMSMTT) that solves measurement to track association conflicts in sensor data reports and capable of initiating and terminating new tracks. The proposed solution enhances the Multiple Hypotheses Tracking (MHT) algorithm with multidimensional assignment approach by modified scoring procedure which improves performance of the tracker. The work exploits a generalized assignment problem structure; formulate the new framework which treats measurement to track assignment for each target as a random variable and transform it to the particle filter for non maneuvering multi-target models. The algorithm is tested with non maneuvering targets with sparse scenarios in presence of clutter against traditional Multiple Hypothesis Tracker (MHT) under extreme conditions such as track swap and coalescences and analyzes it using different performance metrics.