Performance evaluation for shape estimation of extended objects using a modified hausdorff distance
Lifan Sun, Sen Zhang, Baofeng Ji, Jiexin Pu · 2016
To verify the validity of an tracking algorithm, its performance should be evaluated and compared with other algorithms. Unlike point target, the task of extended object tracking is to estimate the object shape in addition to its kinematic state. Especially for the shape estimation performance evaluation, there is a pressing need for measuring the degree of the similarity between the estimated shape and true one. In this context, this paper proposes a modified Hausdorff distance to handle this problem, which fully considers different parameterization of the star-convex model for extended object tracking. A example illustrating the modified Hausdorff distance for extended object shape evaluation is presented in simulations. Results demonstrate that it is a qualified measure to evaluate the performance of extended object shape estimation.