Open AccessImproved IMM algorithm based on XGBoost
Author(s) -
Da Li,
Pei Zhang,
Ruifang Li
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1748/3/032017
Subject(s) - computer science , artificial intelligence , adaptability , filter (signal processing) , process (computing) , hidden markov model , algorithm , motion (physics) , machine learning , markov process , markov decision process , information source (mathematics) , pattern recognition (psychology) , computer vision , mathematics , statistics , ecology , biology , operating system
Based on Markov hypothesis, IMM uses multiple motion models to match the moving states of the target, and assumes the transfer probability of each model according to the prior knowledge, which has strong adaptability to tracking maneuvering target. However, it is not direct enough to obtain prior knowledge from source data statistics and then make decision according to maximum likelihood, and the information of source data is not fully utilized. Therefore, we use XGBoost in machine learning algorithm to replace this process. We propose XGBoost-IMM model algorithm. XGBoost can fully learn the information of the source data and make decision on the target motion model, and then IMM can perform multi-filter filtering based on the decision. Experimental results show that our algorithm has good performance.