{"title":"Multitarget track-before-detect from image observations based on multi-object particle PHD filter","authors":"Ran Zhu, Yunli Long, Zhichao Sha, W. An","doi":"10.1109/PIERS.2017.8262281","DOIUrl":null,"url":null,"abstract":"In order to deal with more complicated situations such as closely spaced objects and target crossings, we propose a recursive multitarget TBD algorithm for image observations based on multi-object particle PHD (MOP-PHD) filter. Instead of sampling from the single target PHD intensity, multi-object set particle sampling is utilized in the approximation of predicted multi-object density. Update of the multi-object state incorporates the multi-object set likelihood function corresponding to a more general observation model to accommodate the overlapping illumination of closely spaced point targets. Each multi-object set particle contains random number of possible single target states, and thus combined with the generalized observation model, the effect of multi-object states can be taken into account simultaneously during the multi-object measurement update procedure. Based on the standard Sequential Monte Carlo PHD (SMC-PHD) filter, multi-object particle PHD filter for image observations is developed and evaluated. Simulation results demonstrate that the proposed method can achieve more accurate estimation without the restriction of non-overlapping assumption, especially when the moving targets become closely spaced.","PeriodicalId":387984,"journal":{"name":"2017 Progress In Electromagnetics Research Symposium - Spring (PIERS)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2017-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2017 Progress In Electromagnetics Research Symposium - Spring (PIERS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/PIERS.2017.8262281","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1
Abstract
In order to deal with more complicated situations such as closely spaced objects and target crossings, we propose a recursive multitarget TBD algorithm for image observations based on multi-object particle PHD (MOP-PHD) filter. Instead of sampling from the single target PHD intensity, multi-object set particle sampling is utilized in the approximation of predicted multi-object density. Update of the multi-object state incorporates the multi-object set likelihood function corresponding to a more general observation model to accommodate the overlapping illumination of closely spaced point targets. Each multi-object set particle contains random number of possible single target states, and thus combined with the generalized observation model, the effect of multi-object states can be taken into account simultaneously during the multi-object measurement update procedure. Based on the standard Sequential Monte Carlo PHD (SMC-PHD) filter, multi-object particle PHD filter for image observations is developed and evaluated. Simulation results demonstrate that the proposed method can achieve more accurate estimation without the restriction of non-overlapping assumption, especially when the moving targets become closely spaced.