上海交通大学学报 ›› 2022, Vol. 56 ›› Issue (2): 143-155.doi: 10.16183/j.cnki.jsjtu.2021.037

• • 上一篇    下一篇

自适应插值飞蛾扑火优化的多特征粒子滤波车辆跟踪算法

黄鹤1,2(), 吴琨1,2, 李昕芮1, 王珺2,3, 王会峰1, 茹锋1,2   

  1. 1. 长安大学 电子与控制工程学院, 西安 710064
    2. 长安大学 西安市智慧高速公路信息融合与控制重点实验室, 西安 710064
    3. 西北大学 信息科学与技术学院,西安 710127
  • 收稿日期:2021-02-01 出版日期:2022-02-28 发布日期:2022-03-03
  • 作者简介:黄鹤(1979-),男,河南省南阳市人,副教授,主要研究方向为人工智能与信息融合、图像处理及嵌入式开发. E-mail: huanghe@chd.edu.cn.
  • 基金资助:
    国家重点研发计划项目(2018YFB1600600);陕西省重点研发计划项目(2021GY-285);陕西省重点研发计划项目(2021SF-483);陕西省自然科学基础研究计划项目(2021JM-184);长安大学中央高校基本科研业务费专项资金项目(300102329401);长安大学中央高校基本科研业务费专项资金项目(300102329501);西安市智慧高速公路信息融合与控制重点实验室长安大学开放基金项目(300102321502)

A Multi-Feature Particle Filter Vehicle Tracking Algorithm Based on Adaptive Interpolation Moth-Flame Optimization

HUANG He1,2(), WU Kun1,2, LI Xinrui1, WANG Jun2,3, WANG Huifeng1, RU Feng1,2   

  1. 1.School of Electronics and Control Engineering, Chang’an University, Xi’an 710064, China
    2.Xi’an Key Laboratory of Intelligent Expressway Information Fusion and Control, Chang’an University, Xi’an 710064, China
    3.School of Information Science and Technology, Northwest University, Xi’an 710127, China
  • Received:2021-02-01 Online:2022-02-28 Published:2022-03-03

摘要:

针对现有飞蛾扑火优化算法精度低、全局搜索能力差的问题,提出一种自适应插值飞蛾扑火优化算法,并将其嵌入多特征粒子滤波中优化,构建自适应插值飞蛾扑火优化的多特征粒子滤波车辆跟踪算法.首先,在飞蛾的位置更新机制中加入自适应权值,改善所提算法的全局搜索能力.其次,采用改进的插值飞蛾扑火优化算法对采样过程进行优化,结合多特征自适应融合优化粒子滤波车辆跟踪算法,根据最新观测信息不断调整粒子分布,使低权值层粒子向权值较高的区域移动,增强粒子质量,避免样本退化.实验结果表明,本文算法能够有效降低状态预测所需的样本粒子数,提高算法的跟踪性能,在车辆目标发生遮挡、光照、姿态及尺度变化等干扰下仍然能够准确、稳定地跟踪目标车辆.

关键词: 车辆跟踪, 粒子滤波, 多特征融合, 飞蛾扑火

Abstract:

In order to solve the problem of low accuracy and the poor global searching ability of the moth-flame optimization algorithm, an improved adaptive interpolation moth-flame optimization algorithm is proposed, which is embedded into multi-feature particle filter to optimize. Besides, a multi-feature particle filter vehicle tracking algorithm based on adaptive interpolation moth-flame optimization is constructed. First, adaptive weights are added to the moths’ position updating mechanism to improve the global searching ability of the proposed algorithm. Next, the adaptive interpolation moth-flame optimization algorithm is used to optimize the sampling process. Then, in combination with the multi-feature adaptive fusion particle filter vehicle tracking algorithm, the particle distribution according to the latest observation information is continuously adjusted, so that the particles in the low weight layer can move to the area with higher weight to enhance the particle quality and avoid sample degradation. The experimental results show that the proposed algorithm can effectively reduce the number of sample particles required for state prediction, improve the tracking performance of the algorithm, and track the target vehicle accurately and stably under the interferences of occlusion, illumination, attitude, and scale changes.

Key words: vehicle tracking, particle filter, multi-feature fusion, moth-flame

中图分类号: