航空学报 > 2018, Vol. 39 Issue (6): 321828-321828   doi: 10.7527/S1000-6893.2018.21828

基于视觉注意机制的认知雷达数据关联算法

王树亮, 毕大平, 阮怀林, 周阳   

  1. 国防科技大学 电子对抗学院, 合肥 230037
  • 收稿日期:2017-10-25 修回日期:2018-02-09 出版日期:2018-06-15 发布日期:2018-02-09
  • 通讯作者: 毕大平,E-mail:DAPEEI@163.com E-mail:DAPEEI@163.com
  • 基金资助:
    国家自然科学基金(61671453);安徽省自然科学基金(1608085MF123)

Cognitive radar data association algorithm based on visual attention

WANG Shuliang, BI Daping, RUAN Huailin, ZHOU Yang   

  1. Institute of Electronic Countermeasure, National University of Defence Technology, Hefei 230037, China
  • Received:2017-10-25 Revised:2018-02-09 Online:2018-06-15 Published:2018-02-09
  • Supported by:
    National Natural Science Foundation of China (61671453); Anhui Province Natural Science Fund Project (1608085MF123)

摘要: 针对传统关联波门设计方法在应用于机动目标跟踪时容易引起失跟、以及概率数据关联算法不适于多交叉目标跟踪的问题,提出了一种基于人类视觉选择性注意机制和知觉客体的"特征整合"理论的认知雷达数据关联算法。算法以综合交互式多模型概率数据关联算法为基础,采取假设目标最大机动水平已知的"当前"统计模型和匀速运动模型作为模型集,通过实时交互使关联波门能够随目标机动动态调整,较好地兼顾了雷达计算耗时和跟踪成功率。在利用目标位置特征的基础上,进一步提取、整合目标运动特征,对关联波门交叉区域公共量测进行分类,使多交叉目标跟踪问题转化为多个单目标跟踪问题,优化了传统概率数据关联算法。仿真结果表明:与传统关联波门设计方法相比,算法跟踪失败率和计算耗时明显降低;而且在计算资源增加不大的情况下,杂波环境适应性也得到了显著增强。

关键词: 视觉注意, 认知雷达, 数据关联, 特征整合, 机动目标跟踪

Abstract: Maneuvering target tracking loss often occurs with the traditional association gate design method, and the probabilistic data association algorithm is not fit for multi-cross target tracking. To overcome these problems, a cognitive radar data association algorithm is proposed based on the human visual attention model and the perceptual object "feature integration" theory. The algorithm is based on the comprehensive interacting multiple model probabilistic data association algorithm, and uses the "current" statistical model, in which the maximum maneuvering level is known, and the uniform motion model as the model set. The association gate can be adjusted via target maneuvering, balancing the radar computing time and tracking success rate. Based on the features of target location, the motion features of the target are extracted and integrated to classify the public measure in the cross area of the association gate. The multi-cross target tracking problem is then transferred into the problem of tracking of multiple single targets, so the traditional probabilistic data association algorithm is optimized. Simulation results show that the proposed algorithm can significantly reduce the tracking failure rate and computing time, and the optimized data association algorithm with slightly more time consumption is more environment adaptable than the traditional algorithm.

Key words: visual attention, cognitive radar, data association, feature integration, maneuvering target tracking

中图分类号: