SBIR Phase I: Using Hidden Markov Models to Track Human Targets

SBIR 第一阶段：使用隐马尔可夫模型跟踪人类目标

• 批准号: 9760988
• 负责人: Scott Thompson
• 金额: $ 9.97万
• 依托单位: ChT Engineering Systems
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1998
• 资助国家: 美国
• 起止时间: 1998-01-01 至 1998-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9760988&HistoricalAwards=false
• 关键词: SBIR; Phase; Using; Hidden; Markov

This Small Business Innovation Research Phase I project will investigate the feasibility Of implementing a novel multi-sensor data fusion technique for low cost target tracking. Literature suggests that Hidden Markov Models significantly out perform other methods of multiple target tracking when ambiguous detections, noise, and clutter are present. The primary objective here is to demonstrate the feasibility of utilizing Hidden Markov Models to track human targets. The target tracking problem can be viewed as one of developing a model that minimizes the difference between the model's estimate of the target coordinates {Xout(t), Yout(t)} and the actual coordinates given by {Xactual(t), Yactual(t)}. In practice, this must be done simultaneously for a varying number of independent targets. However, this Phase I effort will focus on single target tracking and leave multiple target tracking to the subsequent Phase II project. Portable code used to implement models in Phase I will be incorporated in a microprocessor-based system following multiple target refinement in Phase II. Synthesized sensor data will be used initially to develop tracking models. These data will then be supplemented with measured data from ultrasonic sensors, in order to further refine the tracking model. Direct applications of the human tracking system include video surveillance for security, safety, and human or animal behavioral observation systems. With spotlight attachments the device will provide an illuminated path for remote search and rescue operations, emergency evacuation in poorly lit buildings, crime deterrence, and night surveillance tracking of escaping prisoners. This system will also be well suited as a supplement to a video target tracking and identification system. This award is supported by the Small Business Innovation Research Program and the Experimental Program to Stimulate Competitive Research (EPSCoR).

该小型企业创新研究第一阶段项目将研究实施新型多传感器数据融合技术以实现低成本目标跟踪的可行性。 文献表明，当存在模糊检测、噪声和杂波时，隐马尔可夫模型的性能显着优于其他多目标跟踪方法。 这里的主要目标是证明利用隐马尔可夫模型跟踪人类目标的可行性。 目标跟踪问题可以看作是开发一个模型，最小化模型估计的目标坐标 {Xout(t), Yout(t)} 与由 {Xactual(t), Yactual( t)}。 实际上，必须对不同数量的独立目标同时完成此操作。 然而，第一阶段的工作将集中于单目标跟踪，并将多目标跟踪留给后续的第二阶段项目。 在第二阶段的多目标细化之后，用于实现第一阶段模型的可移植代码将被合并到基于微处理器的系统中。 合成的传感器数据最初将用于开发跟踪模型。 然后，这些数据将补充超声波传感器的测量数据，以进一步完善跟踪模型。 人体跟踪系统的直接应用包括用于安全、保障和人类或动物行为观察系统的视频监控。 借助聚光灯附件，该设备将为远程搜索和救援行动、光线不足的建筑物中的紧急疏散、威慑犯罪以及对逃跑囚犯的夜间监视跟踪提供照明路径。 该系统也非常适合作为视频目标跟踪和识别系统的补充。 该奖项由小型企业创新研究计划和刺激竞争研究实验计划 (EPSCoR) 提供支持。