超宽带生命探测雷达多观测点数据关联处理方法研究

The detection mode of multi-observation-points using UWB Life-Detection radar conforms to the actual requirement of ruin rescue. However, the associative processing and detection of the radar data including radar-echoes and detection-values obtained from the multi-observation-points is difficult. Until now, it cannot be used in the practical application, which makes the rescuers get less life information of the trapped person and leads to the low rescue efficiency. Aiming at this problem, this project firstly combines multi-observation-points radar-echoes to establish the observation model of ruin clutter and determine the corresponding probability distribution, according to the diversity of the vital signs of the trapped person. Based on the probability distribution of ruin clutter, the range, azimuth, respiration frequency and body micro-movement texture are taken as the detection objects and several detection models are established to obtain the accurate and related results of the vital signs. Secondly, the Markov state estimation model is established to associate these results of vital signs. The multi-observation-points associative detection can be realized by revealing the inner connection between the vital signs and the trapped posture, respiration status and body micro-movement status of the trapped person. Finally, by introducing the ruin sparse degree estimation and combining the range and azimuth values, the trapped person’s spatial information is obtained to study the multi-observation-points search strategy of the trapped person. The consequent research outcome will bring broad application prospect, and play an important supporting role on the development of national life-detection radar technology and the high rescue efficiency.

超宽带生命探测雷达的多观测点探测模式是符合废墟救援实际需要的工作方式，但雷达的多观测点数据（包括雷达回波和检测值）关联检测处理难度大，至今没有获得实际应用，使得救援人员无法获取更多被困人员生命信息，导致救援效率降低。本项目针对该问题，首先，依据被困人员生命特征多样性特点，综合多观测点雷达回波建立废墟杂波观测模型并确定其概率分布，基于此将距离、方位、呼吸频率、肢体微动纹理作为生命特征检测对象，建立单观测点下的多种检测模型并确保检测值的准确性与相关性。其次，通过建立Markov状态估计模型关联这些检测值，揭示其与被困人员受困姿势、呼吸状态及肢体微动状态之间的内在联系，实现多观测点关联检测。最后，引入废墟稀疏程度估计，并联合距离、方位检测值构成被困人员空间信息，研究多观测点的被困人员搜救策略。该研究工作及成果具有广阔应用前景，将对我国生命探测雷达技术发展，提高救援效率，起着非常重要的支撑作用。

本项目旨在围绕超宽带生命探测雷达的多观测点探测的工作模式，探索并突破多观测点数据关联处理方法，以期提高被困人员搜救效率，并具备一定实用条件。由于废墟救援数据处理方法关注的重点是微弱信号检测能力，本项目主要围绕多观测点微弱信号数据增强、微弱生命特征检测模型及检测算法、被困人员搜救策略等方面开展研究工作，并取得重要研究进展与学术成果。在微弱信号数据增强方面，由于超宽带生命探测雷达B-Scan图像中包含的生命特征信号非常微弱，必须进行增强处理，故引入循环归一化增强算法实现整个采集时窗内的回波信号均得到不同程度的增强，通过设计距离向带通滤波器和慢时滑动平均滤波器对B-Scan图像滤波，以提高微弱生命特征在B-Scan图像中的对比度，便于后续检测处理。在微弱生命特征检测模型及检测算法方面，分别围绕微弱生命特征检测和基于多观测点成像检测模型的生命特征方位检测两个核心，在微弱信号预处理增强方法基础上，开展高效的微弱呼吸信号检测算法的设计开发工作。在被困人员搜救策略方面，围绕“多次探测，一次判定”和“多次判定，最终确定”的被困人员搜救策略，实现综合多次探测结果，初步给出救援区域判定结果，再综合多次判定结果，最终确定救援位置的搜救目的，进而在一定程度上保证稳定性以提高救援效率。该研究工作及成果在多观测点探测方面具有广阔应用前景，可以相应在不同频段雷达探测领域进行拓展应用，提高探测数据处理效率，发挥多观测点探测威力。