基于正弦高斯混合模型的磁目标通用快速检测

在诸如潜艇等磁目标的探测问题中,被测磁目标信号可能表现出参数待定的磁异信号或极低频磁信号的特征。提出了一种可同时检测参数待定的微弱磁异信号和微弱极低频磁信号的通用快速检测方法。通过抓住磁异信号和极低频磁信号在统计上共同具有的分段类正弦特征,建立了正弦高斯混合模型,实现了对参数待定的磁异信号和极低频磁信号的通用建模。基于正弦高斯混合模型和序贯检测理论构建了一种累积和检测器,实现了对微弱磁异信号和微弱极低频磁信号的通用快速检测,给出了未知模型参数的确定方法,分析了所构建检测器的序贯检测性能。并进一步研究了所构建检测器对不同的微弱磁异信号和微弱极低频磁信号的检测性能。在以地磁为背景的磁目标探测系统实验中,验证了所构建检测器对微弱磁异信号和微弱极低频磁信号的通用性和快速性。实验表明,所构建检测器的信噪比可低至-8 dB,计算量相比于传统检测方法降低4个量级。

Universal fast detection for magnetic objects based on sine Gauss mixture model

The signal characteristic of a magnetic object, such as a submarine, is indeterminate in the magnetic object detection, which may appear as a magnetic anomaly signal or an extremely low frequency magnetic signal with undetermined coefficients. This paper proposes a universal fast detection method for both the weak magnetic anomaly signal and the weak extremely low frequency magnetic signal with undetermined coefficients. A sine Gauss mixture model is built based on the piecewise sinusoidal statistical characteristics of the magnetic anomaly signal and extremely low frequency magnetic signal. This universal model can represent both the magnetic anomaly signal and extremely low frequency magnetic signal with undetermined coefficients. A detector is developed based on the sine Gauss mixture model and the sequential detection theory, which realizes the universal fast detection for both types of signals mentioned above. The unknown parameters in the detector are identified, and the sequential detection performance of the detector is studied. In addition, the detection performance for different weak magnetic anomaly signals and weak extremely low frequency magnetic signals is analyzed. The universality and rapidity of the designed detector for both the weak magnetic anomaly signal and the weak extremely low frequency magnetic signal are verified based on experiment system with geomagnetism. The experimental results indicate that the signal to noise rate can be -8 dB, the amount of calculation is reduced by 4 orders compared with the traditional detection methods.