中文摘要：

水声信号被动检测中广泛使用LOFAR图对接收信号进行处理和分析。针对LOFAR图中线谱信号检测问题,根据线谱信号特征设计特征函数,提出频域滑动窗线谱特征累积检测法。该方法在频率轴移动观察窗,用多步决策算法计算每个观察窗的最优解,得到最优路径,如果最优路径特征值大于阈值,则累积LOFAR图像素点被该最优路径经过的次数,次数越多对应点为线谱点的概率越大。仿真研究表明,该方法对频率时变、低信噪比的线谱信号具有良好的检测能力,可实现多根线谱的增强与检测。海试数据处理结果证明了该方法的可行性和稳健性。该算法对于辐射线谱信号的水下目标远距离探测识别有较高的参考价值。

英文摘要：

In field of passive signal detection, lofargram is widely used to process and analyze the received noise signal. Aiming at the problem of lofargram spectrum line detection, a new method named frequency domain accumulate de- tection algorithm based on spectrum line feature is designed. The key point of the method is moving observation window in the frequency axis; using the multistage decision algorithm to calculate the optimal solution of every window; and accumulating the counter of every image point. That is, if the function value of the optimal path is greater than thre- shold, then the counter of every image point on the optimal path increases one. The image point has higher probability to be a spectrum line point if the number of the point＇s counter is larger. Simulation research indicates that this method can detect several spectrum lines correctly when the SNR is low on the lofargram. The method is verified with sea trial data. So it is valuable in the field of remote detection of underwater radiant target.