为了实现对高速运动目标的快速识别, 设计了一种基于尾焰光谱分布特征分析的识别系统。系统由跟踪成像模块和光谱分析模块组成, 跟踪成像模块用于对准目标, 光谱分析模块用于目标识别。在此基础上, 研究了基于多特征波段峰均值与谷均值比例的区分因子识别算法。该算法设置了有效区分区间, 并通过区分因子比例关系实现高速目标的识别。分析了目标高速运动对光谱获取的影响, 给出了决定光谱偏移量大小的径向和切向速度的函数关系。实验采用少量火箭弹燃烧部作为被测目标, 在距探测系统 0.5 km、1.0 km、2.0 km以及 4.0 km处分别以 10.0 m/s～100.0 m/s的速度进行光谱探测实验。实验结果显示, 对于不同速度的被测目标而言, 在相等的采样周期内系统检测的光谱分布产生了明显的光谱偏移, 但其光谱分布形态基本一致; 对于不同测试距离而言, 距离越远能量衰减越强, 虽然光谱整体振幅差异很大, 但光谱形态无明显变化。通过计算相应波段上峰均值和谷均值的区分因子与区分区间的包含关系, 完成了对被测目标的有效识别。

In order to realize the fast identification of high speed moving targets, a target recognition system was designed based on the characteristics analysis of spectrum distribution of plume. The system consists of tracking imaging module and spectral analysis module. The tracking module is used to align the target and the spectral analysis module is used for target recognition. On this basis, the identification algorithm by distinguishing factor had been studied, and it’s based on the ratio of peak mean and valley mean of multi-feature band. The valid distinction interval was set in the algorithm, and the high-speed target recognition was realized. It was analyzed the effect of high velocity of target impact on spectrum acquisition, and it was given the function of the radial and tangential velocity to decide how much amount of spectrum offset. In the experiments, a small amount of rocket combustion section was used as a test target at the 0.5 km, 1.0 km, 2.0 km and 4.0 km away from the detection system, and spectrum detection experiments were carried out separately at a speed of approximately 10.0 m/s to 100.0 m/s. Experimental results show that it produced a significant shift for the spectrum distribution in the equivalent sampling period for the different speeds measured target, but it had basically the same form of spectrum distribution. For the different test distance, the energy is weaker when the distance farther. Although the overall amplitude of the spectrum is very different, the spectral shape does not change significantly. By calculating inclusion relationship with the distinction factor and distinguish range on the respective band, effective recognition of the measured object was completed.