论文部分内容阅读
污染云团光谱自动鉴别报警中的虚警是十分重要而必需研究解决的问题。理论上，满足分类方程的解是无限的。对未参与训练的干扰谱，特别是背景与系统噪声，由于其波形的随机性，有很大的概率满足污染云团光谱的判决条件，从而造成完全随机的虚警。根据对光谱频谱的分析，本文提出了一种判决算法。经系统动态训练选择出判决阈值，解决了虚警问题，从而在外场鉴别试验中获得圆满成功。
It is very important and must research and solve the problem that the false alarm in the automatic identification of pollution cloud spectrum is alarmed. In theory, the solution to the classification equation is infinite. Due to the randomness of their waveforms, there is a great probability that the spectrum of interferences that do not participate in the training, especially the background and system noise, satisfy the decision criteria of contaminated cloud spectra, resulting in a completely random false alarm. Based on the analysis of spectrum, this paper presents a decision algorithm. Through the systematic dynamic training, the judgment threshold is selected and the false alarm problem is solved, so that it is successful in the field experiment.