摘要
采用Relax算法提取雷达目标参数,选择散射点强度及其相对位置作为模式识别特征;用高斯混合模型来逼近模式特征的概率分布密度;通过隐马尔可夫状态描述雷达目标不同方位之间的概率关联。研究了飞机类目标的识别问题。实测数据的计算机仿真结果表明,当训练数据与识别测试数据取之完全不同数据段时,三类飞机的100次识别的平均识别率分别为76%,74%,83%。
The Relax algorithm is utilized to extract the parameters of radar targets. The relative location and the amplitude of point scattering are chosen as the features of pattern recognition; Probability relationship in different orientation of radar targets is associated to the states of HMMs. The research into the recognition of airplane-like targets is made with the technique above; While the observation state probability of HMM is simulated by a Gaussian mixture model. Computer simulations with radar data show that the averaged recognition rate of three types of airplane reach 76% （Yake-42）, 74%（An-26）,83% （Cessna） respectively, in 100 simulations, with the training data and recognition data coming from different data segments.
出处
《机电一体化》
2009年第10期69-72,共4页
Mechatronics
