红外激光主动成像和识别  

Infrared laser active imaging and recognition technology

作　　者：赵建川 王弟男[2] 陈长青[2] 郭劲[2] 

机构地区：[1]海军驻长春地区航空军事代表室,吉林长春130033 [2]中国科学院长春光学精密机械与物理研究所激光与物质相互作用国家重点实验室,吉林长春130033

出　　处：《中国光学》2013年第5期795-802,共8页Chinese Optics

基　　金：吉林省重大科技攻关专项资金资助项目(No.20120615)

摘　　要：在传统激光主动成像系统的基础上,结合目标识别技术搭建了一台激光主动成像识别系统实验平台,研究了激光主动成像后的目标识别技术。由7个不变Hu矩构成特征量,用由136个权值系数构成BP神经网络算法对黑夜条件下450 m处的运动目标—43式冲锋模具枪进行了实验研究。研究显示,采用该方法成功获得了清晰的红外激光主动成像效果,对2 740 frame激光主动成像图像的统计目标识别率达到了68.87%,其中旋转变换下的统计识别率可达80.05%。该项研究对实际黑夜暗小目标的探测识别具有重要意义。An experiment platform for laser active imaging and target recognition was built combining a laser active image system and the target recognition technology, and the target recognition after laser active imaging was mainly researched. The feature vector was comprised of seven invariant Hu moments. The BP neural net-work algorithm comprised of 136 weight coefficients was used to study the moving target, a 43 submachine gun model at 450 m from the experiment platform at night, and excellent experiment results were obtained. It shows clear imaging effects by 68.87 % of target recognition statistic probability in 2 740 frames of laser active imaging, and the probability of rotation transformation reaches 80. 05 %. These researches are significant to the detection and recognition of little targets at night.

关 键 词：激光主动成像 红外成像 目标识别 成像距离模型 Hu矩特征量 BP神经网络 

分 类 号：TN249[电子电信—物理电子学] TP391[自动化与计算机技术—计算机应用技术]