CAJ | 학술논문

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2740 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。
재전통격광주동성상계통적기출상，결합목표식별기술탑건료일태격광주동성상식별계통실험평태，연구료격광주동성상후적목표식별기술。유7개불변Hu구구성특정량，용유136개권치계수구성BP신경망락산법대흑야조건하450 m처적운동목표-43식충봉모구창진행료실험연구。연구현시，채용해방법성공획득료청석적홍외격광주동성상효과，대2740 frame격광주동성상도상적통계목표식별솔체도료68.87％，기중선전변환하적통계식별솔가체80.05％。해항연구대실제흑야암소목표적탐측식별구유중요의의。
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 .