基于深度卷积神经网络的蛇形机动航迹图像识别  

Deep Convolutional Neural Networks for the Image Recognition of“S-Maneuver”Target

作　　者：郑昌艳 梅卫[1] 王刚[1] 

机构地区：[1]军械工程学院,石家庄050003

出　　处：《火力与指挥控制》2016年第5期66-70,共5页Fire Control & Command Control

基　　金：国防"十一五"预研基金(40405020204);国防"十二五"预研基金资助项目(40405070102)

摘　　要：为提高防空武器系统对空袭目标的拦截防御能力,针对现有蛇形机动识别算法鲁棒性较差的问题,提出了将航迹坐标数据转化为图像,利用深度神经卷积神经网络进行航迹模式分类的方法。针对航迹数据直接转化为图像时存在机动幅度不明显或过大的问题,提出了有效解决方案。基于CAFFE平台进行了大量仿真实验,确定了适宜于航迹模式分类的深度卷积网络结构和网络参数。实验结果表明,该方法能有效提高蛇形机动航迹识别的鲁棒性。To improve the interception ability of anti -aircraft weapons and solve the existingrobustness problem in the “S-maneuver”target recognition algorithm,the tracks coordinate data intoimages are transformed and then took advantage of deep convolutional neural networks to recognize it.The feasible method to solve the maneuver range inconformity problem is proposed which existed as thecoordinate data transformed into images. The suitable deep convolutional architecture and networkparameters have been identified after plenty of experiments based on CAFFE software platform. It isproved to be an efficient method to improve the robustness of target“S-maneuver”recognition.

关 键 词：蛇形机动 图像识别 深度卷积神经网络 CAFFE 

分 类 号：TP271[自动化与计算机技术—检测技术与自动化装置]