面向军事目标识别的DRFCN深度网络设计及实现

Design and implementation of DRFCN in-depth network for military target identification

自动目标识别(ATR)技术一直是军事领域中急需解决的重点和难点。本文设计并实现了一种新的面向军事目标识别应用的DRFCN深度网络。首先,在DRPN部分通过卷积模块稠密连接的方式,复用深度网络模型中每一层的特征,实现高质量的目标采样区域的提取;其次,在DFCN部分通过融合高低层次特征图语义特征信息,实现采样区域目标类别和位置信息的预测;最后,给出了DRFCN深度网络模型结构以及参数训练方法。与此同时,进一步对DRFCN算法开展了实验分析与讨论:1)基于PASCAL VOC数据集进行对比实验,结果表明,由于采用卷积模块稠密连接的方法,在目标识别平均准确率、实时性和深度网络模型大小方面,DRFCN算法均明显优于已有基于深度学习的目标识别算法;同时,验证了DRFCN算法可以有效解决梯度弥散和梯度膨胀问题。2)利用自建军事目标数据集进行实验,结果表明,DRFCN算法在准确率和实时性上满足军事目标识别任务。

Automatic target recognition(ATR) technology has always been the key and difficult point in the military field. This paper designs and implements a new DRFCN in-depth network for military target identification. Firstly, the part of DRPN is densely connected by the convolution module to reuse the features of each layer in the deep network model to extract the high quality goals of sampling area;Secondly, in the DFCN part, we fuse the information of the semantic features of the high and low level feature maps to realize the prediction of target area and location information in the sampling area;Finally, the deep network model structure and the parameter training method of DRFCN are given. Further, we conduct experimental analysis and discussion on the DRFCN algorithm: 1) Based on the PASCAL VOC dataset for comparison experiments, the results show that DRFCN algorithm is obviously superior to the existing algorithm in terms of average accuracy, real-time and model size because of the convolution module dense connection method. At the same time, it is verified that the DRFCN algorithm can effectively solve the problem of gradient dispersion and gradient expansion. 2) Using the self-built military target dataset for experiments, the results show that the DRFCN algorithm implements the military target recognition task in terms of accuracy and real-time.