多尺度特征提取和多层次注意力机制的迷彩伪装目标分割算法  

Camouflage Target Segmentation Algorithm Using Multi-Scale Feature Extraction and Multi-Level Attention Mechanism

作　　者：梁新宇 权冀川 杨辉 肖铠鸿 王中伟 Liang Xinyu;Quan Jichuan;Yang Hui;Xiao Kaihong;Wang Zhongwei(College of Command&Control Engineering,Army Engineering University of PLA,Nanjing 210007;Unit 73676 of PLA,Wuxi 214400;Unit 73658 of PLA,Chuzhou 239000)

机构地区：[1]陆军工程大学指挥控制工程学院,南京210007 [2]中国人民解放军73676部队,无锡214400 [3]中国人民解放军73658部队,滁州239000

出　　处：《计算机辅助设计与图形学学报》2022年第5期683-692,共10页Journal of Computer-Aided Design & Computer Graphics

摘　　要：针对现阶段语义分割算法在野外战场环境中对迷彩伪装目标分割边界不理想、小目标分割精度低的问题,提出一种结合多尺度特征提取和多层次注意力机制的迷彩伪装目标语义分割算法CSS-Net.该算法由编码结构和解码结构组成.编码部分利用轻量化的深度可分离卷积联合残差结构构建特征编码器,对迷彩伪装目标图像进行特征提取;解码部分设计了策略选择的多尺度特征融合模块和多层次注意力特征增强模块,用以获取图像的多尺度信息和通道信息,在逐步精细化分割结果的同时进一步增强图像的语义解码过程.实验结果表明,CSS-Net算法能够有效实现复杂战场环境下伪装目标的分割识别,总体分割的平均交并比指标达到91.98%,分割边界得到改善.与DeepLabv3+算法相比,CSS-Net算法用于迷彩伪装小目标图像分割时的平均交并比指标增长3.71个百分点,对于多尺度目标分割的平均交并比指标均超过85%,分割效果提升明显.There are two problems when semantic segmentation algorithm is used in the battlefield environment:The camouflage target segmentation boundary is not ideal,and the small target segmentation accuracy is low.So a camouflage target semantic segmentation algorithm CSS-Net is proposed,which combines multi-scale feature extraction method and multi-level attention mechanism.The algorithm consistes of an encoder-decoder structure.In the encoder part,a lightweight and deep separable convolution with residual structure is used to construct the feature encoder to extract features from the reconnaissance images.In the decoder part,a strategy-selected multi-scale feature fusion module and a multi-level attention feature enhancement module are designed to obtain the multi-scale representation information and the channel information of the images.It furtherly enhances the semantic decoding process while gradually refining the segmentation results.The experiment results show that the CSS-Net algorithm can effectively complete the segmentation and recognition of camouflage targets in complex battlefield environment.The overall mIoU value reaches 91.98%,and the segmentation boundary is improved.Compared with DeepLab v3+algorithm,the mIoU value of CSS-Net on camouflage small targets is increased by 3.71 percents,and the mIoU value on multi-scale targets exceeds 85%.The segmentation effects are improved significantly.

关 键 词：语义分割 迷彩伪装目标 多尺度特征融合 多层次注意力机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]