弹载融合图像深度卷积网络视觉解释

薛松; 钱立志; 杨传栋

doi:10.15892/j.cnki.djzdxb.2022.05.019

弹箭与制导学报 >

2022 , Vol. 42 >Issue 5: 102 - 107

DOI: https://doi.org/10.15892/j.cnki.djzdxb.2022.05.019

弹载融合图像深度卷积网络视觉解释

薛松 ¹^,² ,
钱立志 ¹^,² ,
杨传栋 ³

展开

¹ 陆军炮兵防空兵学院兵器工程系,合肥 230031
² 陆军炮兵防空兵学院高过载弹药制导控制与信息感知实验室,合肥 230031
³ 陆军炮兵防空兵学院研究生队,合肥 230031

薛松(1989—),男,陕西韩城人,讲师,博士研究生,研究方向:信息化弹药运用、图像智能化处理。

收稿日期: 2022-04-28

网络出版日期: 2025-01-16

收起

Missile Borne Fusion Image Visual Explanations for Deep Convolutional Networks

XUE Song ¹^,² ,
QIAN Lizhi ¹^,² ,
YANG Chuandong ³

Expand

¹ Department of Weapons Engineering,Army Academy of Artillery and Air Defense,Hefei 230031,China
² High Overload Ammunition Guidance Control and Information Perception Laboratory,Army Academy of Artillery and Air Defense,Hefei 230031,China
³ Postgraduate Team,Army Academy of Artillery and Air Defense,Hefei 230031,China

Received date: 2022-04-28

Online published: 2025-01-16

Fold

摘要

近年来,卷积神经网络的决策过程受到了越来越多的关注,其内部运行机制促使研究者们开展了深入研究,并形成了基于显著性映射的视觉解释理论方法。文中提出一种适用于弹载融合图像的深度卷积网络视觉解释方法,该方法通过置信度提升重组神经网络梯度映射,并结合权重参数获得显著图。实验结果表明,与经典的视觉解释方法相比,文中方法具有良好的主观视觉效果,在平均下降和平均提升两类指标上都达到了最优,同时具备较为准确的定位能力。

关键词： 融合图像; 显著性映射; 卷积神经网络; 计算机视觉

本文引用格式

薛松 , 钱立志 , 杨传栋 . 弹载融合图像深度卷积网络视觉解释[J]. 弹箭与制导学报, 2022 , 42(5) : 102 -107 . DOI: 10.15892/j.cnki.djzdxb.2022.05.019

Abstract

In recent years, the decision-making process of convolutional neural networks has attracted more and more attention. Its internal operating mechanism prompts researchers to conduct in-depth research, and forms a visual interpretation theory method based on saliency map. The paper presents a deep convolutional network vision explanations method for missile borne fusion images. In this method, the gradient map of neural network is reconstructed by the “increase of confidence”, and the saliency map is obtained by combining the weight parameters. The experimental results show that compared with the classical visual interpretation methods, the paper method has good subjective visual effect, and achieves the best in the two indicators of average drop and average increase, and has more accurate positioning ability.

Key words： fusion image; saliency map; convolutional neural networks; computer vision

[an error occurred while processing this directive]

参考文献

原文顺序 | 文献年度倒序 | 文中引用次数倒序

[1]	KRIZHEVSKY A, SUTSKEVER I, HINTON G E. Imagenet classification with deep convolutional neural networks. advances[J]. Neural Information Processing Systems, 2012, 25: 1097-1105.

[2]	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]// IEEE. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 770-778.

[3]	HUANG G, LIU Z L, MAATEN V D, et al. Densely connected convolutional networks[C]// IEEE. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2017: 2261-2269.

[4]	REDMON J, FARHADI A. YOLOv3: an incremental improvement[C]// IEEE. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 2767-2773.

[5]	HE K M, ZHANG X Y, REN S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transcations on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.

[6]	SHELHAMER E, LONG J, DARRELL T. Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(4): 640-651. DOI PMID

[7]	SELVARAJU R R, COGSWELL M, DAS A, et al. Grad-CAM: visual explanations from deep networks via gradient-based localization[J]. International Journal of Computer Vision, 2020, 128(2): 336-359.

[8]	ZEILER M D, FERGUS R. Visualizing and understanding convolutional networks[C]// ECCV. Proceedings of the European conference on computer vision. Berlin:Springer, 2014: 818-833.

[9]	ZHOU B L, KHOSLA A, LAPEDRIZA A, et al. Learning deep features for discriminative localization[C]// IEEE. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 2921-2929.

[10]	CHATTOPADHAY A, SARKAR A, HOWLADER P, et al. Grad-CAM++: generalized gradient-based visual explanations for deep convolutional networks[C]// IEEE. Proceedings of the Winter Conference on Applications of Computer Vision. New York: IEEE, 2018: 839-847.

[11]	DESAI S, RAMASWAMY H G. Ablation-CAM: visual explanations for deep convolutional network via gradient-free localization[C]// IEEE. Proceedings of the Winter Conference on Applications of Computer Vision. New York: IEEE, 2020: 983-991.

[12]	FU R G, HU Q Y, DONG, X H, et al. Axiom-based Grad-CAM: towards accurate visualization and explanation of CNNs[EB/OL]. (2020-08-05) [2022-04-15]. https://arxiv.org/abs/2008.02312.

[13]	薛松, 钱立志, 张航, 等. 多源末制导弹载融合图像目标检测研究进展[J]. 弹箭与制导学报, 2021, 41(3):67-75.

[14]	霍星, 邹韵, 陈影, 等. 双尺度分解和显著性分析相结合的红外与可见光图像融合[J]. 中国图象图形学报, 2021, 26(12):2813-2825.

[15]	SURAJ S, FLEURET F. Full-gradient representation for neural network visualization[EB/OL]. (2019-05-02) [2022-04-15]. https://arxiv.org/abs/1905.00780.

[16]	ANG H F, WANG Z F, DU M N, et al. Score-CAM: score-weighted visual explanations for convolutional neural networks[C]// IEEE. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. New York: IEEE, 2020: 111-119.

[17]	VITALI P, ABIR D, KATE S. RISE: randomized input sampling for explanation of black-box models[EB/OL]. (2018-06-19)[2022-04-15]. https://arxiv.org/abs/1806.07421.

[18]	ZHANG J M, ZHE L, BRANDT J, et al. Top-down neural attention by excitation backprop[C]// ECCV. Proceedings of the European Conference on Computer Vision. Berlin:Springer, 2016: 543-559.

Options

文章导航