Convolutional Neural Network Based Multi-Band Ship Target Recognition with Feature Fusion

Feng Liu; Tongsheng Shen; Xinxing Ma

doi:10.3788/AOS201737.1015002

Acta Optica Sinica, Volume. 37, Issue 10, 1015002(2017)

Convolutional Neural Network Based Multi-Band Ship Target Recognition with Feature Fusion

Feng Liu^1、*, Tongsheng Shen², and Xinxing Ma¹

Author Affiliations

¹ Department of Control Engineering, Naval Aeronautical and Astronautical University, Yantai, Shandong 264001, China

² China Defense Science and Technology Information Center, Beijing 100142, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(9)

Fig. 1. Flow of the proposed algorithm

Download full size

Fig. 2. Diagram of the network structure

Download full size

Fig. 3. Contrast of parameters for different layers

Download full size

Fig. 4. Examples in target identification database. (a) Cruise A; (b) cruise B; (c) fisher; (d) railway ferry; (e) warship; (f) merchant ship

Download full size

Fig. 5. Dimension selection of fusion features

Download full size

Fig. 6. Recognition rate matrices of the proposed method. (a) Proposed fusion recognition; (b) visible light; (c) medium-wave infrared; (d) long-wave infrared

Download full size

Fig. 7. Partial images of false recognition

Download full size

Table 1. Structure and parameters of the neural network

View table

Table 1. Structure and parameters of the neural network

Parameter Layer	Structure of network
Parameter Layer	Conv_1	Conv_2	Conv_3	Conv_4	Fc_5	Fc_6
Structure	C+R+L+P	C+R+P	C+R+P	C+R+P	F+R+D	F+R+D
Input	227×227×3	27×27×96	13×13×256	13×13×384	6×6×256	4096
Neuron	96	256	384	256	4096	4096
Kernal size	11×11	5×5	3×3	3×3	1×1	1×1
Stride	4	2	1	1
Pooling	3×3	3×3		3×3
Pooling stride	2	2		2
Train parameter	96×(11×11×3+1)	256×(5×5×96+1)	384×(3×3×256+1)	256×(3×3×384+1)	4096×(3×3×256+1)	4096×(4096+1)

Table 2. Recognition rate comparison of different methods
View table
Table 2. Recognition rate comparison of different methods
Method Recognition rate /%
Visible light Medium-wave infrared Long-wave infrared Fusion recognition
HOG+SVM 63.2 55.0 53.5
SIFT 67.3 60.2 55.2
AlexNet 75.6 66.5 67.7
VGG-16 77.3 68.3 69.2
Proposed 75.1 67.2 68.1 84.5

Tools

Get Citation

Copy Citation Text

Feng Liu, Tongsheng Shen, Xinxing Ma. Convolutional Neural Network Based Multi-Band Ship Target Recognition with Feature Fusion[J]. Acta Optica Sinica, 2017, 37(10): 1015002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites