Decision fusion of CNN and SRC with application to SAR target recognition

Layer	Convolution/Pooling kernel	Size of feature map
Input	—	88×88×1
Convolution 1	5×5×20	84×84×20
Pooling 1	2×2×20	42×42×20
Convolution 2	5×5×40	38×38×40
Pooling 2	2×2×40	19×19×40
Convolution 3	4×4×80	16×16×80
Pooling 3	2×2×80	8×8×80
Convolution 4	3×3×160	6×6×160
Pooling 4	2×2×160	3×3×160
Convolution 5	3×3×N	1×1×N
softmax	—	N

Table 2. Training and testing samples under experiment 1: Including 10-class targets

View table

View in Article

Table 2. Training and testing samples under experiment 1: Including 10-class targets

Class	Training		Testing
Class	Elevation angle/(°)	Sample amount	Elevation angle/(°)	Sample amount
BMP2	17	214	15	174
BTR70		214		175
T72		213		175
T62		278		256
BRDM2		277		257
BTR60		234		174
ZSU23/4		278		249
D7		278		249
ZIL131		278		249
2S1		278		249

Table 3. Average recognition rates under experiment 1
View table
View in Article
Table 3. Average recognition rates under experiment 1
Recognition method Ours SVM SRC CNN
Average recognition rate 99.36% 98.64% 98.23% 99.08%

Table 4. Training and testing samples under experiment 2: Including 3-class targets

View table

View in Article

Table 4. Training and testing samples under experiment 2: Including 3-class targets

Class	Training			Testing
Class	Elevation angle/(°)	Configuration	Sample amount	Elevation angle/(°)	Configuration	Sample amount
BMP2	17	9 563	214	15	9 566	175
BMP2		9 563	214		c21	175
BTR70		c71	214		c71	175
T72		132	213		812	174
T72		132	213		s7	167

Table 5. Average recognition rates under experiment 2
View table
View in Article
Table 5. Average recognition rates under experiment 2
Recognition method Ours SVM SRC CNN
Average recognition rate 95.42% 92.58% 92.14% 93.96%

Table 6. Training and testing samples under experiment 3: Including 3-class targets
View table
View in Article
Table 6. Training and testing samples under experiment 3: Including 3-class targets
Class Training Testing
Elevation angle/(°) Sample amount Elevation angle/(°) Sample amount
2S1 17 277 30 267
45 285
BRDM2 276 30 266
45 285
ZSU23/4 277 30 267
45 285

Table 7. Average recognition rates under experiment 3
View table
View in Article
Table 7. Average recognition rates under experiment 3
Recognition method Ours SVM SRC CNN
Average recognition rate 30° 97.56% 94.52% 95.87% 97.04%
45° 71.64% 66.64% 65.42% 67.56%

Tools

Get Citation

Copy Citation Text

Jianhua Lu. Decision fusion of CNN and SRC with application to SAR target recognition[J]. Infrared and Laser Engineering, 2022, 51(3): 20210421

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image processing

Received: Dec. 20, 2021

Accepted: --

Published Online: Apr. 8, 2022

The Author Email:

DOI:10.3788/IRLA20210421

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Descriptions of different layers in CNN

Table 1. Descriptions of different layers in CNN

Table 2. Training and testing samples under experiment 1: Including 10-class targets

Table 2. Training and testing samples under experiment 1: Including 10-class targets

Table 3. Average recognition rates under experiment 1

Table 3. Average recognition rates under experiment 1

Table 4. Training and testing samples under experiment 2: Including 3-class targets

Table 4. Training and testing samples under experiment 2: Including 3-class targets

Table 5. Average recognition rates under experiment 2

Table 5. Average recognition rates under experiment 2

Table 6. Training and testing samples under experiment 3: Including 3-class targets

Table 6. Training and testing samples under experiment 3: Including 3-class targets

Table 7. Average recognition rates under experiment 3

Table 7. Average recognition rates under experiment 3