Synthetic Aperture Radar Image Change Detection Based on Difference Image Construction of Log-Hyperbolic Cosine Ratio and Multi-Region Feature Convolution Extreme Learning Machine

Zhikang Lin; Wei Liu; Chaoyang Niu; Gui Gao; Wanjie Lu

doi:10.3788/AOS221491

Acta Optica Sinica, Volume. 43, Issue 12, 1228001(2023)

Synthetic Aperture Radar Image Change Detection Based on Difference Image Construction of Log-Hyperbolic Cosine Ratio and Multi-Region Feature Convolution Extreme Learning Machine

Zhikang Lin¹, Wei Liu^1、*, Chaoyang Niu¹, Gui Gao², and Wanjie Lu¹

Author Affiliations

¹School of Data and Target Engineering, PLA Strategic Support Force Information Engineering University,Zhengzhou 450000, Henan, China

²Faculty of Geosciences and Environmental Engineering, Southwest Jiaotong University, Chengdu 611756, Sichuan, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(14)

Fig. 1. Flow chart of proposed method

Download full size

Fig. 2. Generation process of log-hyperbolic cosine ratio (LHCR) difference image (DI)

Download full size

Fig. 3. Multi-region feature extraction

Download full size

Fig. 4. Multi-temporal SAR images and their change references. The first row is Yellow River dataset A, the second row is Yellow River dataset B, the third row is Zhengzhou flood dataset C, and the fourth row is Zhengzhou flood dataset D. (a) Images at time 1; (b) images at time 2; (c) change reference images

Download full size

Fig. 5. Final change detection images of Yellow River dataset A. (a) NR_ELM; (b) GarborPCANet; (c) CWNN; (d) DDNet;(e) LHCR_MRFCELM; (f) change reference image

Download full size

Fig. 6. Final change detection images of Yellow River dataset B. (a) NR_ELM; (b) GarborPCANet; (c) CWNN; (d) DDNet;(e) LHCR_MRFCELM; (f) change reference image

Download full size

Fig. 7. Final change detection images of Zhengzhou flood dataset C. (a) NR_ELM; (b) GarborPCANet; (c) CWNN; (d) DDNet;(e) LHCR_MRFCELM; (f) change reference image

Download full size

Fig. 8. Final change detection images of Zhengzhou flood dataset D. (a) NR_ELM; (b) GarborPCANet; (c) CWNN; (d) DDNet;(e) LHCR_MRFCELM; (f) change reference image

Download full size

Fig. 9. Difference image analysis chart

Download full size

Fig. 10. Results of three DI. (a) Proposed LHCR DI; (b) log-ratio DI; (c) neighborhood-based ratio DI

Download full size

Fig. 11. Change detection results of NR_ELM and MRFCELM for three DI generation methods. (a) LHCR_ELM, Kappa value is 0.6364; (b) LR_ELM, Kappa value is 0.6473; (c) NR_ELM, Kappa value is 0.5967; (d) LHCR_MRFCELM, Kappa value is 0.6950; (e) LR_MRFCELM, Kappa value is 0.3973; (f) NR_ MRFCELM, Kappa value is 0.6454

Download full size

Fig. 12. Relationship between different sample block sizes and PCC

Download full size

Table 1. SAR dataset parameters

View table

Table 1. SAR dataset parameters

Dataset	Yellow River dataset A	Yellow River dataset B	Zhengzhou flood dataset C	Zhengzhou flood dataset D
Sensor	Radarsat-2	Radarsat-2	Gaofen-3	Gaofen-3
Location	Yellow River， China	Yellow River， China	Zhengzhou， China	Zhengzhou， China
Data	2008.06 2009.06	2008.06 2009.06	2021.07.20 2021.07.24	2021.07.20 2021.07.24
Size	296 $\times$ 184	233 $\times$ 356	300 $\times$ 300	300 $\times$ 300
Change	Water fades	Water flooding and water fades	Flood disaster	Water fades and farm irrigation

Table 2. Evaluation metrics of different methods on four datasets

View table

Table 2. Evaluation metrics of different methods on four datasets

Method	Yellow River dataset A
Method	FN	FP	OE	PCC /%	k /%	t /s
NR_ELM^［13］	9	926	935	98.28	55.04	3.9
GarborPCANet^［23］	3	2078	2081	96.19	35.34	1755.3
CWNN^［21］	9	1667	1676	96.92	40.24	281.4
DDNet^［22］	21	2492	2513	95.39	30.19	287.8
LHCR_MRFCELM	65	100	165	99.70	86.44	20.7
Method	Yellow River dataset B
Method	FN	FP	OE	PCC /%	k /%	t /s
NR_ELM^［13］	944	221	1165	98.60	74.73	8.6
GarborPCANet^［23］	178	2049	2227	97.32	68.34	3472.8
CWNN^［21］	111	10162	10273	87.62	29.99	259.7
DDNet^［22］	151	4856	5007	93.96	48.28	312.1
LHCR_MRFCELM	603	696	1299	98.43	75.82	21.7
Method	Zhengzhou flood dataset C
Method	FN	FP	OE	PCC /%	k /%	t /s
NR_ELM^［13］	3127	16	3143	96.51	76.44	5.6
GarborPCANet^［23］	3257	13	3270	96.37	75.29	1593.9
CWNN^［21］	1989	108	2097	97.67	85.37	282.4
DDNet^［22］	2886	112	2998	96.67	77.98	266.4
LHCR_MRFCELM	1633	390	2023	97.75	86.37	19.6
Method	Zhengzhou flood dataset D
Method	FN	FP	OE	PCC /%	k /%	t /s
NR_ELM^［13］	3388	537	3925	95.63	59.67	5.8
GarborPCANet^［23］	3385	374	3759	95.82	59.38	1537.3
CWNN^［21］	433	15282	15715	82.54	36.08	319.4
DDNet^［22］	730	7136	7866	91.26	54.63	352.6
LHCR_MRFCELM	2065	1362	3427	96.19	69.50	21.2

Tools

Get Citation

Copy Citation Text

Zhikang Lin, Wei Liu, Chaoyang Niu, Gui Gao, Wanjie Lu. Synthetic Aperture Radar Image Change Detection Based on Difference Image Construction of Log-Hyperbolic Cosine Ratio and Multi-Region Feature Convolution Extreme Learning Machine[J]. Acta Optica Sinica, 2023, 43(12): 1228001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Remote Sensing and Sensors

Received: Jul. 18, 2022

Accepted: Sep. 22, 2022

Published Online: Jun. 20, 2023

The Author Email: Liu Wei (greatliuliu@163.com)

DOI:10.3788/AOS221491

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology