Classification Method of Remote Sensing Image Based on Dynamic Weight Transform and Dual Network Self Verification

Qingfang Zhang; Ming Cong; Ling Han; Jiangbo Xi; Qingqing Jing; Jianjun Cui; Chengsheng Yang; Chaofeng Ren; Junkai Gu; Miaozhong Xu; Yiting Tao

doi:10.3788/LOP231381

Laser & Optoelectronics Progress, Volume. 61, Issue 8, 0828001(2024)

Classification Method of Remote Sensing Image Based on Dynamic Weight Transform and Dual Network Self Verification

Qingfang Zhang¹, Ming Cong^1、*, Ling Han¹, Jiangbo Xi¹, Qingqing Jing², Jianjun Cui¹, Chengsheng Yang¹, Chaofeng Ren¹, Junkai Gu¹, Miaozhong Xu³, and Yiting Tao³

Author Affiliations

¹College of Geology Engineering and Geomatics, Chang'an University, Xi'an 710054, Shaanxi, Chian

²China Aero Geophysical Survey & Remote Sensing Center for Land and Resources, Beijing 100083, China

³State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, Hubei, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(15)

Fig. 1. The basic structure of dynamic weight deformation

Download full size

Fig. 2. The application of point multiplication of dynamic weight deformation

Download full size

Fig. 3. Diamond feature extraction structure based on residual connection

Download full size

Fig. 4. Classification network structure with dynamic deformation of multi-diamond weights

Download full size

Fig. 5. Schematic diagram of YOLO network improved by weight dynamic deformation

Download full size

Fig. 6. Schematic diagram of the dual model network structure

Download full size

Fig. 7. Experiment 1. (a) Original remote sensing image; (b) feature category composition; (c) GT; (d) results of FCN; (e) results of Attention-Unet; (f) results of MASK-RCNN; (g) results of DWTCN; (h) results of DWTDN

Download full size

Fig. 8. Experiment 2. (a) Original remote sensing image; (b) feature category composition; (c) GT; (d) results of FCN; (e) results of Attention-Unet; (f) results of MASK-RCNN; (g) results of DWTCN; (h) results of DWTDN

Download full size

Fig. 9. Experiment 3. (a) Original remote sensing image; (b) feature category composition; (c) GT; (d) results of FCN; (e) results of Attention-Unet; (f) results of MASK-RCNN; (g) results of DWTCN; (h) results of DWTDN

Download full size

Fig. 10. Experiment 4. (a) Original remote sensing image; (b) feature category composition; (c) GT; (d) results of FCN; (e) results of Attention-Unet; (f) results of MASK-RCNN; (g) results of DWTCN; (h) results of DWTDN

Download full size

Table 1. Classification accuracy of experiment 1

View table

Table 1. Classification accuracy of experiment 1

Classification accuracy of different networks	FCN		Attention-Unet		MASK-RCNN		DWTCN		DWTDN
Classification accuracy of different networks	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa
villa	0.84	0.27	0.88	0.36	0.91	0.43	0.84	0.28	0.90	0.42
building	0.74	0.18	0.86	0.35	0.83	0.32	0.89	0.25	0.87	0.35
vegetation	0.84	0.70	0.94	0.69	0.94	0.71	0.83	0.67	0.95	0.75
road	0.76	0.16	0.57	0.22	0.57	0.25	0.88	0.18	0.67	0.32
water	0.88	0.97	0.98	0.96	0.98	0.97	0.98	0.96	0.99	0.98
bare land	0.77	0.35	0.67	0.40	0.66	0.40	0.77	0.36	0.67	0.41
overall	0.72	0.60	0.70	0.57	0.79	0.69	0.73	0.61	0.82	0.72

Table 2. Classification accuracy of experiment 2

View table

Table 2. Classification accuracy of experiment 2

Classification accuracy of different networks	FCN		Attention-Unet		MASK-RCNN		DWTCD		DWTDN
Classification accuracy of different networks	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa
building	0.67	0.20	0.54	0.15	0.62	0.28	0.64	0.10	0.70	0.31
vegetation	0.82	0.27	0.86	0.41	0.88	0.45	0.89	0.41	0.90	0.48
greenhouses	0.73	0.45	0.82	0.65	0.80	0.60	0.78	0.57	0.83	0.67
bare land	0.70	0.39	0.72	0.34	0.71	0.33	0.74	0.47	0.72	0.34
overall	0.60	0.38	0.69	0.51	0.66	0.47	0.66	0.48	0.71	0.52

Table 3. Classification accuracy of experiment 3

View table

Table 3. Classification accuracy of experiment 3

Classification accuracy of different networks	FCN		Attention-Unet		MASK-RCNN		DWTCN		DWTDN
Classification accuracy of different networks	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa
building	0.93	0.59	0.90	0.43	0.91	0.54	0.92	0.61	0.93	0.60
vegetation	0.88	0.53	0.89	0.52	0.89	0.52	0.88	0.55	0.89	0.54
water	0.49	0.14	0.61	0.24	0.67	0.26	0.48	0.06	0.69	0.28
greenhouses	0.85	0.64	0.89	0.73	0.90	0.75	0.89	0.71	0.91	0.76
bare land	0.72	0.48	0.70	0.56	0.74	0.57	0.75	0.52	0.75	0.57
overall	0.71	0.55	0.66	0.51	0.68	0.53	0.71	0.56	0.72	0.56

Table 4. Classification accuracy of experiment 4

View table

Table 4. Classification accuracy of experiment 4

Classification accuracy of different networks	FCN		Attention-Unet		MASK-RCNN		DWTCN		DWTDN
Classification accuracy of different networks	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa
building	0.69	0.43	0.81	0.63	0.83	0.67	0.72	0.49	0.84	0.68
vegetation	0.80	0.56	0.83	0.64	0.84	0.65	0.81	0.57	0.85	0.68
road	0.52	0.05	0.59	0.25	0.61	0.29	0.62	0.18	0.66	0.32
bare land	0.76	0.01	0.77	0.43	0.77	0.43	0.76	0.01	0.77	0.43
overall	0.59	0.37	0.63	0.48	0.65	0.49	0.62	0.42	0.70	0.53

Table 5. Classification accuracy of all experiments

View table

Table 5. Classification accuracy of all experiments

Classification accuracy of different networks	FCN		Attention-Unet		MASK-RCNN		DWTCD		DWTDN
Classification accuracy of different networks	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa	OA	Kappa
Exp.1	0.72	0.60	0.70	0.57	0.79	0.69	0.73	0.61	0.82	0.72
Exp.2	0.60	0.38	0.69	0.51	0.66	0.47	0.66	0.48	0.71	0.52
Exp.3	0.71	0.55	0.66	0.51	0.68	0.53	0.71	0.56	0.72	0.56
Exp.4	0.59	0.37	0.63	0.48	0.65	0.49	0.62	0.42	0.70	0.53

Tools

Get Citation

Copy Citation Text

Qingfang Zhang, Ming Cong, Ling Han, Jiangbo Xi, Qingqing Jing, Jianjun Cui, Chengsheng Yang, Chaofeng Ren, Junkai Gu, Miaozhong Xu, Yiting Tao. Classification Method of Remote Sensing Image Based on Dynamic Weight Transform and Dual Network Self Verification[J]. Laser & Optoelectronics Progress, 2024, 61(8): 0828001

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: May. 26, 2023

Accepted: Jul. 24, 2023

Published Online: Mar. 15, 2024

The Author Email: Ming Cong (mingc@chd.edu.cn)

DOI:10.3788/LOP231381

CSTR:32186.14.LOP231381

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology