Remote Sensing Image Segmentation Network Based on Multi-Level Feature Refinement and Fusion

Table 1. Vaihingen training sample pixel category distribution
View table
Table 1. Vaihingen training sample pixel category distribution
Class Impervious surface Building Low vegetation Tree Car Background
Pixel percentage /% 29.03 26.69 20.06 22.08 1.21 0.93

Table 2. Object classification accuracy on Vaihingen dataset

View table

Table 2. Object classification accuracy on Vaihingen dataset

Method	IoU					Mean IoU	Mean F₁	Overall accuracy
Method	Impervious surface	Building	Low vegetation	Tree	Car	Mean IoU	Mean F₁	Overall accuracy
PSPNet	77.26	87.89	65.67	77.48	55.86	72.83	85.08	87.83
UNet	77.45	84.42	66.03	76.24	58.65	73.29	85.57	87.97
RefineNet	82.14	87.12	66.69	74.57	56.87	73.48	85.82	88.16
Deeplab v3+	83.77	89.92	67.93	78.67	62.46	76.55	87.90	89.65
MRFNet (proposed）	84.37	90.15	68.04	79.21	63.49	77.05	88.24	90.34

Table 3. Object classification accuracy on Potsdam dataset
View table
Table 3. Object classification accuracy on Potsdam dataset
Method IoU Mean IoU Mean F₁ Overall accuracy
Impervious surface Building Low vegetable Tree Car
MRFNet（proposed） 84.37 90.15 68.04 79.21 63.49 77.05 88.24 91.47

Tools

Get Citation

Copy Citation Text

Yongsheng Jian, Daming Zhu, Zhitao Fu, Shiya Wen. Remote Sensing Image Segmentation Network Based on Multi-Level Feature Refinement and Fusion[J]. Laser & Optoelectronics Progress, 2023, 60(4): 0428002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites