Robust Semantic Segmentation Model for Greenhouses Based on Remote Sensing Feature Optimization

Table 1. Number of images in the dataset
View table
Table 1. Number of images in the dataset
Month Number of images Month Number of images
January 97 July 78
February 70 August 78
March 78 September 80
April 78 October 82
May 87 November 83
June 78 December 73

Table 2. Greenhouse feature set

View table

Table 2. Greenhouse feature set

Feature subset	Feature	Description	Feature	Description
Spectral features	Aerosol	Aerosol band	Red edge 3	Red edge 3 band
	Blue	Blue band	NIR	Near-infrared
	Green	Green band	Red edge 4	Red edge 4 band
	Red	Red band	Water vapor	Water vapor band
	Red edge 1	Red edge 1 band	SWIR 1	Short-wave infrared 1
	Red edge 2	Red edge 2 band	SWIR 2	Short-wave infrared 2
Vegetation features	EVI	Enhanced vegetation index^［21］	REP	Red edge position^［22］
	SAVI	Soil-adjusted vegetation index^［23］	NDRE	Normalized difference red edge index^［24］
	NDVI	Normalized difference vegetation index^［25］	WDRVI	Wide dynamic range vegetation index^［26］
	GNDVI	Green normalized difference vegetation index^［27］	MTCI	Meris terrestrial chlorophyll index^［28］
Greenhouse features	PMLI	Plastic-mulched landcover index^［29］	PGI	Plastic greenhouse index^［30］
Greenhouse features	V_I	Greenhouse vegetable land index^［31］	APGI	Advanced plastic greenhouse index^［32］

Table 3. Dataset sample information
View table
Table 3. Dataset sample information
Dataset Number of bands Sample area /km² Number of labeled sample images Number of enhanced sample images
Monthly images 12 1258.29 576 576
Annual images 12 1258.29 48 576
Greenhouse feature set 24 1258.29 576 576

Table 4. Accuracy evaluation results of different datasets

View table

Table 4. Accuracy evaluation results of different datasets

Method	Annual images	Monthly images	Greenhouse feature set	Optimized feature dataset
SVM	94.82	91.68	91.72	91.78
RF	93.49	91.44	91.48	91.63
BPNN	93.89	89.14	90.18	91.49
FCN	95.52	96.96	94.55	96.54
U-Net	96.20	97.83	97.67	97.73
RLUNet	97.13	98.92	98.65	98.89

Table 5. Accuracy evaluation results of different semantic segmentation methods

View table

Table 5. Accuracy evaluation results of different semantic segmentation methods

Validation images	Method	Training images
		Annual images			Monthly images
		OA	IoU	F1	OA	IoU	F1
Annual images	FCN	95.52	72.05	83.75	96.41	52.78	68.11
	U-Net	96.20	82.70	90.53	96.91	50.85	66.46
	RLUNet	97.13	87.67	93.43	97.14	57.67	72.41
Monthly images	FCN	93.32	12.78	16.52	96.96	74.28	85.24
	U-Net	93.76	13.92	20.62	97.83	84.24	91.38
	RLUNet	93.82	14.97	22.73	98.92	88.98	94.17

Tools

Get Citation

Copy Citation Text

Guolong Li, Zhen Zhang, Jing Ding, Wanli Wang, Heling Sun, Chao Deng. Robust Semantic Segmentation Model for Greenhouses Based on Remote Sensing Feature Optimization[J]. Laser & Optoelectronics Progress, 2025, 62(12): 1228005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites