Traditional remote sensing methods. AGs mapping has been widely concerned. Some pioneers extracted AGs from RS imagery by constructing novel spectral indices. Yang et al. (2017) proposed a new plastic greenhouse index PGI that can be used to identify transparent greenhouses in atmospherically corrected Landsat images; Guo and Li (2020) proposed a Normalized Difference Plastic Index NDPI using WorldView-3 SWIR bands; Zhang et al. (2022) constructed a novel spectral index APGI using Sentinel-2 images. However, these carefully constructed spectral indices may fail when the spectral response of AGs varies with seasons. Other works map AGs through an object-oriented approach.Wu et al. (2016) integrated RS data from Landsat-8 and an object-oriented classification method to implement an inheritance classification for AGs; Ji et al. (2019) proposed a threshold model created by 7 discriminative features to extract AGs from high-resolution GE imagery; Balcik et al. (2020) employed an object-based image classification method with 3 different classifiers to detect AGs in SPOT-7 and Sentinel-2 Multispectral Instrument (MSI) images. Although such methods are steady to the spectral features of AGs, they have limited feature extraction capabilities and are susceptible to isolated noise.

传统的遥感方法中，AGs的制图已经受到了广泛的关注。一些先驱利用构建新的光谱指数从RS图像中提取AGs。例如，Yang等人（2017）提出了一种新的塑料温室指数PGI，可用于在大气校正的Landsat图像中识别透明温室; Guo和Li（2020）利用WorldView-3 SWIR波段提出了一种归一化的塑料指数NDPI; Zhang等人（2022）利用Sentinel-2图像构建了一种新的光谱指数APGI。然而，这些精心构建的光谱指数在AGs的光谱响应随季节变化时可能会失效。其他的研究通过面向对象的方法来制图AGs。例如，吴等人（2016）将来自Landsat-8的RS数据和面向对象分类方法相结合，实现了对AGs的继承分类; Ji等人（2019）提出了一个由7个判别特征创建的阈值模型，用于从高分辨率GE图像中提取AGs; Balcik等人（2020）利用基于对象的图像分类方法和3种不同的分类器来检测SPOT-7和Sentinel-2多光谱仪（MSI）图像中的AGs。尽管这些方法对于AGs的光谱特征是稳健的，但它们的特征提取能力有限，容易受到孤立噪声的影响。