Greenhouse Crop Identification from Multi-Temporal Multi-Sensor Satellite Imagery Using Object-Based Approach: A Case Study from Almería (Spain)

Abstract

M.A.A.); faguilar@ual.es (F.J.A.)Abstract: A workflow headed up to identify crops growing under plastic-covered greenhouses (PCG) and based on multi-temporal and multi-sensor satellite data is developed in this article. This workflow is made up of four steps: (i) data pre-processing, (ii) PCG segmentation, (iii) binary pre-classification between greenhouses and non-greenhouses, and (iv) classification of horticultural crops under greenhouses regarding two agronomic seasons (autumn and spring). The segmentation stage was carried out by applying a multi-resolution segmentation algorithm on the pre-processed WorldView-2 data. The free access AssesSeg command line tool was used to determine the more suitable multi-resolution algorithm parameters. Two decision tree models mainly based on the Plastic Greenhouse Index were developed to perform greenhouse/non-greenhouse binary classification from Landsat 8 and Sentinel-2A time series, attaining overall accuracies of 92.65% and 93.97%, respectively. With regards to the classification of crops under PCG, pepper in autumn, and melon and watermelon in spring provided the best results (F β around 84% and 95%, respectively). Data from the Sentinel-2A time series showed slightly better accuracies than those from Landsat 8.For instance, a combination of data from Sentinel-2 (2A and 2B) and Landsat 8 provides a global median average revisit interval of 2.9 days [9].Regarding PCG mapping from remote sensing, an increasing amount of scientific literature has been published during the last decade that has mainly focused on Landsat imagery [4,[10][11][12][13][14][15][16]. Novelli et al. [17] compared single-date Sentinel-2 and Landsat 8 data to automatically classify PCG. A few indices especially adapted to plastic sheet detection, such as the Index Greenhouse Vegetable Land Extraction (Vi) [18], Plastic-Mulched Landcover Index (PMLI) [10], Moment Distance Index (MDI) [12,19], Plastic Greenhouse Index (PGI) [13], and Greenhouse Detection Index (GDI) [16] have been recently proposed.In relation to the classification of crops via remote sensing, Badhwar [20] published one of the first works where Landsat imagery multi-temporal data (only three dates) were used for corn and soybean crops mapping. In fact, crop types classification from medium resolution satellite imagery was mainly conducted by using pixel-based approaches until approximately 2011. Just before Petitjean et al. [21] argued that the increasing spatial resolution of available spaceborne sensors was enabling the application of the object-based image analysis (OBIA) paradigm to extract crop types from satellite image time series, Peña-Barragán et al. [22] developed a methodology for outdoor crop identification and mapping using OBIA and decision tree algorithms. This methodology was also applied to a Landsat time series to map sugarcane over large areas [23]. This OBIA approach consisted of two main consecutive phases: (i) the delimitation of crop fields by image segmentation and, (ii) the application of decision rules based...

Evaluating the conservation state of the páramo ecosystem: An object-based image analysis and CART algorithm approach for central Ecuador

“…Our CDT serves as an exploratory tool for data-driven discovery and prediction to gain new insights on LULC in central EPEs. Although other classifiers could achieve better classification results, CDT provides clear decision rules with fixed threshold values that can be used in future research without any training phase [44]. It has been shown that the CART algorithm can provide stable performance and reliable results in machine learning and data mining research [45].…”

“…It has been used to identify spectral bands with the highest discriminative capabilities between classes and low misclassification rates [45]. Classification utilizing the DT algorithm has already been applied in identifying both field-grown and greenhouse crops from remote sensing data with excellent and robust results [44]. However, the CDT found in this study can become unstable, as a change in one node affects all of the nodes that are connected below.…”

Mapping Plastic Greenhouses Using Spectral Metrics Derived From GaoFen-2 Satellite Data

“…The exhibited plastic greenhouses belong to both arch and single plastic greenhouses, which are the main types in China for planting various vegetables and typically feature a size of 400-1200 m² [40]. In other areas, such as in Spain even larger cover areas are characteristic (around 10 000 m 2 ) [37]. Tomato, pimento, tabasco, cucumber, muskmelon, and watermelon are the most representative crops under the plastic greenhouses in this area.…”

“…Methods of spectral unmixing in combination with textural features are shown to be a prerequisite for high mapping accuracies of plastic greenhouses on medium resolution images of 91.2% [21]. However, for assessing the performances of different classification products based on medium resolution images [37], VHR images are always needed to verify and correct the mapping accuracies [38], [39]. Although medium resolution images seem more economical, reported accuracies are significantly lower.…”

Improving Ecotope Segmentation by Combining Topographic and Spectral Data

“…Alternatively, Geographic Object-Based Image Analysis can be used to delineate spatial regions by grouping adjacent pixels into homogeneous areas according to the objectives of the study [23,24]. For biodiversity research, image segmentation has been used to automatically derive homogeneous vegetation units based on spectral [25] or a combination of spectral and structural (height) information [16,26].…”