Automatic extraction of greenhouses from remote sensing images

Greenhouse detection is important with respect to urban and rural planning, yield estimation and crop planning, sustainable development, natural resource management, risk analysis and damage assessment. When generating Greenhouse Information Systemthe detection of greenhouses using traditional techniques may be very time consuming and expensive, especially in the residential areas that include intensive greenhouse areas. Therefore, fast and accurate detection of greenhouses and their types automaticaly from remote sensing imagery is important for saving labour and time. In this study, it is aimed to automatically detect greenhouse areas by using color and infrared orthophoto (RGBIR), topographic map and Digital Surface Model (DSM). There are two main steps in the study: (i) Determination of greenhouse areas using Object Based Image Analysis (OBIA); and (ii) Obtaining greenhouse boundaries and separating plastic and glass greenhouses. Color and infrared orthophotos, normalized Digital Surface Model (nDSM), Normalized Difference Vegetation Index (NDVI) and Visible Red-based Built-up Index (VrNIR_BI) were used in the determination of greenhouse areas using OBIA. In this process, the optimum scale parameter was determined automatically by the Estimation of Scale Parameter2 (ESP2) tool and Multi-Resolution Segmentation (MRS) was used as the segmentation algorithm. In the classification stage, K-Nearest Neighbor (K-NN), Random Forest (RF) and Support Vector Machine (SVM) classification techniques were used and the accuracies of the classification results were compared. The classification with the highest accuracy was reduced to two classes as greenhouse and non-greenhouse areas. In the second stage of the study, the boundaries of the greenhouse areas obtained from classification were determined, and the glass and plastic greenhouses were separated by using Python programming language from classification results and nDSM data. The study was implemented in Kumluca, Antalya, which includes intensive greenhouse areas and there are glass greenhouses as well as plastic greenhouses. Obtained results showed that greenhouse areas can be determined from color and infrared orthophoto and DSM data successfully by using object-based classification techniques. The highest overall accuracy was obtained when the SVM classifier was used with 94.80%, followed by the K-NN classification with 86.14%. The lowest overall accuracy was obtained as a result of RF classification with 81.46%. Furthermore, it is seen that greenhouse boundaries and plastic-glass greenhouse separations can be made with the proposed approach. However, it is thought that if more accurate DSM is used, the results can be improved, the algorithm can be applied in larger areas and more successful results can be obtained.