ISSN: 1513-6728

New Publication| Asian Journal of Geoinformatics

OilseedcropNet: Discriminating edible oilseed crops using PlanetScope temporal dataset

Saloni Awaghade, Anil Kumar*, Uttara Singh, Sunil Tiwari


Objective of this research work was to evaluate the significance of spectral bands at different stages of crop growth while mapping oilseed crops. Study area considered was in the surroundings of Merta, Nagaur, Rajasthan, India. The target oilseed crops for the study were Mustard and Taramira. To handle spectral overlap between these crops, temporal MASVI2 index from PlanetScope data were generated. Optimum number of temporal dates was selected from separability analysis using a temporal indices database. Optimum temporal indices data was classified with OilseedcropNet to map Oilseed crops. Mustard crop found in the field was homogeneous and was mapped as homogeneous fields. Taramira crop fields were not homogeneous, as it has soil patches in between crop patches within the same field. Due to this issue OilseedcropNet, has not mapped Taramira crop fields with acceptable accuracy. Further fuzzy Modified Possibilistic c-Means employing 'Individual Sample as Mean' Training approach was used to map Taramira crop, where each sample served as an input parameter for the model's training. It was observed that OilseedcropNet gave good classification results for homogeneous mustard crop fields with sufficient training data size, but failed to map heterogeneous Taramira crop with small training samples. On the other hand, the MPCM approach using the ‘Individual Sample as Mean’ training approach gave good classification results for the Taramira crop with fewer training samples. Red and NIR band combination for temporal indices database for mustard crop was found best. For Taramira mapping, Red and NIR bands for the initial stage, while yellow and red bands for later stages were found most effective.

Keywords: MSAVI2, MPCM, ‘Individual Sample as Mean’, MMD, OilseedcropNet

Full Text: