Report on Comparison of NDVI and Mahalanobis Distance in Paddy Fields and Field Survey for Anomaly Field Detection

Abstract

In the field of agriculture, crop analysis techniques using satellite imagery have been developed to conduct field survey effectively for grasping agricultural information, such as crop acreage, yield and grain quality. However, in case that the estimation model is constructed based on the field data, abnormal field data should be excluded to prevent adverse effects. The abnormal fields are extracted based on Mahalanobis’ distance and NDVI of fields calculated using SPOT satellite imagery. We conducted a field survey in Hokkaido to accumulate knowledge about the condition of the extracted fields on the ground. The extracted fields included undamaged fields including ‘growth delay’ and ‘no growth delay’ and damaged fields including ‘rice blast’, ‘sediment inflow’, ‘cool weather’, ‘flooded’, and ‘lodging’. Experimental results show that Mahalanobis’ distance is more suitable for detecting these field than NDVI.