Mapping of Rice Plant Growth From Airborne Line Scanner Using ANFIS Method.

Abstract

A map of rice plant growth stage is important information to support food security. Remote sensing data on agricultural land have been taken using an airborne line scanner with three channels (NIR, Red, and Green). Airborne camera used is an engineering model of LISAT (LAPAN IPB satellite) camera. Adaptive Neuro fuzzy interferences system (ANFIS) has been applied to these data to mapping of rice plant growth stage. Twenty four classification scenarios were generated to obtain the best of accuracy. A classification scenario consists of input combination from the original image or image rationing and four types of fuzzy membership function. There are four rice plant growth stages: 1) new rice planting, 2) vegetative rice, 3) reproductive rice, and 4) ripening rice. Other objects classification are non vegetation / fallow and trees. Best accuracy occurred in scenario having input images rationing (MPRI, NDVSI, and SAVI) with trapezoid fuzzy membership function, which gives kappa value 95.76%. Best prediction class is new rice planting and ripening rice with user’s accuracy of more than 99% followed by vegetative rice with user’s accuracy of 93.48 %, while worst prediction class is reproductive rice with user’s accuracy of 74.38 %.