Relationships of Class Number Variation and Image Classification Accuracy in the LAPAN-A3 Multispectral Imager

LAPAN-A3 has a multispectral imager payload that can be used for earth observation. One of its uses is for land use and land cover classification. To find out the suitability of the classification results with the actual data, it is necessary to calculate accuracy. This study aims to find out the relationship between variations in the number of classes and the accuracy of the classification results of LAPAN-A3 compare with Landsat-8. The research was conducted in 4 study areas in Indonesia, namely Mandailing Natal Regency (North Sumatra), Pandeglang Regency (Banten), Semarang City (Central Java), and Kupang Regency (NTT). It can be concluded that the accuracy is very good in the classification of 2 classes where the accuracy value is more than 95%. Good accuracy in the classification of 4 classes with an accuracy value of more than 85%. The accuracy is good enough in the 6 class classification with an accuracy value of 80%. The blur effect is the reason of the decrease in accuracy due to the less optimal ability to separate spectrals.