Genetic algorithm-based decision tree classifier for remote sensing mapping with SPOT-5 data in the HongShiMao watershed of the loess plateau, China

摘要

The loess plateau in China has faced severe soil erosion and runoff. Check-dams are effective measures for soil and water conservation; concomitantly check-dam planning and construction urgently require current land use maps. Remote sensing technique plays a key role in achieving up-to-date land use maps. However, limited by the impact of hilly and gully terrain in the loess plateau, commonly used classifier for remote sensing data cannot achieve satisfactory results. In this paper, HongShiMao watershed in the loess plateau was chosen as the study area. Decision tree classifier (DTC) based on a genetic algorithm (GA) was applied to the land use classification automatically. Compared with the results by iterative self-organizing data analysis technique (ISODATA), GA-based DTC had much better results. Its total accuracy was 83.2% with a Kappa coefficient 0.807. The results also showed that most part of the study area belonged to the barren land with sparse grass or crop cover that attributed to the soil erosion and runoff.