Abstract: In order to improve extraction accuracy of satellite images for complex mountainous tobacco planting areas, Gaofen No.6 2 m resolution satellite images were used as remote sensing data sources．Based on multi temporal satellite remote sensing image sequences and satellite image features of different growth stages of tobacco, an object-oriented classification method was adopted to complete preliminary extraction of tobacco planting areas from three different temporal satellite images．multi temporal vegetation index rules were used for preliminary extraction of tobacco planting areas, and finally obtaining tobacco planting areas through cross validation, and completing accuracy verification．Results showed that research method could accurately extract complex mountainous tobacco planting areas, maintain a low misclassification rate（5.3%）, and achieve an extraction accuracy of 94.67%, significantly improving accuracy of extracting tobacco planting areas based on a single temporal satellite image（90%）．