Vegetation Extraction Based on the Visual Characteristics of Plants from Remote Sensing Image

In urban planning, the measure and mapping of the green vegetation play an important role. On high resolution remote sensing images, the vegetation can be extracted accurately without any priori knowledge by using the tone characteristics of the image itself. Firstly, this paper introduces the principle of vegetation extraction method based on visual characteristics of plants. Then we propose an improved method to increase the accuracy of the final results. Morphological close operation is conducted before detecting the vegetation region contours so as to get qualifying results. The NDVI is also applied to decrease misclassifications of non-vegetation and improve the poor accuracy. The experiments showed that our improved method can reduce misclassifications on non-vegetation significantly and obtain higher accuracy.