Research on Vegetation Information Extraction from Visible UAV Remote Sensing Images

Abstract

Vegetation index has been widely used to detect the vegetation condition on the surface. UAV (Unmanned Aerial Vehicle) remote sensing system has advantages in flexibility, economy, high efficiency and real time monitoring. In this paper, 10 visible light vegetation indices based on UAV image were calculated and compared in their performance in extracting vegetation information. After the indices calculation, 15 same-sized AOI regions were selected for each of the six types of land objects (including vegetation-grassland, forest land, crops and non-vegetation-buildings, roads, and bare land) for further analysis of the performance of different vegetation indices. Bimodal histogram and the histogram entropy threshold method were used to determine the threshold value of each vegetation index for extracting vegetation information. Then, accuracy and efficiency were compared for different indices. Finally, RGBDI was chosen extract the vegetation from the UAV image owing to its extraction accuracy and efficiency. The results showed that RGBDI extraction accuracy was as highly as 99% and threshold value was easily determined by using bimodal histogram threshold method. The accuracy of VDVI and EXG were lower than RGBDI, VDVI has the accuracy of 97% and the EXG has 90%. Their threshold value can also be easily determined by histogram entropy method.