Application of UAV imaging platform for vegetation analysis based on spectral-spatial methods

摘要

This paper presents application of UAV imaging platforms for vegetation analysis. Remote sensing using a UAV, also known as low altitude remote sensing is performed to acquire RGB images for vegetation analysis. Two UAV platforms, a VTOL quadcopter and a fixed wing UAV, were used to obtain the images. Crop region classification was carried out on images acquired from VTOL quadcopter to demonstrate its use in applications such as inspections that require hovering of UAVs while tree crown classification was carried out on images acquired from the fixed wing UAV to demonstrate its use in applications that requires coverage over a relatively larger area. Classification was performed for crop region mapping and tree crown mapping using spectral-spatial method. In this proposed method, Bayesian information criterion was used to determine the constraint of optimal number of clusters for a given image. Keeping this constraint, divisive approach was performed using k-means and EM algorithm for clustering the dataset. On these clusters, the agglomerative approach was used to merge the dataset. The merging was done using percentage voting. Further, to improve the classification efficiency, spatial classification was applied. UAV images obtained using the two UAV platforms were used to demonstrate the performance of the proposed algorithm. A performance comparison of the proposed spectral-spatial classification with the other classification methods is presented. From the obtained results, it was concluded that the proposed spectral-spatial classification performs better and was more robust than the other algorithms in the literature. (C) 2017 Elsevier B.V. All rights reserved.