A new framework for UAV-based remote sensing data processing and its application in almond water stress quantification

With the rapid development of small imaging sensors and unmanned aerial vehicles (UAVs), remote sensing is undergoing a revolution with greatly increased spatial and and temporal resolutions. While more relevant detail becomes available, it is a challenge to analyze the large number of images to extract useful information. This research introduces a new general framework to process high-resolution multispectral images based on Principle Component Analysis (PCA) for crop stress quantification. As a case study, this framework is applied in almond water stress quantification using UAV-based remote sensing images. First, crop distributions of pixel value of sample trees are obtained as histograms consisted of 255 bins, assuming the stress information lies in the overall canopy pixels and ignoring the spatial relations among pixels. Second, PCA is applied to extract principle components out of histograms of 255 dimensions. This approach is advantageous in that it makes no assumption about the underlying canopy distribution of pixel values. It is shown that the first principle component has a significant correlation with stem water potential. This method is also compared with the traditional method of using the mean values of canopy Normalized Difference Vegetation Index (NDVI) as a baseline, and it shows improved performance in predicting the water stress.