Agriculture, food security and weather changes have been global concern. Water stress is an essential indicator in agriculture management. Owing to the unfavorable revisit time for most remote sensing based water stress detection, an automatic real-time detection framework is highly demanded. Currently, the biggest challenge that prevents the realization of such system is the complexity of real scenes in images. This paper aims to propose a deep learning framework that is built upon deep neural networks, big data, and modern computational power to detect water stress using hyperspectral images. The Coffe deep learning framework is adopted and implemented by NVidia's GPU in this paper. An embedded system will be built on the NVIDIA~R Jetson TX1 Developer Kit for real-time learning.
展开▼