FPGA Accelerator for Meta-Recognition Anomaly Detection: Case of Burned Area Detection

Optical remote sensing instruments accumulate abundant data from across all of the earth's land surfaces, making it possible both to understand effects climate change and monitor, investigate, manage ground-level events in detail. Processing using resources located near on-board satellite sensors can bring major benefits terms minimizing analysis time quickly initiating active actions critical situations. In missions, long-term production algorithms may encounter unexplored samples, i.e., abnormal events, need be able discriminate take correct action. this matter, authors present a field programmable gate array (FPGA)-based solution for natural anomaly detection multispectral imagery deep convolutional neural networks. The weather-induced hazards disasters, considered anomalies sense, are discovered by modeling an detector on hybrid system that is hardware efficient. proposed approach assembled Xilinx Zynq UltraScale+ XCZU9EG multiprocessor system-on-chip (MPSoC) device, where model scaled into FPGA logic, followed downstream statistical meta-recognition predictor. accelerator has produced notable results identifying contemporary hazard, burned areas, scenes acquired Sentinel-2 over Europe, Spain France. implemented algorithm achieved equivalent speedup 4.46× 4.5× lower power consumption than implementation Tesla K80 GPU.