FPGA-Based CNN for Real-Time UAV Tracking and Detection

Neural networks (NNs) are now being extensively utilized in various artificial intelligence platforms specifically the area of image classification and real-time object tracking. We propose a novel design to address problem unmanned aerial vehicle (UAV) monitoring detection using Zynq UltraScale FPGA-based convolutional neural network (CNN). The biggest challenge while implementing algorithms on FPGAs is limited DSP hardware resources available FPGA platforms. Our proposed overcomes autonomous UAV tracking Xilinx’s XCZU9EG system chip (SoC) platform. explores provides solution for overcoming floating-point maintaining performance. consists two modules: module network–based module. uses our background-differencing algorithm, based modified CNN designed give maximum field-programmable gate array (FPGA) These modules complement each other enabled simultaneously provide an enhanced any given video input. has been tested real-life flying UAVs, achieving accuracy 82%, running at full frame rate input camera both (NN) detection, similar performance than equivalent deep learning processor unit (DPU) with HD implementation but lower resource utilization as shown by results.