Analysis of Hardware-Computing Modules and Neural Networks for Aviation Surveillance Systems with Computer Vision

The paper proposes a hardware and software system for the tasks of recognizing physical ground objects in the flow of radar frames on the basis of a single-board computing module located on a small aircraft. Hardware computing modules with neural network processing blocks are analyzed. Small-size neural network models for their deployment on the hardware-computing module are analyzed and selected. The hardware-software system of airborne basing with realization of recognition of physical ground objects in the stream of radar frames is described.