Real-Time FMCW Radar CA-CFAR Pipeline

Engineered CA-CFAR and GO-CFAR target detectors in C++17 using FFTW3 single-precision FFT, with an adaptive sliding-window noise estimator and per-cell threshold calibrated for a constant false alarm rate. The detection core runs in well under a microsecond per 1024-bin range profile.

Built a three-tier streaming architecture: the C++17 binary emits newline-delimited JSON frames over stdout at 25 fps, a Python FastAPI bridge spawns it as a subprocess and rebroadcasts every frame to all WebSocket clients with bounded per-client queues that drop stale frames instead of leaking memory.

Shipped a React dashboard with per-frame animated SVG range-profile morphing, a 64×64 Canvas Doppler heatmap, a live detections table, and interactive CA/GO-CFAR parameter controls. Four pre-built radar scenes plus a live C++ stream mode, with the whole pipeline modelled in UML component diagrams and verified by a full Google Test suite covering false alarm rate and JSON serialisation.