CFAR (Constant False Alarm Rate)

The CFAR is an adaptive detection that dynamically adjusts the detection threshold according to the surrounding noise.

It continuously estimates background noise from data arrays such as a spectrum, spectrogram, or scalogram

to maintain a stable false alarm rate and suppress unwanted detections caused by clutter or noise.

1. Detection Process

The detection process first measures the background noise within a localized reference window to establish a baseline.

Based on this real-time noise estimation, the processor dynamically scales a threshold across the entire signal matrix.

Signal regions exceeding the adaptive threshold are identified as target candidates and forwarded for subsequent processing.

Condition	CFAR Type( CA/GO/SO/OS-CFAR)	CA-CFAR
	Training Cell Size	y: 2 cells x: 2 cells
	Guard Cell Size	y: 1 cell x: 1 cell
	Alpha or P_fa Adaptive Threshold = α × Noise Estimation	α = 3.0
Input Array Map & Energy Level		Cell Under Test: CUT(5,5) & Its neighbors
Local Noise Map		(1)Calculate Local Noise – CUT(5,5) example – Total number of training cell: 40 – Local Noise = sum of training cells / 40 ≈ 10.2
Adaptive Threshold Map		(2)Calculate Adaptive Threshold –  CUT(5,5) example Adaptive Threshold = Local Noise × α = 10.25×3.0≈30.8
Detection		(3)Compare CUT energy vs. Threshold – CUT(5,5) example – Energy Level of CUT(5,5) = 10 – Local Noise of CUT(5,5) = 30.8 = 10<30.8→Target is absence at CUT(5,5)