Table 4 Performance comparison, single target in clutter edge (S) and multiple targets in clutter (M), \(P=6\)
From: A neural network framework for binary classification of radar detections
Detector (trained on) | S: \(P_{\mathrm{D}}\) | S: \(P_{\mathrm{FA}}\) | M: \(P_{\mathrm{D}}\) | M: \(P_{\mathrm{FA}}\) |
|---|---|---|---|---|
NN (N) O1, 50x1 | 0.61 | \(2.92\times 10^{-4}\) | 0.39 | \(4.00\times 10^{-4}\) |
NN (N) O2, 50x1 | 0.78 | \(1.89\times 10^{-4}\) | 0.65 | \(1.95\times 10^{-4}\) |
NN (N) O3, 50x1 | 0.81 | \(3.77\times 10^{-3}\) | 0.69 | \(3.76\times 10^{-3}\) |
NN (N) O1, 50x2 | 0.72 | \(1.17\times 10^{-3}\) | 0.62 | \(1.47\times 10^{-3}\) |
NN (N) O2, 50x2 | 0.80 | \(1.62\times 10^{-3}\) | 0.70 | \(1.76\times 10^{-3}\) |
NN (N) O3, 50x2 | 0.82 | \(9.04\times 10^{-3}\) | 0.73 | \(9.20\times 10^{-3}\) |
NN (N+C) O1, 50x1 | 0.61 | \(2.51\times 10^{-5}\) | 0.41 | \(2.38\times 10^{-5}\) |
NN (N+C) O2, 50x1 | 0.76 | \(8.59\times 10^{-6}\) | 0.62 | \(8.40\times 10^{-6}\) |
NN (N+C) O3, 50x1 | 0.77 | \(4.04\times 10^{-7}\) | 0.66 | \(4.86\times 10^{-7}\) |
NN (N+C) O1, 50x2 | 0.66 | \(3.45\times 10^{-5}\) | 0.58 | \(3.31\times 10^{-5}\) |
NN (N+C) O2, 50x2 | 0.79 | \(4.80\times 10^{-5}\) | 0.68 | \(4.73\times 10^{-5}\) |
NN (N+C) O3, 50x2 | 0.81 | \(1.98\times 10^{-5}\) | 0.71 | \(1.99\times 10^{-5}\) |
CMSO-CFAR | 0.82 | \(1.89\times 10^{-2}\) | 0.74 | \(1.90\times 10^{-2}\) |
SO-CFAR | 0.77 | \(8.00\times 10^{-3}\) | 0.66 | \(3.85\times 10^{-3}\) |
GO-CFAR | 0.60 | \(7.63\times 10^{-5}\) | 0.14 | \(7.03\times 10^{-5}\) |