CFAR.rar_CFAR algorithm_CFAR检测_CFAR检测 matlab_matlab cfar 3-D

% cfar.m % % simple cell-averaging lead/lag CFAR % % Mark A. Richards % April 2000 % Updated Nov. 2006 clear all, close all load cfar_even.mat; % Compute threshold for Pfa = 10^(-2) and mean power = 1. % Because the mean power = 1, this is also the threshold multiplier % for any mean power level. Pfa = 10^(-2); Tmult = -log(Pfa) % Set up the ideal threshold T_ideal = Tmult*ones(size(z)); % The CFAR has 50 each lead and lag cells, and 3 guard % cells on each side of the cell under test. Nref = 50; Nguard = 3; Nz=length(z); % Compute cell-averaging CFAR multiplier Nc = 2*Nref; alpha = Nc*((Pfa^(-1/Nc)) - 1) % Do a brute force sliding window CFAR over the range of % indices where the reference windows fully overlap the data. % This is inefficient but easy to understand. % Use NaNs ("not-a-numbers") for the undefined values in order % to create a full length output sequence for plotting convenience. first = Nref + Nguard + 1; last = Nz - Nref - Nguard; avg = NaN*ones(size(z)); for k=first:last avg(k) = mean([z(k-Nguard-Nref:k-Nguard-1);z(k+Nguard+1:k+Nguard+Nref)]); end T_cfar = alpha*avg; % plot the results. First, data and thresholds figure plot([z T_ideal T_cfar]); xlabel('sample'); ylabel('power');grid % now the threshold histogram figure hist(T_cfar,50); axis([0 20 0 1200]); xlabel('threshold value'); ylabel('number of occurrences'); title('Histogram of CFAR Threshold Values') % add vertical marker at the value of the ideal threshold hold on; v = axis; plot([Tmult,Tmult],[v(3),v(4)],'r') hold off; % Observed Pfa Ncross = sum(z>T_cfar) Pfa_obs = Ncross/(Nz-2*Nref-2*Nguard) % Now repeat the whole business for the CFAR_uneven.mat data file load cfar_uneven.mat; % Set up the ideal thresholds T_ideal1 = [Tmult*ones(Nz/2,1); NaN*ones(Nz-Nz/2,1)]; T_ideal10 = [NaN*ones(Nz/2,1); 10*Tmult*ones(Nz-Nz/2,1)]; % Do the CFAR. avg = NaN*ones(size(z)); for k=first:last avg(k) = mean([z(k-Nguard-Nref:k-Nguard-1);z(k+Nguard+1:k+Nguard+Nref)]); end T_cfar = alpha*avg; % plot the results. First, data and thresholds figure plot([z T_ideal1 T_ideal10 T_cfar]); xlabel('sample'); ylabel('power');grid % re-plot, but just the transition region figure plot([z(2300:2700) T_ideal1(2300:2700) T_ideal10(2300:2700) T_cfar(2300:2700)]); xlabel('sample'); ylabel('power');grid; axis([0 400 0 60]) % now the threshold histogram figure hist(T_cfar,100); axis([0 100 0 1200]); xlabel('threshold value'); ylabel('number of occurrences'); title('Histogram of CFAR Threshold Values') % add vertical marker at the values of the ideal thresholds hold on; v=axis; plot([Tmult,Tmult],[v(3),v(4)],'r') plot(10*[Tmult,Tmult],[v(3),v(4)],'r') hold off; % Observed Pfa Ncross = sum(z>T_cfar) Pfa_obs = Ncross/(Nz-2*Nref-2*Nguard) % Now repeat for the target case load cfar_target.mat Nz = length(z); % Standard CA CFAR first = Nref + Nguard + 1; last = Nz - Nref - Nguard; avg_ca = NaN*ones(size(z)); for k=first:last avg_ca(k) = mean([z(k-Nguard-Nref:k-Nguard-1);z(k+Nguard+1:k+Nguard+Nref)]); end % Repeat for SOCA CFAR avg_soca = NaN*ones(size(z)); for k=first:last avg_soca(k) = min((sum(z(k-Nguard-Nref:k-Nguard-1))), ... sum(z(k+Nguard+1:k+Nguard+Nref)))/Nref; end % Repeat for GOCA CFAR avg_goca = NaN*ones(size(z)); for k=first:last avg_goca(k) = max((sum(z(k-Nguard-Nref:k-Nguard-1))), ... sum(z(k+Nguard+1:k+Nguard+Nref)))/Nref; end alpha_soca = 5.175; alpha_goca = 6.686; T_ca = alpha*avg_ca; T_soca = alpha_soca*avg_soca; T_goca = alpha_goca*avg_goca; T_ideal = Tmult*ones(size(z)); % plot the results. First, data and thresholds figure plot(db([z T_ideal T_ca T_soca T_goca],'power')); axis([450 650 0 25]); xlabel('sample'); ylabel('power');grid