日本人做的一个雷达系统仿真的matlab的toolbox.zip

**************************************** * Radar Toolbox - release version 0.11 * **************************************** ___IMPORTANT___ This software is released in the same manner as any academic work. A lot of routines are based on published theory but a great deal of effort was necessary to get numerical stability. You should give an acknowledgement if this toolbox is used in any publication or software: e.g. "Software by Glen Davidson at www.radarworks.com was used in this work". ___INSTALLATION___ Windows, Linux and Solaris should run without modification on Matlab 6.5 R13 On Windows, you need version 2000 or higher, the dlls probably won't run on anything below. OR, files like make* include the mex commands for building on any system. This is recommended anyway as only basic optimisation flags have been used. Check the Statlib/dcdflib/src/ directory as well. In total there should be 6 files to compile via: makecdf.m makeinvcdf.m makereclassrho.m makecovfullint.m makecovfullX.m makecfarmex.m But, an old single precision random library is also present (RT_rand) - it is not recommended but can be built using makeRT_rand.m For x86 Linux, mex files compiled with Intel ICC are provided, they are optimised for a Pentium 3 but will run on any system. If you have a Pentium 4, it is HIGHLY recommended that you recompile using ICC version 7 and change the optimisation flags in each make*.m file. The SSE2 instructions can almost double the speed of some routines. I would prefer Gnu GCC 3 for portability, but don't have it on my system. For Solaris, mex files compiled with Solaris CC are provided. They are not optimised, and better results are expected by recompiling with GCC. Anything else (OS X etc..), the makefiles are present and the code is GCC compatible. ___USE___ For pretty pictures and well commented code run the *_demo.m programs, currently these include: change_demo.m k_demo.m weibull_demo.m polarisation_demo.m separation_demo.m statistics_demo.m correlation_demo.m k_detection_demo.m segmentation_demo.m cfar_sim_demo.m cfar_theory_demo.m k_noise_demo.m ___DOCUMENTATION___ 1) See the index.html file in the directory HTML_help 2) The website at www.radarworks.com 3) For for each m file type > help filename 4) For mex files, run without any arguments to get a summary 5) Comments in source code. If www.radarworks.com is down, there is an an older mirror at: http://www.env.kochi-tech.ac.jp/ouchi/radarworks/index.htm But this may be severely out of date in the future. Report all bugs to gleninjapan@hotmail.com , or glen@radarworks.com if you get no response. Have fun! Glen Davidson Mar 17, 2003 ___KNOWN ISSUES___ 1. In the pdf and cdf statistics functions the limiting value of abscissa x->0 is not calculated. At present this usually returns NaN. 2. For low look number, the code to generate polarimetric samples can be slightly inaccurate, this seems to be a problem with floating point accuracy and that symmetric matrices are not explicitly recognised by Matlab. 3. For RT_Kcdf_noise and RT_Kpdf_noise.m the numerical integration to infinity is sensitive to extreme values for the mean clutter (mu) and noise component (rho). The routine rescales the input but if rho >> mu or mu >> rho by a factor of 1e3, the corresponding limit distribution is used. In general, any extreme values will cause problems in this routine due to the numerical integration routines. 4. In RT_cgaussian_fft.m , the two sided autocorrelation function has been artificially padded with a single 0 to make it an even length (allowing a fast power-2 FFT). The results may not be correct for short acf sequences that do not end at zero, in such cases a direct filter method such as RT_cgaussian_iir.m would be preferred anyway. 5. segfullX may ungracefully crash on Solaris with low memory - this may be because I didn't use the GNU compiler, or it may be a problem with the MEMCHECK code in segfullX.c . Anyway, it runs if you have lots of memory so I am in no mood to debug.... 6. Matlab 5 will require some simple modifications to the files, I can't test it anymore though. ___CHANGELOG___ v0.11: Mar 17, 2003 Major rewrite of openairsar.m to open old data files as well, I completely hacked the old version thinking that hdxxxx.log files were always present. Previous version is now renamed to openairsar_new.m but it only works for recent datasets. Included Solaris .mexsol files (may crash with low memory) Changed make*.m files so they run on all systems (not just Lin/Win) Removed accidental C++ comments from .c files Improved documentation of Weibull functions Added warning when RT_Kcdf_noise.m and RT_Kpdf_noise.m are using limiting distribution v0.10: Mar 4, 2003 Alterations to RT_Kpdf_noise.m and RT_Kcdf_noise.m for better stability (still slow) Added HTML help files (in directory HTML_help) v0.09: Mar 3, 2003 Added RT_Kpdf_noise.m and RT_Kcdf_noise.m Added k_noise_demo.m BUG: RT_cgaussian_fft.m incorrectly formed the two sided autocorrelation function, this meant that zero correlation could not be achieved. (See Known Issues above) Changed RT_Kcdf.m so that it returns [p,q,uerr] rather than [p,uerr] in line with other CDF functions. Corrected slight typo in cgamnn.m , now slightly more robust. v0.08: Feb 26, 2003 Improved the accuracy checks in separation_demo.m, changed to default 4 looks Hopefully improved the hack in single_polsep.m and multiple_polsep.m that allows 2x2 covariance matrices to be used with the same code. v0.07: Feb 25, 2003 Changed multiple_polsep.m so that it no longer calls gaminv.m (required Statistics Toolbox) Documented separation_demo.m (previously included as separation_test.m) Minor code and documentation changes. v0.06: Feb 20, 2003 Added Weibull statistical functions, based on shape parameter a and mean mu. Note: The standard Weibull distribution uses a scale parameter b which is more difficult to understand (it's not a linear function of the mean). Modified statistics_demo.m to include Weibull PDF and CDF Added weibull_demo.m to use parameter estimator weib_est.m v0.05: Feb 17, 2003 Added Constant False Alarm Rate (CFAR) functions for CA, CAGO, CASO, OS and TM This includes theoretical Threshold, Probability of False Alarm and Detection. Also the processor architectures have been implemented in C for speed. CFAR Theory demonstration added (cfar_theory_demo.m) CFAR Simulation demonstration added (cfar_sim_demo.m) v0.04: Feb 12, 2003 Added F distribution functions Added change detection demo using F distribution (change_demo.m) Replaced RT_cgaussian.m with RT_cgaussian_irr.m and RT_cgaussian_fft.m Updated correlation_demo.m to use new functions Added Swerling Probability of Detection Function (swerling_pd.m) Added random generator for K with coherent noise (RT_Krnd_noise.m) Added Target Detection demo for IID K (k_detection_demo.m) Improved comments in demos v0.03: Feb 6, 2003 Removed dlls for Windows ME, they were far too slow using the Borland compiler. Added spatially correlated Gamma generation functions Added correlation_demo.m BUG: Shape parameter in segfullX.dll for multiple frequencies was counted twice, i.e. an input parameter of s for F frequencies for 3 polarisations was used as s * F * 3 * F ... now correctly set as s * F * 3. Included RT_geomean - corrects bug in previous version. v0.02: Corrected function name in polarisation_demo.m Included digamma.m and modified calling files for when psi.m present Additional dlls included for Windows ME v0.01: Initial test release onto the net ___LEGAL____ NO WARRANTY (APPLIED TO ALL RELATED PROGRAMS AND THEORY) WE PROVIDE ABSOLUTELY NO WARRANTY OF ANY KIND EITHER