IPP Software Navigation Tools IPP Links Communication Pan-STARRS Links

Ignore:
Timestamp:
Aug 31, 2013, 5:55:16 AM (13 years ago)
Author:
eugene
Message:

merge changes from EAM branch (pmModel_CentralPixel functions for sersic-like model support; add interactive mode to PCM LMM fitting; harder PCM LMM damping (nu scaled by 3 not 2); add EXT_AND SKY and SHAPE fitting modes, include sky in INDEX-only fits; ifdefs for trace-only values; allow positions to go to 100,000; major re-work of central pixel & flux for sersic-like models; add pmPCMMakeModel function to generate the flux for an arbitrary model

Location:
trunk/psModules
Files:
2 edited

Legend:

Unmodified
Added
Removed

  • trunk/psModules

  • trunk/psModules/src/objects/models/pmModel_DEV.c

    r35768 r36085  
    1616   * PM_PAR_SYY 5   - Y^2 term of elliptical contour (sqrt(2) / SigmaY)
    1717   * PM_PAR_SXY 6   - X*Y term of elliptical contour
    18    * PM_PAR_7   7   - normalized dev parameter
    1918
    2019   note that a standard dev model uses exp(-K*(z^(1/2n) - 1).  the additional elements (K,
    … …  
    4948#include "pmPSFtry.h"
    5049#include "pmDetections.h"
     50#include "pmModel_CentralPixel.h"
    5151
    5252#include "pmModel_DEV.h"
    … …  
    6363# define PM_MODEL_SET_LIMITS      pmModelSetLimits_DEV
    6464
    65 // f = exp(-z^0.125)
     65// f = exp(-kappa*r^(1/index))
     66// f = exp(-kappa*z^(0.5/index))
     67// index = 4, 0.5/index = 0.125
    6668# define ALPHA 0.125
    67 // # define ALPHA 0.25
    6869
    6970// the model is a function of the pixel coordinate (pixcoord[0,1] = x,y)
    … …  
    7374// Lax parameter limits
    7475static float paramsMinLax[] = { -1.0e3, 1.0e-2, -100, -100, 0.001, 0.001, -1.0 };
    75 static float paramsMaxLax[] = { 1.0e5, 1.0e8, 1.0e4, 1.0e4, 100, 100, 1.0 };
     76static float paramsMaxLax[] = { 1.0e5, 1.0e9, 1.0e5, 1.0e5, 100, 100, 1.0 };
    7677
    7778// Moderate parameter limits
    … …  
    8687static float *paramsMinUse = paramsMinLax;
    8788static float *paramsMaxUse = paramsMaxLax;
    88 static float betaUse[] = { 1000, 3e6, 5, 5, 1.0, 1.0, 0.5 };
     89static float betaUse[] = { 2, 3e6, 5, 5, 10.0, 10.0, 0.5 };
    8990
    9091static bool limitsApply = true;         // Apply limits?
    91 
    92 # include "pmModel_SERSIC.CP.h"
    9392
    9493psF32 PM_MODEL_FUNC (psVector *deriv,
    … …  
    109108    psAssert (z >= 0, "do not allow negative z values in model");
    110109
    111     float index = 0.5 / ALPHA;
    112     float par7 = ALPHA;
    113     float bn = 1.9992*index - 0.3271;
    114     float Io = exp(bn);
    115 
    116     psF32 f2 = bn*pow(z,ALPHA);
    117     psF32 f1 = Io*exp(-f2);
    118 
    119     psF32 radius = hypot(X, Y);
    120     if (radius < 1.0) {
    121 
    122         // ** use bilinear interpolation to the given location from the 4 surrounding pixels centered on the object center
    123 
    124         // first, use Rmajor and index to find the central pixel flux (fraction of total flux)
    125         psEllipseAxes axes;
    126         pmModelParamsToAxes (&axes, PAR[PM_PAR_SXX], PAR[PM_PAR_SXY], PAR[PM_PAR_SYY], true);
    127 
    128         // get the central pixel flux from the lookup table
    129         float xPix = (axes.major - centralPixelXo) / centralPixeldX;
    130         xPix = PS_MIN (PS_MAX(xPix, 0), centralPixelNX - 1);
    131         float yPix = (index - centralPixelYo) / centralPixeldY;
    132         yPix = PS_MIN (PS_MAX(yPix, 0), centralPixelNY - 1);
    133 
    134         // the integral of a Sersic has an analytical form as follows:
    135         float logGamma = lgamma(2.0*index);
    136         float bnFactor = pow(bn, 2.0*index);
    137         float norm = 2.0 * M_PI * PS_SQR(axes.major) * index * exp(bn) * exp(logGamma) / bnFactor;
    138 
    139         // XXX interpolate to get the value
    140         // XXX for the moment, just integerize
    141         // XXX I need to multiply by the integrated flux to get the flux in the central pixel
    142         float Vcenter = centralPixel[(int)yPix][(int)xPix] * norm;
    143        
    144         float px1 = 1.0 / PAR[PM_PAR_SXX];
    145         float py1 = 1.0 / PAR[PM_PAR_SYY];
    146         float z10 = PS_SQR(px1);
    147         float z01 = PS_SQR(py1);
    148 
    149         // which pixels do we need for this interpolation?
    150         // (I do not keep state information, so I don't know anything about other evaluations of nearby pixels...)
    151         if ((X >= 0) && (Y >= 0)) {
    152             float z11 = z10 + z01 + PAR[PM_PAR_SXY]; // X * Y positive
    153             float V00 = Vcenter;
    154             float V10 = Io*exp(-bn*pow(z10,par7));
    155             float V01 = Io*exp(-bn*pow(z01,par7));
    156             float V11 = Io*exp(-bn*pow(z11,par7));
    157             f1 = interpolatePixels(V00, V10, V01, V11, X, Y);
    158         }
    159         if ((X < 0) && (Y >= 0)) {
    160             float z11 = z10 + z01 - PAR[PM_PAR_SXY]; // X * Y negative
    161             float V00 = Io*exp(-bn*pow(z10,par7));
    162             float V10 = Vcenter;
    163             float V01 = Io*exp(-bn*pow(z11,par7));
    164             float V11 = Io*exp(-bn*pow(z01,par7));
    165             f1 = interpolatePixels(V00, V10, V01, V11, (1.0 + X), Y);
    166         }
    167         if ((X >= 0) && (Y < 0)) {
    168             float z11 = z10 + z01 - PAR[PM_PAR_SXY]; // X * Y negative
    169             float V00 = Io*exp(-bn*pow(z01,par7));
    170             float V10 = Io*exp(-bn*pow(z11,par7));
    171             float V01 = Vcenter;
    172             float V11 = Io*exp(-bn*pow(z10,par7));
    173             f1 = interpolatePixels(V00, V10, V01, V11, X, (1.0 + Y));
    174         }
    175         if ((X < 0) && (Y < 0)) {
    176             float z11 = z10 + z01 + PAR[PM_PAR_SXY]; // X * Y positive
    177             float V00 = Io*exp(-bn*pow(z11,par7));
    178             float V10 = Io*exp(-bn*pow(z10,par7));
    179             float V01 = Io*exp(-bn*pow(z01,par7));
    180             float V11 = Vcenter;
    181             f1 = interpolatePixels(V00, V10, V01, V11, (1.0 + X), (1.0 + Y));
    182         }
     110    // for DEV, we can hard-wire kappa(4):
     111    // float index = 4.0;
     112    float kappa = 7.670628;
     113
     114    // r = sqrt(z)
     115    float q = kappa*pow(z,ALPHA);
     116    float f0 = exp(-q);
     117
     118    assert (isfinite(q));
     119    assert (isfinite(f0));
     120
     121    // only worry about the central pixels at most
     122    float radius = hypot(X, Y);
     123    if (radius <= 1.5) {
     124        // Nsub ~ 10*index^2 + 1
     125        psEllipseAxes axes = pmPSF_ModelToAxes(PAR, pmModelClassGetType ("PS_MODEL_DEV"));
     126        int Nsub = 2 * ((int)(25 / axes.minor)) + 1;
     127        Nsub = PS_MIN (Nsub, 121);
     128        Nsub = PS_MAX (Nsub, 11);
     129        f0 = pmModelCP_SersicSubpix (X, Y, PAR[PM_PAR_SXX], PAR[PM_PAR_SXY], PAR[PM_PAR_SYY], 4.0, Nsub);
    183130    }   
    184131
    185     psF32 z0 = PAR[PM_PAR_I0]*f1;
    186     psF32 f0 = PAR[PM_PAR_SKY] + z0;
    187 
    188     assert (isfinite(f2));
     132    float f1 = PAR[PM_PAR_I0]*f0;
     133    float f = PAR[PM_PAR_SKY] + f1;
     134
    189135    assert (isfinite(f1));
    190     assert (isfinite(z0));
    191     assert (isfinite(f0));
     136    assert (isfinite(f));
    192137
    193138    if (deriv != NULL) {
    … …  
    195140
    196141        dPAR[PM_PAR_SKY]  = +1.0;
    197         dPAR[PM_PAR_I0]   = +2.0*f1; // XXX extra damping..
    198 
    199         // gradient is infinite for z = 0; saturate at z = 0.01
    200         psF32 z1 = (z < 0.01) ? z0*bn*ALPHA*pow(0.01,ALPHA - 1.0) : z0*bn*ALPHA*pow(z,ALPHA - 1.0);
    201 
    202         assert (isfinite(z1));
    203 
    204         dPAR[PM_PAR_XPOS] = +1.0*z1*(2.0*px/PAR[PM_PAR_SXX] + Y*PAR[PM_PAR_SXY]);
    205         dPAR[PM_PAR_YPOS] = +1.0*z1*(2.0*py/PAR[PM_PAR_SYY] + X*PAR[PM_PAR_SXY]);
    206         dPAR[PM_PAR_SXX]  = +2.0*z1*px*px/PAR[PM_PAR_SXX];
    207         dPAR[PM_PAR_SYY]  = +2.0*z1*py*py/PAR[PM_PAR_SYY];
    208         dPAR[PM_PAR_SXY]  = -1.0*z1*X*Y;
    209     }
    210     return (f0);
     142        dPAR[PM_PAR_I0]   = +f0;
     143
     144        if (z > 0.01) {
     145          float z1 = f1*kappa*ALPHA*pow(z,ALPHA-1.0);
     146          dPAR[PM_PAR_XPOS] = +1.0*z1*(2.0*px + Y*PAR[PM_PAR_SXY]);
     147          dPAR[PM_PAR_YPOS] = +1.0*z1*(2.0*py + X*PAR[PM_PAR_SXY]);
     148          dPAR[PM_PAR_SXX]  = +2.0*z1*px*px/PAR[PM_PAR_SXX];
     149          dPAR[PM_PAR_SYY]  = +2.0*z1*py*py/PAR[PM_PAR_SYY];
     150          dPAR[PM_PAR_SXY]  = -1.0*z1*X*Y;
     151        } else {
     152          // gradient -> 0 for z -> 0, but has undef form
     153          float z1 = f1*kappa*ALPHA*pow(z,ALPHA);
     154          dPAR[PM_PAR_XPOS] = +1.0*z1*(2.0/PAR[PM_PAR_SXX] + PAR[PM_PAR_SXY]);
     155          dPAR[PM_PAR_YPOS] = +1.0*z1*(2.0/PAR[PM_PAR_SYY] + PAR[PM_PAR_SXY]);
     156          dPAR[PM_PAR_SXX]  = +2.0*z1*px/PAR[PM_PAR_SXX]/PAR[PM_PAR_SXX];
     157          dPAR[PM_PAR_SYY]  = +2.0*z1*py/PAR[PM_PAR_SYY]/PAR[PM_PAR_SYY];
     158          dPAR[PM_PAR_SXY]  = -1.0*z1;
     159        }
     160    }
     161    return (f);
    211162}
    212163
    … …  
    302253    }
    303254
    304     // the normalization is modified by the slope
    305     float index = 0.5 / ALPHA;
    306     float bn = 1.9992*index - 0.3271;
    307     float Io = exp(0.5*bn);
    308 
    309255    // set the model normalization
    310256    if (!pmModelSetNorm(&PAR[PM_PAR_I0], source)) {
    311257      return false;
    312258    }
    313     PAR[PM_PAR_I0] /= Io;
    314259
    315260    // set the model position
    … …  
    328273    psEllipseAxes axes;
    329274    pmModelParamsToAxes (&axes, PAR[PM_PAR_SXX], PAR[PM_PAR_SXY], PAR[PM_PAR_SYY], true);
    330     float AspectRatio = axes.minor / axes.major;
    331 
    332     float index = 4.0;
    333     float bn = 1.9992*index - 0.3271;
    334 
    335     // the integral of a Sersic has an analytical form as follows:
    336     float logGamma = lgamma(2.0*index);
    337     float bnFactor = pow(bn, 2.0*index);
    338     float norm = 2.0 * M_PI * PS_SQR(axes.major) * index * exp(bn) * exp(logGamma) / bnFactor;
    339    
    340     psF64 Flux = PAR[PM_PAR_I0] * norm * AspectRatio;
    341 
    342     return(Flux);
     275
     276    float norm = 0.00168012;
     277    float flux = PAR[PM_PAR_I0] * 2.0 * M_PI * axes.major * axes.minor * norm;
     278
     279    return(flux);
    343280}
    344281
    … …  
    359296    pmModelParamsToAxes (&axes, PAR[PM_PAR_SXX], PAR[PM_PAR_SXY], PAR[PM_PAR_SYY], true);
    360297
    361     // f = Io exp(-z^n) -> z^n = ln(Io/f)
    362     psF64 zn = log(PAR[PM_PAR_I0] / flux);
    363     psF64 radius = axes.major * sqrt (2.0) * pow(zn, 0.5 / ALPHA);
     298    // static value for DEV:
     299    float kappa = 7.670628;
     300
     301    // f = Io exp(-kappa*z^n) -> z^n = ln(Io/f) / kappa
     302    psF64 zn = log(PAR[PM_PAR_I0] / flux) / kappa;
     303    psF64 radius = axes.major * pow(zn, 0.5 / ALPHA);
    364304
    365305    psAssert (isfinite(radius), "fix this code: radius should not be nan for Io = %f, flux = %f, major = %f (%f, %f, %f)",
Note: See TracChangeset for help on using the changeset viewer.