B Qa* @spddlZddlZddlZddlZddlZddlZddlmZddl m Z m Z ddl m ZddlmZddlmZmZmZmZmZdZdd d d d d dgZdZedZdZddZdBddZddZddZ dCddZ!dDd d!Z"d"d#Z#dEd$d%Z$dFd'd(Z%dGd*d+Z&dHd-d.Z'dId0d1Z(dJd2d3Z)dKd5d6Z*d7d8Z+dLd:d;Z,dMdd?Z.e/d@krle.ej0dAddS)ON)linalg) ifgramStackcluster) decorrelation)get_template_content)readfile writefileptimeutils arg_groupzmintpy.networkInversion.obsDatasetName numIfgram weightFunc maskDataset maskThreshold minRedundancyminNormVelocityaexample: ifgram_inversion.py inputs/ifgramStack.h5 -t smallbaselineApp.cfg --update ifgram_inversion.py inputs/ifgramStack.h5 -w no # turn off weight for fast processing ifgram_inversion.py inputs/ifgramStack.h5 -c no # turn off parallel processing # offset ifgram_inversion.py inputs/ifgramStack.h5 -i rangeOffset -w no -m waterMask.h5 --md offsetSNR --mt 5 ifgram_inversion.py inputs/ifgramStack.h5 -i azimuthOffset -w no -m waterMask.h5 --md offsetSNR --mt 5 invert_networkureferences: Berardino, P., Fornaro, G., Lanari, R., & Sansosti, E. (2002). A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. IEEE TGRS, 40(11), 2375-2383. doi:10.1109/TGRS.2002.803792 Pepe, A., and Lanari, R. (2006), On the extension of the minimum cost flow algorithm for phase unwrapping of multitemporal differential SAR interferograms, IEEE-TGRS, 44(9), 2374-2383. Perissin, D., and Wang, T. (2012), Repeat-pass SAR interferometry with partially coherent targets, IEEE TGRS, 50(1), 271-280, doi:10.1109/tgrs.2011.2160644. Samiei-Esfahany, S., Martins, J. E., Van Leijen, F., and Hanssen, R. F. (2016), Phase Estimation for Distributed Scatterers in InSAR Stacks Using Integer Least Squares Estimation, IEEE TGRS, 54(10), 5671-5687. Seymour, M. S., and Cumming, I. G. (1994), Maximum likelihood estimation for SAR interferometry, 1994. IGARSS '94., 8-12 Aug 1994. Yunjun, Z., Fattahi, H., and Amelung, F. (2019), Small baseline InSAR time series analysis: Unwrapping error correction and noise reduction, Computers & Geosciences, 133, 104331, doi:10.1016/j.cageo.2019.104331. Yunjun, Z., Fattahi, H., Brancato, V., Rosen, P., Simons, M. (2021), Oral: Tectonic displacement mapping from SAR offset time series: noise reduction and uncertainty quantification, ID 590, FRINGE 2021, 31 May – 4 Jun, 2021, Virtual. c Cstjdtjtdtdtd}|jddd|jddd d d |jd d ddtdd|jddddd |jddddddd|jdddd |jd d!d"d#d$d%|d&d'}|jd(d)d*d+d+d,d-d.hd/d0|jd1d2d3d4d%|jd5d6d7d8d7gd9d0|jd:d;dd?}|jd@dAdBdCdDd |jdEdFdGdHdIt dJdKdL|jdMdNdOdPdIt dQdRdLt |}t |}|jdSdTd#dUd%|S)VNz2Invert network of interferograms into time-series. ) descriptionformatter_classepilogifgramStackFilez(interferograms stack file to be inverted)helpz-tz --template templateFileztemplate text file with options)destrz-iz-dz--dsetr zkdataset name of unwrap phase / offset to be used for inversion e.g.: unwrapPhase, unwrapPhase_bridging, ...)rtyperz-mz --water-mask waterMaskFilez4Skip inversion on the masked out region, i.e. water.z-oz--outputoutfile)ZTS_FILEZ TCOH_FILEZ NUM_INV_FILEz)Output file name. (default: %(default)s).)rnargsmetavarrz --ref-dateref_datez&Reference date, first date by default.z--skip-referencez --skip-refskip_ref store_truez:[for offset and testing] do not apply spatial referencing.)ractionrsolverz(solver for the network inversion problemz-wz --weight-funcrvarZfimcohnozfunction used to convert coherence to weight for inversion: var - inverse of phase variance due to temporal decorrelation (default) fim - Fisher Information Matrix as weightcoh - spatial coherence no - no/uniform weight)rdefaultchoicesrz--min-norm-phaser store_falsezoEnable inversion with minimum-norm deformation phase, instead of the default minimum-norm deformation velocity.z--norm residualNormL2L1z;Optimization mehtod, L1 or L2 norm. (default: %(default)s).z--stdz --calc-stdcalcStdzdCalculate time-series STD via linear propagation from the network of interferograms or offset pairs.maskz&mask observation data before inversionz --mask-dsetz--mask-datasetz--mdrzBdataset used to mask unwrapPhase, e.g. coherence, connectComponentz --mask-thresz--mask-thresholdz--mtrNUMg?zIthreshold to generate mask when mask is coherence (default: %(default)s).)rr!rr*rz --min-redunz--min-redundancyz--mrrg?zWminimum redundancy of interferograms for every SAR acquisition. (default: %(default)s).z--update update_modezEnable update mode, and skip inversion if output timeseries file already exists, readable and newer than input interferograms file) argparseArgumentParserRawTextHelpFormatter REFERENCETEMPLATEEXAMPLE add_argumentstradd_argument_groupfloatr add_memory_argumentadd_parallel_argument)parserr&r1rAM/home/centos/operations/rsmas_insar/sources/MintPy/mintpy/ifgram_inversion.py create_parserCsJ             rCc Cs0t}|j|d}t|j}|ddkrsz&read_template2inps..)rrr)rr)r-rzmintpy.compute.cs g|]}|kr|qSrA)r`)rrrs)dask_key_prefixrlrArBrts)rconfig)r[) maxMemory)rr)) rpvarsr read_templateutcheck_template_auto_valuelistr`rqr=r;) template_filerjiDictkeyListrmvaluerA)rurlrBrWs<           rWc stdtdd}tddjDs>d}tdjntjdd }|d jd }WdQRXtj jd}||krd}td jdntd jtj d (}t |j j dtjj }WdQRXtddjD}||krd}tdj ntdj |dkrtj tjttj jdd_fdddD}tfddtDrd}tdtnB|rtfdd|Drd}td|ntdttd||S)Nz2--------------------------------------------------zupdate mode: ONskipcss|]}tj|VqdS)N)r]r^r_)rrrsrArArB szrun_or_skip..runz"1) NOT ALL output files found: {}.rr timeseriesz'1) output file {} is NOT fully written.z"1) output files already exist: {}.MODIFICATION_TIMEcss|]}tj|VqdS)N)r]r^getmtime)rrrsrArArBr sz52) output files are NOT newer than input dataset: {}.z22) output dataset is newer than input dataset: {}.T) dropIfgramcsg|]}|kr|qSrA)r`)rrrs)atr_tsrArBrtszrun_or_skip..)rOrNc3s.|]&}tt|t|dkVqdS)NoneN)r;rxgetrq)rrrm)rrjrArBrsz83) NOT all key configuration parameters are the same: {}c3s|]}||kVqdS)NrA)rrrm)atr_ifgrrArBrsz(3) NOT all the metadata are the same: {}z53) all key configuration parameters are the same: {}.zrun or skip: {}.)r\allrrVh5pyFilesizer]r^getsizerr=r attrsrrminrrTrflenrget_date12_listr any configKeys)rjflagf fsize_reffsizetito meta_keysrA)rrrjrB run_or_skipsD,     rTh㈵>?temporalCoherencec  CsF||jdd}|dk r,||jdd}|jdd} |jd} tj| | ftjd} |dkrhtj} nd} d}t||||gd\}\}}}ttj|dkdd |kr| | |fS|dk ryt t |j |Wnt j k r| | |fSXy&|r|dk r2t jt||t|||d dd \}}nt j|||d dd \}}t||||||| d } |t|d| f}tj|dd | ddddf<n~|dk rt jt||t|||d dd \}}nt j|||d dd \}}t||||||| d } || ddddf<Wnt j k r0YnX|jd}| | |fS) a Estimate time-series from a stack/network of interferograms with Least Square minimization on deformation phase / velocity. Problem: A X = y opt 1: X = np.dot(np.dot(numpy.linalg.inv(np.dot(A.T, A)), A.T), y) opt 2: X = np.dot(numpy.linalg.pinv(A), y) opt 3: X = np.dot(scipy.linalg.pinv(A), y) opt 4: X = scipy.linalg.lstsq(A, y)[0] [recommend and used] opt 4 supports weight. scipy.linalg provides more advanced and slighted faster performance than numpy.linalg. This function relies on the LAPACK routine gelsd. It computes the minimum-norm solution to a linear least squares problem using the singular value decomposition of A and a divide and conquer method. opt 4 is faster than opt 1/2/3 because it estimates X directly without calculating the A_inv matrix. opt 2/3 is better than opt 1 because numpy.linalg.inv() can not handle rank defiency of design matrix B Traditional Small BAseline Subsets (SBAS) algorithm (Berardino et al., 2002, IEEE-TGRS) is equivalent to the setting of: min_norm_velocity=True weight_sqrt=None Parameters: A - 2D np.ndarray in size of (num_pair, num_date-1) B - 2D np.ndarray in size of (num_pair, num_date-1), design matrix B, each row represents differential temporal baseline history between reference and secondary date of one interferogram y - 2D np.ndarray in size of (num_pair, num_pixel), phase/offset of all interferograms with no-data value: NaN. tbase_diff - 2D np.ndarray in size of (num_date-1, 1), differential temporal baseline history weight_sqrt - 2D np.ndarray in size of (num_pair, num_pixel), square root of weight of all interferograms min_norm_velocity - bool, assume minimum-norm deformation velocity, or not rcond - cut-off ratio of small singular values of A or B, to maintain robustness. It's recommend to >= 1e-5 by experience, to generate reasonable result. min_redundancy - float, min redundancy defined as min num_pair for every SAR acquisition inv_quality_name - str, inversion quality type/name temporalCoherence for phase residual for offset Returns: ts - 2D np.ndarray in size of (num_date, num_pixel), phase time-series inv_quality - 1D np.ndarray in size of (num_pixel), temporal coherence (for phase) or residual (for offset) num_inv_obs - 1D np.ndarray in size of (num_pixel), number of observations (ifgrams / offsets) used during the inversion rN)dtyperesidualg)mat_list)axis)cond)inv_quality_name weight_sqrtrL)reshapeshapenpzerosfloat32nanskip_invalid_obsrsumrinvdotT LinAlgErrorlstsqmultiplycalc_inv_qualitytilecumsum)ABy tbase_diffrmin_norm_velocityrcondmin_redundancyrrLnum_date num_pixelts inv_quality num_inv_obsXe2Zts_diffrArArBestimate_timeseries)sZ3           rcCs||jdd}||jdd}tj|jddftjdtj}t|||gd\}\}}ttj|dkdd|kr||St dt | }t t |j||tj}||||k<t|}|S) aEstimate the time-series STD from network of STD via linear propagation. Pixel by pixel only. For a system of linear equations: A X = y, propagate the STD from y to X. Parameters: G - 2D np.ndarray in size of (num_pair, num_date-1), design matrix y - 2D np.ndarray in size of (num_pair, 1), stack of obs y_std - 2D np.ndarray in size of (num_pair, 1), stack of obs std. dev. Returns: ts_std - 2D np.ndarray in size of (num_date, 1), time-series obs std. dev. rrr)r)rg)rg?)rrrrrrrrrdiagsquareflattenrrrrastypesqrt)Gry_stdrrts_stdZ stack_cov_invZts_varrArArBestimate_timeseries_stds  &  rcCsvtt|rnt|dddf}||ddf}x0t|D]$\}}|dk rF||ddf||<qFW||fS)aSkip invalid observations in the stack of phase/offset and update corresponding matrices. This applies to the pixel-wised inversion only, because the region-wised inversion has valid obs in all pairs. Parameters: obs - 2D np.ndarray in size of (num_pair, num_pixel), observations (phase / offset) of all interferograms with no-data value: NaN. mat_list - list of 2D np.ndarray in size of (num_pair, *) or None Returns: obs / mat_list Nr)rrisnanr enumerate)obsrrrsmatrArArBrsrc Cs|j\}}tj|tjd} ttd|} || krtt|| } |r^td || | xt | D]b} | | } t | d| |}|dkr|dd| |ft ||dd| |f}t tjtd|dd || | |<n|d kr|dk rZ|dd| |ft ||dd| |f}ttjt |d dd | | |<n4|| |jdkrt|| |ntj| | |<tdtt| d }|rj| d|dkrjtd | d| qjWn|dkr|t ||}t tjtd|dd |} nn|d krr|dk rT|| |}ttjt |d dd } n|jdkrjt|ntj} ntd|| S)aCalculate the inversion quality of the time series estimation. Parameters: G - 2D np.ndarray in size of (num_pair, num_date-1), design matrix A or B X - 2D np.ndarray in size of (num_date-1, num_pixel), solution y - 2D np.ndarray in size of (num_pair, num_pixel), phase or offset e2 - 1D np.ndarray in size of (num_pixel,), square of the sum of the 2-norm residual inv_quality_name - str, name of the inversion quality parameter temporalCoherence for phase residual for offset weight_sqrt - 2D np.ndarray in size of (num_pair, num_pixel), weight square root, None for un-weighted estimation. Returns: inv_quality - 1D np.ndarray in size of (num_pixel), temporalCoherence / residual )rgjAz=calculating {} in chunks of {} pixels: {} chunks in total ...rrNy?r)rrrz chunk {} / {}z&un-recognized inversion quality name: )rrrrintrz round_to_1ceilr\rVrangerrabsrexprrrmaxrU)rrrrrrrLnum_pairrr chunk_size num_chunkrsc0c1eZ chunk_steprArArBrs@   0,  0(4 $   rrcCst|}|jdd|j}|j}|jd|jd||d}tt|d|d}tt ||dd}tt||}|r|d krt d |t d ||ft d |g} x>t |D]2} | |} t || |g} d | || f} | | qW| |fS)a4Split into chunks in rows to reduce memory usage Parameters: dataset_shape - tuple of 3 int max_memory - float, max memory to use in GB print_msg - bool Returns: box_list - list of tuple of 4 int num_box - int, number of boxes F)rLrrrg?i@ rzmaximum memory size: %.1E GBz-split %d lines into %d patches for processingz" with each patch up to %d linesr)rrclengthwidthr numDaterrrrintr\rrappend) ifgram_file max_memoryrLZifg_objrrds_sizenum_boxy_stepbox_listrsy0y1boxrArArB split2boxes1s&      rr'cCst|jdd}t|d}|dkrNtj||jdkrtdtdnTtj||jdkrtdtd td td td td tdt|S)z< Check Rank of Design matrix for weighted inversion T)rrr)rz@WARNING: singular design matrix! Inversion result can be biased!z-continue using its SVD solution on all pixelszERROR: singular design matrix!z; Input network of interferograms is not fully connected!z6 Can not invert the weighted least square solution.zYou could try:zC 1) Add more interferograms to make the network fully connected:z6 a.k.a., no multiple subsets nor network islandsz8 2) Use '-w no' option for non-weighted SVD solution.) rrget_design_matrix4timeseriesrr matrix_rankrr\ Exception)r weight_func date12_listrrArArBcheck_design_matrixTs  rrGc Cs|j|dd}|r&td||||j|||dd|d}|dk rV|rtdnJd |jkr|rltd t|jd }|d dd}WdQRXnt d xDt |D]8} || ddfd k} || ddf| || 8<qW|S)a&Read unwrapPhase / azimuthOffset / rangeOffset from ifgramStack file Parameters: stack_obj - ifgramStack object box - tuple of 4 int ref_phase - 1D array or None Returns: stack_obs - 2D array of unwrapPhase in size of (num_pair, num_pixel) )rrzreading {} in {} * {} ...F) datasetNamerrrLrNzuse input reference valuerefPhasezread reference phase from filerzJNo reference value input/found on file! unwrapped phase is not referenced!g) get_sizer\rVreadrrdrrfilerr) rnr ref_phase obs_ds_namerrLr stack_obsrrsr1rArArBread_stack_obsls(   $r皙?c Cs|j|dd}|r||jkr|r4td||||j|||dd|d} d| t| <|dkrx|rtd|n\|d kr| |k} |rtd ||n6|d r| |k} |rtd ||nt d |tj || dk<|dk rtj || dk<~ ||fS)z5Mask input unwrapped phase by setting them to np.nan.)rrzreading {} in {} * {} ...F)rrrrLr)connectComponentz3mask out pixels with {} == 0 by setting them to NaN) coherence offsetSNRz3mask out pixels with {} < {} by setting them to NaN OffsetStdz3mask out pixels with {} > {} by setting them to NaNz#Un-recognized mask dataset name: {}gN) rrdr\rVrrrrendswithrUr) rrnr mask_ds_namemask_threshold stack_stdrrLrZmsk_datarArArBmask_stack_obss4 rcCsP|j|dd}|r$td|||jd||dd|d}d|t|<|S) z Read spatial coherence )rrz reading coherence in {} * {} ...r F)rrrrLrg)rr\rVrrrr)rnrrrLrZcoh_datarArArBread_coherencesr順c Cs`tdt|||d}|jd}d|jkr>t|jd}n$t|jdt|jd}|d}tt | td}tt ||}tdj ||d xt |D]} | |} t| d||} | d krd } nd } tj|d d | | f||d| d|d d | | f<t|d d | | f|d d | | f<| ddd krtd | d|qW|S)zTRead coherence and calculate weight_sqrt from it, chunk by chunk to save memory z-calculating weight from spatial coherence ...)rrrZ NCORRLOOKSALOOKSRLOOKSg ףp= ?zLconvert coherence to weight in chunks of {c} pixels: {n} chunks in total ...)cnrTFNg?)LepsilonrLz chunk {} / {})r\rrmetadatar`r=rrrrrrrVrrdecorZcoherence2weightr) rnrrrrweightrrrrsrrrLrArArBcalc_weight_sqrts4  *rc Cstjtj|j}tj|d}t|d}|dkrXtj|d}tj|sXd}|rtj|rtt j |t d t}|sd}|d7}t ||jd d }|jd d }|j||d d }t ||k}|||fS) z6Get the design matrix for time-series STD calculation.zsmallbaselineApp.cfgzmintpy.reference.dateautozreference_date.txtN)rz;reference date is required for time-series STD calculation,z'but NOT found in mintpy.reference.date!T)r)refDater)r]r^dirnamerjoinrryr_r;rloadtxtbytesrrUr get_date_listrarray) rnZ mintpy_dirZcfg_filer"ror date_listA_stdflag_stdrArArBget_design_matrix4stds$   r)Fc 9 Cst|} | jdd|r:|d|d} |d|d} n | j} | j} | | }| jdd}t|}tt |dtj d }t | d d}| j dd}| j|d dd\}}d }d }|d r(|dkrt| ||ddd}| rt| d d\}}|dkrd|}ndt| |dddd}nd|kr| sF|dkrtd|d|t|| j|d|ddd t|d }d|t|<d||dk<tdd|}| rt| d d\}}d|}|dkrd }n"|dkrntd|d|dntd|t| |||dd}d|krLtj||d k<td!|t|| ||||dd"\}}t|tj}|rtd#tj|t !|}t"|d$t"|d%}}||f| j| jfkrtd&t #|}d'd(|Dd} t j|| |d)d$}!|tj|!tjd*9}~!td+||tj%t|dd,9}tj&tj'|d-}"d.}#d|krtj&tj'|d/}"d0}#|"rtj(|"rtd1tj|"t j|"|d2d$}$||$d k9}~$t"t)|}%t*|d}&td3|%||%|d4t+||ftj }'t+||ftj }(t+|tj })d|krT|)tj9})t+|tj,}*|%dkr|' || | }'|( || | }(|) | | })|* | | }*|'|(|)|*|fS|d krRtj%t|dd,}+|+|9}+||+A},~t)|+dkr`td5|t)|+t)|+|%d4t-|||d d |+f|d || |#d6\}-}.}/|-|'d d |+f<|.|)|+<|/|*|+<t)|,dkrtd7|t)|,t)|+|%d4t"t)|,}0t*|,d}1tj.|0d8}2xt/|0D]|}3|1|3}4t-|||d d |4f|d || |#d6\}-}.}/|-$|'d d |4f<|.|)|4<|/|*|4<|2j0|3dd9d:|3d|0d;qW|21ntd<tj.|%d8}2xt/|%D]}3|&|3}4t-|||d d |4f||d d |4f|| |#d6\}-}.}/|-$|'d d |4f<|.|)|4<|/|*|4<|2j0|3dd9d:|3d|%d;qpW|21~| rtd=tj.|%d8}2xrt/|%D]f}3|&|3}4t2||d d |4f|d d |4f| d>}5|5$|(||4f<|2j0|3dd9d:|3d|%d;q*W|21~~|' || | }'|( || | }(|) | | })|* | | }*|d rd t3| j4d?d@tj5}6|'|69}'|(t6|69}(tdAn|dBk| j4dCdDk@rrt78| j4}7|7t3| j4dE}7|'|79}'|(|79}(tdF|7nZ|dGk| j4dCdDk@rt3| j4dH}8|8t3| j4dI}8|'d |89}'|(|89}(tdJ|8|'|(|)|*|fS)KaInvert one patch of an ifgram stack into timeseries. Parameters: box - tuple of 4 int, indicating (x0, y0, x1, y1) of the area of interest or None for the whole image ifgram_file - str, interferograms stack HDF5 file, e.g. ./inputs/ifgramStack.h5 ref_phase - 1D array in size of (num_pair), or None obs_ds_name - str, dataset to feed the inversion. weight_func - str, weight function, choose in ['no', 'fim', 'var', 'coh'] water_mask_file - str, water mask filename if available, to skip inversion on water min_norm_velocity - bool, minimize the residual phase or phase velocity mask_ds_name - str, dataset name in ifgram_file used to mask unwrapPhase pixelwisely mask_threshold - float, min coherence of pixels if mask_dataset_name='coherence' min_redundancy - float, the min number of ifgrams for every acquisition. Returns: ts/ts_std - 3D array in size of (num_date, num_row, num_col) inv_quality - 2D array in size of (num_row, num_col) num_inv_obs - 2D array in size of (num_row, num_col) box - tuple of 4 int Example: ifgram_inversion_patch('ifgramStack.h5', box=(0,200,1316,400)) F)rLrrrrT)rgv@r)rN)rGrR)r)Zsbasi)rrrr'g?rMzreading {} in {} * {} ...Std)rrrrLgY@g{Gzt?z,convert std. dev. to the inverse of variancezun-supported weight_func = z for !z(un-recognized observation dataset name: )rrphasegz/convert zero value in {} to NaN (no-data value))rrrrz6skip pixels (on the water) with zero value in file: {}LENGTHWIDTHz?Input water mask file has different size from ifgramStack file.cSsg|]}|dkr|qS)) waterMaskr1rA)rrrsrArArBrtsz*ifgram_inversion_patch..)rr)rz/skip pixels with {} = NaN in all interferograms)rz../avgSpatialCoh.h5rz../avgSpatialSNR.h5rz'skip pixels with zero value in file: {})rz2number of pixels to invert: {} out of {} ({:.1f}%)dz\estimating time-series for pixels with valid {} in all ifgrams ({:.0f} pixels; {:.1f}%) ...)rrrrrrz\estimating time-series for pixels with valid {} in some ifgrams ({:.0f} pixels; {:.1f}%) ...)maxValuez {}/{} pixels)everysuffixz1estimating time-series via WLS pixel-by-pixel ...zepropagating std. dev. from network of interferograms to time-series (Yunjun et al., 2021, FRINGE) ...)rrr WAVELENGTHg@z.converting LOS phase unit from radian to meterrP PROCESSORcosicorrrz=converting azimuth offset unit from pixel ({:.2f} m) to meterrQRANGE_PIXEL_SIZErz;converting range offset unit from pixel ({:.2f} m) to meter)9rrcrrr$rrr%r date_list2tbaserdiffrrrrerr)rar\rVrrrUrrronesbool_r]r^basenamerrTrget_dataset_listrrr!r r_rwhererint16r progressBarrupdatecloserr=rpirrzazimuth_ground_resolution)9rrrrrwater_mask_filerrrrcalc_stdrnnum_rownum_colrr&rtbaserrrrrrr'r(rr1Zatr_mskZlen_mskZwid_mskdsNamesdsNamer/Zstack_quality_filerquality num_pixel2inv idx_pixel2invrrrrZ mask_all_netZ mask_part_netZtsiZ inv_qualiZnum_obsiZnum_pixel2inv_partZidx_pixel2inv_partprog_barrsidxZts_stdi phase2range az_pixel_sizeZ rg_pixel_sizerArArBifgram_inversion_patch s                     &  & &      rTc* Cs|s t}t}td}t|j}|jdd|jdd}|jdd}|j |j }}|j |j |j dd|_||d}|jd|jdd} } | |_|jrt|d } d | } nd } d } |jrd }nd}| d|7} | d|j7} | d|j7} | d|j| 7} |jr|jdkr4d|j}nN|jdkrRd|j|j}n0|jdrrd|j|j}ntd|j| d|7} n| d7} tj||jdkr| d7} | d7} | d7} | d7} t| td | td!| td"|td#|t |j!}x&t"D]}t#t$|||t%|<q"Wd$|d%<d&|d'<|d|d(<tj&|tj'd)}|j(dd}|j)| f|gtj*| f|gtj*| ||fd*gd+}t+j,|j-||d,|jr t.j/0|j-d}|j 1d-r|d.7}|d/}| |d(<t+,|||d0t.j/2|j34kr2d0}d1|d'<n d2}d|d'<||d%<|5d(|d%tj*||fgi}t+j,|j3||d,d3|d%<d|d'<d3tj*||fgi}t+j,|j6||d,t7|j|j8d4\}}|j|j|j |j|j|j9|j|j|j|jd5 }xt:|D]\}}|d |d}|d6|d}|dkr^td7|d|td8|td9|||d:<|jst;f|d*d;\}}} }!ntd<t<| ||ftj*}t<| ||ftj*}t<||ftj*} t<||ftj*}!tj=|j|j>|j?d=}"|"|"j@t;|||| |!gd>\}}} }!|"Atd?d| |d|d6|d|d g}#t+jB|j-|d$|#d@|jrzt+jB||d$|#d@|d|d6|d|d g}#t+jB|j3| ||#d@t+jB|j6|!d3|#d@|dkrtCt|dA\}$}%tdB|$|%qW|j stD|j!dC}&tD|j!dD}'tdEtdF|&|'|d0kr>dnd}(tdG||(tEF|j3dH})|(|)||&|'f<Wd*QRXtdI| tEF|j6dH})| |)d3|&|'f<Wd*QRXtG|tCt|dA\}$}%tdB|$|%d*S)JzPhase triangulatino of small baseline interferograms Parameters: inps - namespace Example: inps = cmd_line_parse() ifgram_inversion(inps) rFF)rLT)r)unwDatasetNameskip_referencerrrrz with REF_DATE = rKzP------------------------------------------------------------------------------- zdeformation velocityzdeformation phasez/least-squares solution with L2 min-norm on: {} zminimum redundancy: {} zweight function: {} zcalculate STD: {} {} )r z{} == 0)r r z{} < {}r z{} > {}z#Un-recognized mask dataset name: {}zmask out pixels with: {} z mask: no z0***WARNING: the network is NOT fully connected. z! Inversion result can be biased! zF Continue to use SVD to resolve the offset between different subsets. zO-------------------------------------------------------------------------------znumber of interferograms: {}znumber of acquisitions : {}znumber of lines : {}znumber of columns : {}rrEmUNITREF_DATE)rN)datebperpr)rrGZDecorzStd.h5rpixelrr1)r) rrrrrrFrrrrGrz5 ------- processing patch {} out of {} --------------zbox width: {}zbox length: {}rrz6 ------- start parallel processing using Dask -------) config_name)func func_dataresultsz.------- finished parallel processing ------- )datarblock<z(time used: {:02.0f} mins {:02.1f} secs. rOrNz2--------------------------------------------------z.update values on the reference pixel: ({}, {})z$set {} on the reference pixel to {}.zr+z3set # of observations on the reference pixel as {})Hrptimerset_num_threadsrrrcrr$rrget_reference_phaser r#rrrr r0r)rrVrrrrr rUrrrr\rXrrr;rxrqr%string_get_perp_baseline_timeseriesrrr layout_hdf5rfr]r^splitextrer=rgrapoprhrrwrrrTrrYr[rvrrCwrite_hdf5_blockdivmodrrrroll_back_num_threads)*rj start_timenum_threads_dictrnrr&rrrrrZ ref_date4stdZref_msgror4metarmdatespbase ds_name_dictfbaseZ tsStdFilerrr data_kwargsrsrbox_widbox_lenrrrr cluster_objrbrWsref_yref_xref_valrrArArBifgram_inversion_s.                        r~cCsBt|}|jr t|dkr |jS|jdkr4t|ntd|jS)Nrr.z)L1 norm minimization is not fully tested.)rpr3rrr-r~NotImplementedError)rirjrArArBmainas  r__main__r)N)NTrrrT)rr)rNT)rT)r')rGTT)NrNTT)TT)r'Tr) NNrGr'NTNrrF)N)N)1r]sysrdr4rnumpyrscipyrmintpy.objectsrrZmintpy.simulationrrmintpy.defaults.templater mintpy.utilsrrr r rzr rqrr9r8r7rCrprWrrrrrrrrrrrr)rTr~r__name__argvrArArArB sf    ? J(5 } % M #  ) *  /  ?