Comparison of Full-waveform, Single-Photon Sensitive, and Discrete Analog LiDAR Data

Transcription

1 Comparson of Full-waveform, Sngle-Photon Senstve, and Dscrete Analog LDAR Data Angela M. Km, Scott C. Runyon, Rchard C. Olsen Naval Postgraduate School, Remote Sensng Center and Physcs Department, 833 Dyer Road, Monterey, CA, USA; ABSTRACT Full-waveform LDAR data from an AHAB Chroptera I system wth 55 nm and 032 nm lasers ( 0 pts/m 2 ), sngle-photon senstve data from the Sgma Space HRQLS system wth a 532 nm laser ( 9 pts/m 2 ), and dscrete analog data from an Optech Oron C200 system ( 88 pts/m 2 ) were collected from aeral platforms over Monterey, CA, USA n fall 202 and fall 203. The study area contans resdental neghborhoods, forested regons, nland lakes, and the Pacfc Ocean near-shore envronment. Sgnfcant ground truth n the form of GPS measurements and terrestral LDAR scans enable the LDAR data to be compared n terms of measurement precson and degree of tree canopy penetraton, as well as comparsons of derved raster products. Keywords: LDAR, full-waveform, sngle-photon senstve, dscrete analog. INTRODUCTION The motvaton of ths work was to understand f, and to what extent, dfferent modaltes of collectng LDAR data result n data wth dfferent characterstcs, f these data requre dfferent processng strateges, and f there are dfferences n derved products. Three datasets were collected wth dfferent LDAR systems, ncludng a sngle-photon senstve system, a full-waveform system, and a dscrete analog system. The datasets were all collected over the same regon of the Monterey Bay Pennsula n Monterey, CA, at smlar tmes of year n 202 and DATASETS Collecton parameters and system detals for the three aeral LDAR collects are gven n Table. An mage of the Optech Oron C-200 pont cloud, wth RGB values extracted from the concurrently collected Ultracam Eagle orthophotos, s gven n Fgure. The Sgma Space HRQLS data were collected wth east-to-west orented flght lnes. Because the system uses a crcular scannng pattern, each pont on the ground s vewed at least twce - once from the front, and agan from the back at the opposte sde of the scan. The swath wdth of each flght lne was large enough (,500 m) to cover the entre study area, so wth each of the three passes, the entre scene was captured wth both the forward and backward vewng geometry of the sensor. Flght lnes for the AHAB Chroptera I system were orented both east-to-west and north-to-south. For the Optech Oron C-200 data, the study area was surveyed wth opposng east-to-west flght lnes wth sde lap of greater than 60% wth at least 00% overlap. Further author nformaton: A.M.K.: E-mal: amkm@nps.edu, Telephone: Laser Radar Technology and Applcatons XX; and Atmospherc Propagaton XII, edted by Monte D. Turner Gary W. Kamerman, Lnda M. Wasczko Thomas, Earl J. Spllar, Proc. of SPIE Vol. 9465, 94650L 205 SPIE CCC code: X/5/$8 do: 0.7/ Proc. of SPIE Vol L- Downloaded From: on 05/27/207 Terms of Use:

2 Table : System parameters for each of the Monterey Bay LDAR collects. Sngle Photon Senstve Full- Waveform Dscrete Analog System Sgma Space HRQLS AHAB Chroptera I Optech Oron C -200 Scannng Pattern 0x0 square array of 00 Palmer scannng (ellptcal); beamlets, crcular scannng 20 sdeways; 4 forward and aft Sawtooth Wavelength 532 nm Green "Hydro" 55 nm Red "Topo" 064 nm 54 nm Laser Footprnt Each beamlet: 50 cm; 0x0 Green "Hydro" 55 nm - m; beamlet array: 5x5 m Red "Topo" 064 nm - 30 cm 50 cm Collecton Platform Kng Ar B200 Arcraft Partenava Arcraft Bell 206L Helcopter Collecton Alttude 2,300 m 360 m 450 m Swath Wdth,500 m 30 m 280 m Pulse Rate Each beamlet: 25 khz - Green "Hydro" 55 nm - 35 khz; 2.5 mllon measurements /sec. Red "Topo" 064 nm khz 66 khz Green "Hydro" 55 nm: Pont Densty* 8.7 pts/m2 9.4 pts/m2 and 3. pts /m2 88 pts /m2 Overall/ 5.7 pts /m2 Red "Topo" 064 nm: 33 pts /m2 Sngle Flght Lne 9.9 pts/m2 and 3.3 pts/m2 Collecton Date 7 Nov Oct Green laser only; May Red and Green lasers 6 Oct -4 Nov 202 Orthophotography Hasselblad, 4.9 cm GSD, Ultracam Eagle, 5cm None Collecton 3- channel GSD, 4- channel *Pont densty as reported by QT Modeler for the study area of nterest. References: 3 7 Fgure : Pont cloud data from the Optech Oron C-200 system, wth RGB values extracted from the concurrently collected Ultracam Eagle orthophotos, showng the study area. Proc. of SPIE Vol L-2 Downloaded From: on 05/27/207 Terms of Use:

3 2. Optech Oron C-200 Data from the Optech Oron C-200 s the standard commercal LDAR data product. The data s hgh-qualty, wth very hgh accuracy and pont cloud densty. 2.2 AHAB Chroptera I The AHAB Chroptera I s a dual-wavelength system, havng both a green 55 nm Hydro laser and a red 064 nm Topo laser. The AHAB system produces pont cloud data by onlne waveform processng, and also dgtzes and records the full-waveforms from each of the lasers for offlne post-processng. A comparson of the pont cloud data can be made wth the waveforms by assocatng the two data streams based on GPS-tme of the laser shot. Example waveforms extracted from a Coast Lve Oak (Quercus agrfola) tree canopy, and the assocated derved LDAR returns, are shown n Fgure 2. Informaton necessary to precsely locate the waveform n space (such as the pontng angle of the laser) were unavalable. It was therefore mpossble to convert the waveforms to Heght Above Ellpsod values for drect comparson wth the LAS pont cloud returns. Instead, the waveforms were converted to unts of meters, measured from the start of waveform recordng based on the samplng frequency and number of samples n the waveform. In the plots n Fgure 2, the waveform peaks and LAS returns have been manually and approxmately algned. These plots allow a comparson of shape, and demonstrate the possblty of extractng further useful nformaton from the full-waveform sgnals. In ths case, the ground return s mssed wthn the 064 nm Topo waveform (Fgure 2, Left), and the top of the canopy s mssed n the 55 nm Hydro waveform (Fgure 2, Rght). The shape of the waveform peaks may also provde nformaton useful for characterzaton of the tree canopy. Processng of the AHAB Chroptera I data revealed elevaton offsets between data from the two lasers, and elevaton offsets between flght lnes. It was determned that these offsets are due to an mproper calbraton coeffcent beng used when the pont cloud data were processed, and can therefore be corrected wth re-processng of the dataset. However, the re-processed data were not avalable at the tme of ths wrtng..4 v.2 2 'ñ Q 0.8 ÿ E 0.2 Z ?: 508 ÿ E 0.2 z Waveform May 204 Flghtlne 54, Tle 6A, Topo (064 nm) Laser Transmsson tme TO: Dstance From Waveform Start (m) Extracted Returns 4 2 (x,y,z): , , Q 0.8 m 0.6 N 0.4 Eó 02 z (7, c m E 0.8 Ñ 0.6 o 0.4 E 02 Z 30 Waveform May 204 Flghtlne 54, Tle 6A, Hydro (55 nm) Laser Transmsson tme TO: Dstance From Waveform Start (m) Extracted Returns (x,y,z): , , Heght Above Ellpsod (m) Heght Above Ellpsod (m) -6-4 Fgure 2: (Top row) AHAB waveforms and (bottom row) assocated extracted ponts. The gven xyz-locaton s for the return ndcated by the black arrow. The 064 nm Topo waveform shows that the ground return s mssed n the extracted returns, whle the top of the canopy s mssed n the 55 nm Hydro waveform. 2.3 Sgma Space HRQLS The Sgma Space Hgh Resoluton Quantum Ldar System, HRQLS, was desgn wth the goals of hgh-effcency wde-area mappng. 3 The collecton area for the Monterey Bay collect was too small to demonstrate the effcency of the system. Proc. of SPIE Vol L-3 Downloaded From: on 05/27/207 Terms of Use:

4 The data presented n ths work were collected as part of one of the very frst operatonal flghts of the HRQLS system, and therefore contan some artfacts and defcences. Snce then, sgnfcant mprovements have been made n the operaton of the system and processng of the data. Samples of data from more recent collects wth the HRQLS system show sgnfcantly better data qualty than what was acheved wth the Monterey Bay collect. Despte the defcences n the Monterey Bay data, t was stll possble to derve acceptable raster products from the data. 3. COMPARISON OF DIGITAL ELEVATION MODELS (DEMS) AND DIGITAL SURFACE MODELS (DSMS) Dgtal Elevaton Models (DEMs) and Dgtal Surface Models (DSMs) wth m GSD pxels were created from each of the pont cloud datasets usng the commercal ENVI LDAR package from EXELIS Vsual Informaton Solutons wth the default settngs. The processng settngs wthn the ENVI LDAR software can be adjusted to refne the DEM/DSM extracton, but n ths case, the default settngs produced an acceptable product n all cases. Mantanng the settngs n the default mode also helped n the comparson of products, as dfferences can be attrbuted to the LDAR data, rather than dfferences n processng strateges. RGB compostes of DEMs and DSMs were created wth the Optech Oron C-200 products n the red channel, AHAB Chroptera I 064 nm n the green channel, and Sgma Space HRQLS n the blue channel; Fgure 3 shows the entre scene area, and a zoomed-n vew of a small subsecton s shown n Fgure 4. Dfferences n the DEM and DSM products appear as varatons n color n the RGB composte mages. The unformty of color suggests there are only mnor dfferences n the products, manly due to dfferences n vegetaton extracton n the DSM product, and dfferences n bare earth extracton under buldngs n the DEM product. The dfferences n bare earth extracton under buldngs s expected to have more to do wth the grddng algorthm than beng an ndcaton of actual dfferences n the data. Some of the apparent dfferences n vegetaton are due to real changes; for example, tree trmmng was accomplshed between collecton dates. A slght geographc msalgnment s apparent n the RGB composte of the DSM product (Fgures 3 and 4, rght) as a brght green edge on northwest sdes of all of the objects. Fgure 3: RGB composte mages created from Optech Oron C-200, AHAB Chroptera I 064 nm, and Sgma Space HRQLS (L) DEMs and (R) DSMs. Proc. of SPIE Vol L-4 Downloaded From: on 05/27/207 Terms of Use:

5 Fgure 4: Zoomed n vew of RGB composte mages created from Optech Oron C-200, AHAB Chroptera I 064 nm, and Sgma Space HRQLS (L) DEMs and (R) DSMs. A quanttatve comparson was made by examnng the relatve elevaton dfferences between the DEM and DSM products. The products were frst adjusted by means of a vertcal offset to match the measured ground truth to account for dfferences n coordnate systems, and to dscount dfferences due to processng strateges. The Optech Oron C-200 data, due to the low-and-slow collecton parameters, had the hghest pont densty of the collected data, and for ths reason, the Optech data were used as the reference for comparson of the AHAB and Sgma Space products. Hstograms of the relatve dfferences between the derved products are shown n Fgure 5. The 95% threshold level was determned for each of the dfference products as follows: Optech Oron - AHAB Chroptera DEM - 95% of values fall wthn ± 0.7 m Optech Oron - AHAB Chroptera DSM - 95% of values fall wthn ± 0.2 m Optech Oron - Sgma Space Chroptera DEM - 95% of values fall wthn ± 0.34 m Optech Oron - Sgma Space Chroptera DSM - 95% of values fall wthn ± 0.09m 8 4x0 ac 3x0 O 2x0 ó x0 E Z.5 Optech Oron -AHAB Chroptera DEMs Relatve Elevaton Dfference (m) 5 y d 5x0 ; 4x0 O 3x04 0 2x0 Éx0 03 Optech Oron - AHAB Chroptera DSMs Relatve Elevaton Dfference (m) 0,05 m 8x0 8 8x0 O 5 4x0 Optech Oron - HRQLS DEMs 2x0 E Z O5 to 5 Relatve Elevaton Dfference (m).4x0s.2x05.ox05 8.0X 0 B.Ox 0 4.0x0 2.0x04 o Optech Oron - HRQLS DSMs Relatve Elevaton Dfference (m) Fgure 5: Hstograms of relatve elevaton dfferences between the AHAB Chroptera I 064 nm, Sgma Space HRQLS, and Optech Oron C-200 DEMs and DSMs. The Optech Oron C-200 data were used as the reference data. Proc. of SPIE Vol L-5 Downloaded From: on 05/27/207 Terms of Use:

6 There was a slght geographc msalgnment between the AHAB Chroptera I data and the Optech Oron C-200 reference. Ths may partally explan the seemngly large dfferences between the AHAB and Optech Oron derved products. Images colored accordng to relatve elevaton dfferences llustrate the locatons of dfferences n the derved products (Fgure 6 and 7). The largest dfferences seem to mostly be due to dfferences n vegetaton detecton n the DSM product. Because of dfferences n collecton strateges, the Optech Oron reference dataset has a sgnfcantly hgher pont densty than ether of the other two datasets whch may explan the dfferences n DSM results. o Elevaton Dfference (m) < -2.0 _--2.0 to to to to to to.0.0 to.5.5 to 2.0 ;,:; s": > 2.0,,,: Fgure 6: (Top row) Dataset overvew and (Bottom row) zoomed vew of relatve elevaton dfferences between the AHAB Chroptera I and the Optech Oron C-200 (L) DEM and (R) DSM products. Proc. of SPIE Vol L-6 Downloaded From: on 05/27/207 Terms of Use:

7 - Fgure 7: (Top row) Dataset overvew and (Bottom row) zoomed vew of relatve elevaton dfferences between the Sgma Space HRQLS and the Optech Oron C-200 (L) DEM and (R) DSM products. 4. DEGREE OF CANOPY PENETRATION Two trees a Coast Lve Oak (Quercus agrfola), and a Southern Magnola (Magnola grandflora) were selected for the purposes of examnng the degree of canopy penetraton. The Southern Magnola has a very dense canopy. Terrestral laser scannng data acqured wth a Regl VZ-400 system (550 nm) show gaps n the canopy due to falure of the scanner to penetrate to the top of the canopy (Fgure 8). Due to extended drought condtons, the Coast Lve Oak tree canopy n ths case s unhealthy and very sparse. Because of dfferent collecton strateges, a quanttatve comparson of degree of canopy penetraton s not reasonable. The systems have dfferent laser wavelengths, collecton geometres, and scannng patterns, whch result n dfferent pont cloud denstes. The degree of canopy penetraton for any of the systems may be sgnfcantly dfferent usng other collecton strateges. Pont clouds from a sngle flght lne were selected from each of the datasets for vsual comparson (Fgure 9). The Optech Oron C-200 data had the hghest pont densty, and some of the branchng structure s vsble wthn the Coast Lve Oak canopy. In the case of the Sgma Space HRQLS data, these data were collected as part of an ntal system shake down flght, usng prelmnary flterng strateges. An examnaton of the raw data, pror to flterng, reveals the exstence of returns from the ground under the Magnola tree canopy that were erroneously removed due to too aggressve flterng. Proc. of SPIE Vol L-7 Downloaded From: on 05/27/207 Terms of Use:

8 An examnaton of the waveform data from the AHAB system, for example see Fgure 2, shows that addtonal ground and canopy ponts could be extracted from the waveform data. Heght Above Ellpsod (m) T rah `t' -' '~ -4 Y 3 ' f' + :,,.. " r. Fgure 8: (L) Pont clouds from the terrestral laser scans wth the Regl VZ-400 of the Coast Lve Oak (left) and Southern Magnola (rght). (R) Top vew of the two trees from WSI Ultracam Eagle 5 cm GSD orthophoto.,. : _..ra,,.;^.,.<. Ñ f 4, F..Hr:4: y " Return r:53'r. ;54. : : 'et". ' M t. t+r k. r s;?y!. I t h k Vet/ ç!e,3 'f s ;. ^:3ü.!y,y,n. Frst Intermedate,q; Last rt.f I Ff.t;.t:4, +ï }. _ Jy, `: : ; ;:: 2 f?: f f. ' ;.!;: :..s ':T..r: Return =Frst =Intermedate =Last Rahn Frst I= Intermedlate..t!l, `: :-- racsl:.=..r...t T`rr:T. rr. Last.r.7t Ts:O,Yf`!t r yts' l r e! :".. f {."..a...p.3.. :..!'!, ' l':';l Fgure 9: Pont clouds from sngle flght lnes for each of the arborne laser scanners of the Southern Magnola (left) and Coast Lve Oak (rght). Proc. of SPIE Vol L-8 Downloaded From: on 05/27/207 Terms of Use:

9 I I I 5. MEASUREMENT PRECISION Precson refers to lkelhood of achevng the same value wth repeated measurements. A profle from each of the datasets from a smooth, flat surface was used to quantfy the precson each of the systems acheved n these data collects (Fgure 0). Because of the prevously dscussed ssues wth the Monterey Bay Sgma Space HRQLS data (see Secton 2.3), data from a more recent (November 204) collect over Easton, MD were used here. The collecton parameters for the Easton, MD collect were smlar to the Monterey Bay collect 2.3 km collecton alttude and 75 knots arspeed. For all of the data except the AHAB Chroptera I Hydro data, a vertcal precson of approxmately 0. m was acheved. The AHAB Chroptera I Hydro data has a vertcal spread closer to 0.2 m n ths case. As mentoned n Secton 2.2, a vertcal offset (almost 20 cm here) was observed between the AHAB Chroptera I Hydro and Topo lasers. Ths offset s most lkely due to an ncorrect calbraton parameter beng used when processng the raw data, and can be corrected wth re-processng. The re-processed data were not avalable at the tme of ths wrtng. al o a Y -r I r -- -j Ì 0. m m o J l d 4 '_ -' a koztagempsg*fationtter# I f `` r `69... óó6)ó3 e B68M,9.^cóó 8 Opo càóó ßP & r 0 o ó D804Y 4 0cV 8 a$ go 9 0 Fgure 0: Profles from a smooth, flat, surface from (top to bottom) the AHAB Chroptera I Hydro 55 nm and Topo 064 nm lasers, Optech Oron C-200, and Sgma Space HRQLS system. Profles from each of the systems are vertcally offset for clarty; however, the vertcal offset between the Hydro and Topo lasers of the AHAB Chroptera I system does exst n the data due to use of an ncorrect calbraton parameter durng processng of the raw data. 6. DATA MODALITY STRENGTHS AND CHALLENGES 6. Dscrete Analog Pont Cloud Data The Optech Oron C-200 data used n ths work s a good example of dscrete analog LDAR data. Ths s currently the most common type of LDAR data, and s typcally hgh-qualty data wth known processng schemes and readly avalable software. The data s delvered n standard formats, and vrtually no pre-processng s requred pror to extractng data products such as DEMs or DSMs. 6.2 Full Waveform Full-waveform data s becomng more common, but s stll relatvely novel (especally n the topographc mappng realm). Full-waveform bathymetrc systems are more common. It s apparent that extra nformaton s present n full-waveform LDAR sgnals; ths can be observed by examnng full-waveform data and the assocated extracted LDAR returns (for example, see Fgure 2). It s hypotheszed that ths extra nformaton wll be useful for tasks Proc. of SPIE Vol L-9 Downloaded From: on 05/27/207 Terms of Use:

10 such as classfcaton or object detecton, as well as provdng the opton of refnng the peak extracton results as performed onboard the system. At ths pont n tme, software optons for workng wth full-waveform LDAR data are extremely lmted. Whle the ASPRS standard LAS format s almost unversally used for pont cloud data, full-waveform data s often n sensor specfc formats, wth varyng degrees of metadata avalable. 6.3 Sngle Photon Senstve/Photon Countng Ths LDAR modalty offers great potental for collectng hgh-qualty LDAR data extremely effcently. The Monterey Bay data from the Sgma Space HRQLS system suffered from some farly sgnfcant ssues; however, t was stll possble to extract useable DEM and DSM products. More recent data collects show a great mprovement n data qualty. 7. CONCLUSION Data were presented from three separate LDAR systems, ncludng full-waveform, tradtonal dscrete analog, and sngle-photon senstve systems. Some mnor dfferences n the derved data products (DEMs and DSMs) were observed - mostly due to dfferences n vegetaton extracton n the DSM product. These dfferences are most lkely due to the varyng pont denstes of the data, whch s largely a factor of the collecton parameters used. All of the systems had acceptable degrees of penetraton through tree canopes. The Monterey Bay data from the Sgma Space HRQLS system showed some gaps n ground coverage due to too aggressve flterng more recent data collectons have shown sgnfcant mprovement n ths area. Quanttatve comparsons of measurement precson showed smlar performance for each of the systems (approxmately m vertcal spread). Each of the LDAR collecton modaltes has strengths and weaknesses, makng the optmal choce dependent on end-user goals. REFERENCES. Watershed Systems, Inc., Data collecton report. Data collected for Naval Postgraduate School. 2. N. D. Quadros, Characterstcs of bathymetrc LDAR sensors, LDAR Magazne 3(6), J. J. Degnan and C. T. Feld, Moderate to hgh alttude, sngle photon senstve, 3D magng LDARs, Proceedngs of SPIE Defense, Securty, and Sensng 94, 204. Proc. of SPIE Vol L-0 Downloaded From: on 05/27/207 Terms of Use: