Using Digitized Spacecraft Film and a Revised Lunar Control Network for Photogrammetric Mapping

USING DIGITIZED SPACECRAFT FILM AND A REVISED LUNAR CONTROL NETWORK FOR PHOTOGRAMMETRIC MAPPING

Mark R. Rosiek R.L. Kirk B.A. Archinal T.L. Becker L. Weller B. Redding E. Howington-Kraus D. Galuszka U.S. Geological Survey 2255 N. Gemini Dr Flagstaff AZ 86001 [email protected]

ABSTRACT

This paper reports on the results of photogrammetric mapping with Lunar Orbiter, Apollo panoramic and Apollo metric camera digitized photographs using modern softcopy digital mapping techniques and a revised lunar control network to establish control. We will report on the differences between Digital Elevation Models (DEMs) collected from Lunar Orbiter, Apollo panoramic and metric camera digitized photographs. Our test area is the Rima Hadley region, which includes the Apollo 15 landing site. This area is covered by multiple sources of data that can be used for comparison and evaluation of accuracy. The original Lunar Orbiter photographs reconstructed in the 1960s had limited utility for mapping due to a stair-step offset in the reconstructed photographs. These images have been re- cently digitized and digitally reconstructed to fit the reseaux and the fiducials to the camera calibration data. The Apollo imagery was previously used to produce topographic maps using different control networks. These previous networks have kilometer-size offsets between them and were also limited in size and accuracy. A revised lunar control network, the Unified Lunar Control Network 2005, improves the accuracy and the density of control points and includes solved-for for elevations of the control points. We find that all the image sets studied provide useful DEMs with minimal requirement for interactive editing but varying tradeoffs between area covered and resolution. The Apollo metric camera data are the cleanest and would support topomapping of nearly 20% of the Moon at 50 m post spacing. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorse- ment by the U.S. Government.

INTRODUCTION

We are evaluating the utility of using modern softcopy digital mapping techniques for extracting digital elevation models (DEMs) from Lunar Orbiter (LO) and Apollo digitized photographs. These photographs were used in the 1960s-70s for mapping, mission planning, and control purposes. Previous work with LO imagery was difficult due to image artifacts, and the mapping was not controlled to any horizontal or vertical datum (Hansen, 1970). Mapping with Apollo imagery used 2 different control networks. The hardcopy maps that resulted are the Lunar Topographic Orthophotomap (LTO) and Lunar Orthophotomap (LO) map series (Wu and Doyle, 1990). Since these maps were produced, the lunar control network has been improved (Archinal et al., 2006a, b) and the LO images are now available as digitized photographs with the artifacts greatly reduced (Gaddis et al., 2003, Weller et al., 2006). The tools available for softcopy mapping permit the generation of DEMs, from which contour maps can be made if desired, but the digital DEMs contain more detailed topographic information that can also be used for image rectification, photometric correction, slope analyses, etc. Use of the current control and datum en- sures that the topographic data can be used in conjunction with data from current and future missions. Thus the products enabled by this study supersede earlier maps and will be useful for upcoming lunar missions including SELENE – Japan, Chang'e 1 – China, Chandrayaan-1 – India, and Lunar Reconnaissance Orbiter – USA.

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 SOURCE DATA

Lunar Orbiter Imagery LO photographs were collected by five missions, LO-I through -V, during 1966–1967. Stereo models are provided both by the overlap of photographs for global mapping and by deliberate targeting of specific sites of interest. LO IV coverage of the nearside includes 6°-wide bands of stereo at 0° and ±30° latitude, and from about ±60° latitude to the poles (>25% of the nearside in all). The resolution of the imagery is between 30 and 100 m. The other LO missions returned stereo imagery of spot areas with higher resolution (10 - 40 m) (Hansen, 1970). These images were used for Apollo landing site selection, but the full topographic information was not extracted because errors in reconstructing the photographs from sections scanned on the spacecraft produced artifacts in the form of linear "cliffs" in the stereomodels.

Apollo Imagery Apollo 15, 16, and 17 photographed ~20% of the Moon immediately under their orbital tracks using both a frame mapping camera and a panoramic camera. The frame camera was a Fairchild metric camera with a 4.5 x 4.5 in film format. Stereo models are obtained by overlapping photographs along the flight line and between flight lines. When digitized at 10 µm, a metric camera photograph provides a useful resolution of about 15 m/pixel. The panoramic camera was an Itek panoramic camera with 45.24 x 4.5 in film format. Stereo models are obtained by using forward and aft looking photographs acquired along the same flight line. When digitized at 10 µm, a panoramic photograph has a resolution at image center of about 2 m/pixel and at the edge of the image the resolution is about 4 m/pixel. Additional information on the digitization process is provided below.

Clementine In 1994, the Clementine spacecraft acquired digital images of the Moon at visible and near infrared wave- lengths, as well as laser altimeter data (Nozette et al., 1994). The Clementine laser altimeter (Lidar) data have a vertical accuracy of approximately 100 m, horizontal accuracy 3,000m, a surface spot size of 200m, and a spatial resolution of 2.5 . Altimetry data were collected between 79 S - 81 N (Smith et al., 1997, Neumann, 2001). The altimeter data and polar Clementine stereo data (Rosiek et al., 2002) was used to initialize the radii values in the ULCN 2005 control network described in the next paragraph. We will also use the altimeter points to analyze the DEMs that are collected from Apollo and LO images.

Control We used control from two sources. First, Table 1. Lunar Horizontal Control Net Comparison. for horizontal and vertical control we used Name # points # images Horz. Acc. Vert. Acc. points from an interim solution of the Uni- ULCN 1,478 Unknown 100 m to 3 km Few km? fied Lunar control Network 2005 (ULCN Few km to CLCN 271,634 43,871 Sphere 2005). This network combines the Unified some>15 km Lunar Control Network (ULCN) and the ULCN 2005 272,931 43,871 Few km ~ 1 km or less Clementine Lunar Control Network (CLCN). Second, we also used points near the Apollo 15 landing site that were described in Davies and Colvin (2000). The ULCN was described in the last major publication about a lunar control network (Davies et al., 1994). See Table 1 for statistics on this and the other networks discussed here. Images for this network are from the Apollo, Mariner 10, and Galileo missions, and Earth-based photographs. The importance of this network is that its accuracy is relatively well quantified and published information on the network is available. The CLCN includes measurements on 43,871 Clementine 750-nm images. The purpose of this network was to determine the geometry for the Clementine Basemap Mosaic (CBM) (USGS, 1997). After the completion of the CBM, horizontal errors of 15 km or more were noticed and therefore these same errors are present in the CLCN (Malin and Ravine, 1998; Cook et al., 2000; Cook at al, 2002). The errors seem to have arisen for several reasons, including that only a few (22) near side points were fixed to ULCN positions, the camera angles were uncon- strained, and the tie points were all constrained to lie on a mass-centered sphere of radius 1736.7 km. We have merged the ULCN and CLCN and have addressed the horizontal accuracy problems of the CLCN, with the intent to create a new ULCN. Our new solution(s) include 3 changes. 1) The camera angles are constrained to within 0.03° of their a priori Navigation and Ancillary Information Facility (NAIF) values (Acton, 1999). 2) The coordinates of all identifiable ULCN points are constrained to their reported accuracy (Davies et al.,

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 1994). 3) Radii of all tie points are solved for. Our current results show horizontal position changes from the CLCN on average of ~7 km with some changes of dozens of km. The final results are in preparation (Archinal et al., 2006b).

Site Map The site mapped in this project is the Rima Hadley region, including the Apollo 15 landing site. This area has ex- cellent coverage by LO and Apollo images, and previous mapping products exist to compare with our results. We selected images 4102_H3 and 4103_H1 from LO IV, images 5105_MED, 5106_MED, 5107_MED, 5106_H1, 5106_H2, and 5106_H3 from LO V, images 0583, 0585, and 0587 from the Apollo 15 metric camera, and images 9809, 9811, 9814, and 9816 from the Apollo 15 panoramic camera. Figure 1 shows the image footprints.

PROCEDURES

We are using a commercial photogrammetric workstation with SOCET SET (® BAE Systems) software to view the images in stereo, select control and tie points, and collect Figure 1. Footprints of images used in this our DEMs. The LO and Apollo images, obtained by digitiza- study, shown on a Clementine image mosaic tion of film as described below, are imported into the work- with north at top. Blue: LO IV frames 4102_H3 station in TIFF file format. Support data were obtained from and 4103_H1; Green: LO V frames 5105_MED, the National Space Science Data Center and entered by hand. 5106_MED, and 5107_MED; Red: LO V Camera model information for the LO missions and Apollo frames 5106_H1, 5106_H2, and 5106_H3; missions was entered into camera model files used by the White: Apollo 15 metric frames 0583, 0585, and workstation. The workstation provides generic camera mod- 0587; Orange: Apollo 15 forward looking pan els for frame and panoramic sensors, which use the camera frames 9809 and 9811; Yellow: Apollo 15 aft model files to handle specific cameras. looking pan frames 9814 and 9816.

Digitization LO images, including those needed for this study, are being digitized and reconstructed under separate projects (Gaddis et al., 2003, Weller at al, 2006). This process includes scanning the 35 mm framelets at 25 µm spot size, averaging the pixels to 50 µm raster (which preserves image detail better than scanning at 50 µm), and then geomet- rically transforming and mosaicking the framelet images to reconstruct the LO sub-frames. The reconstruction process is guided by measurements of reseau marks on the framelets and fiducials along the frame edges, and largely eliminates the discontinuities present in earlier, hand-mosaicked versions that produce artifacts in stereo mapping. We have 3rd generation negatives and 4th generation film positives of the Apollo photographs in our data hold- ings and selected the latter to digitize. The original 1st generation negatives or 2nd generation positives available at Johnson Space Center or Lunar and Planetary Institute could be scanned for future operational mapping and might yield slightly better image quality, but would also incur additional costs due to the need for special approval and handling and the use of a clean room. Regardless of the image source, images to be used in topomapping must be digitized with a photogrammetric-quality scanner to avoid introducing geometric distortions in the images that will affect mapping, and must allow for digitization of fiducial and timing marks around the image edges. We tested digitizing the images with 5, 10, and 25 µm spot size with a Vexcel Ultrascan 5000 scanner. Al- though the 5 µm spot size provided the most detailed images, the improvement in subjectively evaluated detail over 10 µm was minimal and did not justify the fourfold larger file sizes. Scanning at nominal exposure times and 8 bits/pixel was found to yield equal image quality, even in highlights and shadows, compared to long-exposure scanning at 16 bits/pixel. Therefore the former scanning method was adopted. An Apollo metric photograph can be digitized in 1 pass, and at 10 µm spot size the file size is 140 MB. The 45.24 in length of a panoramic photograph

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 exceeds the length of the scanner used, so these images were digitized in 5 overlapping sections. Timing marks are recorded every 2.5° along the edges of the panoramic film, and the images were digitized so that each section shares a common timing mark with its neighbors. With a 10 µm spot size, the file size is 300 MB for each section. The sections were used individually for mapping, rather than being reconstructed into complete panoramic images.

Interior Orientation Interior orientation is the reconstruction of the location of the scanned pixels within the sensor. For frame sensors this is based on image measurements of fiducial marks and the known locations of these marks; for the panoramic images, timing marks are used. Because the film length for a panoramic image was too large for the scanner, the panoramic image was scanned in five segments (a, b, c, d, and e), as described above. Timing marks that desig- nate the center of the image are located on the edges of the film at the midpoint of the middle scan segment (“c”). The midpoint between these two timing marks is used as the origin for image coordinates. The y coordinates for the two timing marks is set to zero and the x coordinates are set to the distance from the origin, based on their pixel coordinates and the scan size of 10 µm. Timing marks at the edge of the scan segment are picked that are in common with the next scan segment. The midpoint between these timing marks is located, and the distance from the origin midpoint and this midpoint is computed, again based on pixel coordinates and the scan size of 10 µm, and is used as the y coordinate for these timing marks. The x coordinate is the distance from the midpoint. For the next panoramic scan segment, the coordinates for the timing marks that are in common between the segments are used as the starting point. Timing marks that are at the middle of the scan at the other edge of the scan are located. The distance between midpoints are used for the y coordinate and the distance between the midpoint and the timing marks are used for the x coordinate. This is repeated for the next panoramic scan segments. The timing marks are located every 2.5° (y direction), with an additional timing mark at the center of the scan (NASA, 1971). We did not find any information on the nominal location of the timing marks in the x direction. Nominal values for positions of fiducial marks on the LO frame images were obtained from LO calibration reports (USAF, 1968). We picked 6 fiducial marks for interior orientation within SOCKET SET and entered the nominal location for each fiducial. The tips of the saw tooth marks were used for Apollo frame images, since the actual fiducial marks can be difficult to locate. These saw tooth marks are known to be located 56 mm from the indicated principal point. The root mean square (RMS) error for interior orientation of the Apollo metric frame images is less than a pixel, the Apollo panoramic images had errors between 1.5 and 2 pixels, and the LO errors were 2 – 4 pixels in size. The errors on the Apollo panoramic and LO images are higher than expected. The Apollo panoramic image timing marks are very large and in some cases are distorted; locating the center of the timing mark can be problematic. The fiducials are sometimes distorted on the LO images, so there would be more error in their measurement.

Bundle Adjustment Bundle adjustment is a least squares solution that ad- justs camera position and pointing along with control point Table 2. Bundle Adjustment Summary locations to minimize misregistrations of the images to one Control Points another and to ground control. Initial estimates of camera GSD RMS X Y Z position and pointing are part of the manually entered sup- Source (m/pixel) (pixel) (m) (m) (m) port data. For ground control, we visually identified Apollo metric 13.7 0.62 356 516 650 points in the Unified Lunar Control Network 2005 (Archi- Apollo Pan 1.9 – 3.3 1.16 126 400 475 nal et al., 2006a, b) and measured their locations on the LO V med 11.6 – 12.4 1.36 321 429 1485 LO and Apollo images. We also selected tie points and LO V med+high 1.6 – 12.4 1.74 303 422 1177 pass points, whose locations are unknown, to tie the LO LO IV 31.5 – 34.7 1.47 379 494 579 and Apollo images and stereopairs together. Bundle ad- ALL 1.6 – 34.7 2.79 342 524 516 justments were run for each set of images alone and for the GSD=ground sample distance, RMS=root mean square residual of complete set of data. Table 2 summarizes the bundle ad- bundle adjustment in image space, X, Y, Z=RMS horizontal and justment results. vertical components of error in locations of ground control The Apollo metric images provide the best results. Relatively high residuals for LO V (reflected in the higher RMS values in the table) may be the result of residual distortions in the reconstructed images, which are visible in the form of mismatches and even gaps between adjacent framelets. The increased error in the solution with all data is mainly due to having more measurements per point, and matching points across images with different resolutions.

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 DEM Extraction Digital Elevation models were extracted based on each source of imagery, except the unpaired LO V high resolution image. Table 3 provides a list of the DEMs, images used, spacing between points in the DEM, and number of points inside the DEM, and figure 2 provides a layout of the DEMs. The DEMs were exported from SOCET SET in Arc Grid ASCII format. This introduced some extraneous data into the DEM that shows up as thin lines that are outside of the DEM boundary but within the minimum bounding box around the DEM. Default automatic matching parameters were used with success. DEMs based on the Apollo metric images were the cleanest looking. As expected, in the shadowed areas the matcher failed since there was no data to correlate. In a few spots, the matcher did not find the bottom of Rima Hadley, and some minor manual edits fixed these errors. DEMs based on the Apollo panoramic scan segment “c” were better than the DEMs from scan segments “d” and “e”. In segment “c” there were a few areas where there was some y parallax in the model that caused some minor distortions in the DEMs. These errors can be seen in shaded relief images of Table 3. DEM List the DEM as areas where the texture of DEM Image 1 Image2 Spacing (m) Points the shaded relief takes on a rougher APOLLO METRIC DEMS appearance. Along the bottom of Rima AM_585_583 585 583 50 5,353,925 Hadley there are a few spikes that re- AM_587_587 585 587 50 5,591,678 APOLLO PANORAMIC DEMS quire editing. For segments “d” and AP_09_14_C 9809 9814 10 10,119,510 “e” there are more such errors. DEMs AP_09_14_D 9809 9814 10 14,765,075 based on LO V medium resolution im- AP_09_14_E 9809 9814 15 12,480,000 ages have artifacts along the lines AP_11_16_C 9811 9816 10 10,474,792 where the framelets were joined to- AP_11_16_D 9811 9816 10 15,224,271 gether. There is also another pattern AP_11_16_E 9811 9816 15 13,049,698 within the framelets that looks like a AP_11_14_C 9811 9814 10 1,610,488 LO IV DEMS rolling pattern. The DEM based on LO LO_4_H1_H3 4103_H1 4102_H3 150 1,612,478 IV images does not have this pattern, LO V DEMS but the shaded relief image of the DEM LO_5_105_106 5105_MED 5103_MED 50 1,624,137 is noisier with some artifacts in bland LO_5_105_107 5105_MED 5107_MED 50 1,813,794 areas.

Figure 2. DEM layout.

ANALYSIS

The overlapping areas of the DEMs from similar image sources (Apollo panoramic, Apollo metric, and LO V) were compared and the results are presented in table 4. The DEMs for the Apollo panoramic stereo models scan

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 segment “c” did not overlap so we created a DEM from the aft looking image from one stereo model and the forward looking image from the other stereo model. DEMs from the Apollo panoramic Table 4. DEM overlap comparision images provide the most consistent DEM 1 DEM 2 points RMS (m) STD (m) BIAS (m) % Blunders results. The panoramic scan segments AP_09_14_c AP_11_14_c 153,805 7.8 7.5 -2.0 0.5 “c” and “d” overlap with the LO V high AP_11_16_c AP_11_14_c 241,368 7.0 6.8 1.8 2.2 resolution images; this resulted in hav- AP_09_14_d AP_11_16_d 423,083 18.2 15.6 9.4 0.1 ing about a dozen tie points in this area. AP_09_14_e AP_11_16_e 1,128,927 20.2 18.0 9.2 1.2 The overlap area between pano- AP_09_14_c AP_09_14_d 52,462 5.2 4.1 -3.2 0.0 ramic scan segments “d” and “e” have AP_09_14_e AP_09_14_d 137,040 26.4 15.6 21.3 0.1 5 tie points where segments from im- AP_11_16_c AP_11_16_d 190,066 14.8 11.4 9.5 1.1 ages 9809 and 9814 overlap, and 3 tie AP_11_16_e AP_11_16_d 144,854 53.7 36.7 39.1 2.0 points where segments from images AM_583_585 AM_587_585 372,276 73.6 49.9 -54.1 12.6 9811 and 9816 overlap. By increasing LO_5_105_106 LO_5_107_106 10,748 101.0 101.0 3.4 4.3 the tie points in the overlap regions the differences between DEMs should decrease. DEMs from the Apollo metric cam- Table 5. DEMs compared to Apollo metric DEM era overlap in an area with very rough DEM 1 DEM 2 points RMS (m) STD (m) BIAS (m) % Blunders topography; an average elevation of - AP_09_14_c AM_merge 3,902,412 32.2 27.3 17.1 12.6 1,000 m, a standard deviation of 1,200 m, AP_09_14_d AM_merge 4,612,145 29.3 28.9 -4.8 6.2 and a range of 6,250 m. There are 2 large AP_11_14_c AM_merge 578,659 22.3 20.0 9.8 6.6 craters and high mountains in this area. AP_11_16_c AM_merge 4,158,651 15.2 15.1 -1.8 7.4 This is probably the cause for the large AP_11_16_d AM_merge 3,821,960 29.7 29.7 0.2 2.2 differences in this region. LO_4_h1_h3 AM_merge 283,196 150.6 147.6 30.2 1.6 The overlap between the LO V LO_5_105_106 AM_merge 458,129 119.5 113.9 -36.2 2.8 DEMs occurs in a very thin region, with LO_5_107_106 AM_merge 555,316 87.8 87.1 10.8 2.6 only 10,748 points. The DEMs from LO V are noisy and contain artifacts. These DEMs have not been edited. We will edit the DEMs and check to see if the differences change. The DEMs from the Apollo metric cameras were merged and edited. All the other DEMs were compared against the Apollo metric merged DEM and results are presented in Table 5. DEMs from Apollo panoramic scan segment “e” did not overlap with the Apollo metric merged DEM in a significant way (only 2,000 points), so they were dropped from the table. The differences between the Apollo panoramic DEMs and the Apollo metric merged DEM are shown in figure 3. The largest differences occur in mountainous regions, shadow areas, and in the join area of the Apollo metric merged, which is in the upper right of figure 3. There is a repetitive pattern that is evident in the flat areas of the DEMs. This appears to be in the direction of flight. The film was scanned along the film direction shown in figure 3. The pattern is most noticeable in the shaded relief images of the Apollo metric DEMs. We are trying to determine if this is a result of the way the film was digitized or if it occurred in the process of making the 4th generation positive that was scanned. The differences between the LO V DEMs and the Apollo metric merged DEM are shown in figure 4. There Figure 3. Difference between Apollo panoramic is a pattern that outlines the edges of the framelets that DEMs and Apollo metric merged DEM (meters). The make up a LO image. Another pattern appears within the grey areas show where there are shadows on the framelets and appears to be an undulating pattern. This is Apollo metric image. in the direction that the LO framelets were scanned. The

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 largest errors are on the right side of the figure and are in a mountainous region. The left side of the figure is flatter terrain and contains fewer errors.

Figure 4. Difference between LO V DEMs and Apollo metric merged DEM. The grey areas show where there are shadows on the Apollo metric image.

The differences between the LO IV DEM and the Apollo metric merged DEM are shown in figure 5. The pattern outlining the framelets that was visible with LO V is not evident with LO IV. There is a horizontal pattern that looks like there might be a bow in the LO IV DEM, with the edges being higher and the lower part about a third of the way from the top. The seam in the Apollo metric merged data is slightly evident in the top center part of the differences, but is not shown in the bottom part. This might be due to the noise in the Lunar Obiter IV DEM overpowering errors in the Apollo metric merged data. On the left side of the figure, there is a large area with a substantial error. This is due to the blandness of the LO images in this flat area. This can be edited out, but it would be difficult to recover any meaningful data from the LO IV images in this area. The LO IV images covered the largest area and there was little horizontal or vertical control at the edges of the images. Adding additional control at the edges would be help- ful in reducing the bowing of the LO IV DEM. Figure 5. Difference between LO IV DEM and To analyze the absolute accuracy of the DEMs we com- Apollo metric merged DEM. The grey areas show pared the elevation values from the DEMs to Clementine where there are shadows on the Apollo metric image. Lidar points and the XYZ control points used in the bundle adjustment. Figure 6 provides a layout showing the DEMs, the Clementine Lidar points, and points used in the bundle adjustment. All elevations were transformed to refer to a vertical reference surface of radius 1,737,400 m.

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 The Clementine Lidar points were divided into 4 groups based on their longitude values; longitude 1, longitude 3.8, longitude 6, and longitude 9. For each Lidar point the elevations from the DEMs were extracted. The points were sorted based on their latitude values and the elevation profiles were plotted along those lines. In figure 7 the LO IV DEM is the only DEM that covered the points at longitude 9°. This is at the edge of the model and the agreement between the LO IV DEM and elevation values from the Lidar points is much worse than the 100 m accuracy of the Lidar points. There is little control in this area and just a few pass points that tie the LO IV images together. In figure 8 both LO IV and Apollo metric DEMs cover the Lidar points at longitude 6°. There is strong agreement between elevations from the DEMs and the Lidar points. At 25.75° latitude there is one point that is not in agreement. This is at a high point and an area with high slopes. This mismatch is probably caused by poor registra- tion between the DEM and Lidar data, given the horizontal uncertainties of the data and the high slopes. In figure 9 DEMs from seven different models cover the Lidar points at longitude 3.9°; LO IV, both LO V, Apollo metric merged, Apollo panoramic images 9808 and 9814 scan segments “d” and “e,” and images 9811 and 9816 scan segment “e.” For the most part there is strong agreement between the elevations from the DEMs and the Lidar points. Two exceptions would be the LO IV point at 26.75° latitude and the points from the Apollo panoramic images 9811 and 9816 scan segment “e” at the end of the profile. The LO IV point missed the edge of a mountainous feature. With the points at the end of the profile, the DEM tends to decrease in elevation at a faster rate than the Lidar points. The model did not hold to the Z control very well in this area. (See figure 11 and discus- sion later in this paper.) Also, the panoramic sensor model might not have enough Z control in this area, and the panoramic sensor model in the bundle adjustment might be curving down. In figure 10 the LO IV DEM is the only DEM to cover the Lidar points. The topography in this area is flat and there is agreement between the elevations from the LO IV DEM and the Lidar points. There is one area where the LO IV DEM did not correlate well and this error would be removed in the editing of the DEM. The images in this area are very bland and the correlation curve between the images would be very flat. In figure 11 we compare elevations from the DEMs with elevations used for Z control in the bundle adjustment. Z control data came from an interim solution of the ULCN 2005. Points were visually identified in images used in the control network solution and visually transferred to the imagery used in this study. The estimated Z accuracy in the control network is about 1 km. This was used as the weight on the Z control used in the bundle adjustment for this study. The input Z control points are elevation values from ULCN 2005. The output Z control elevation values are the estimated values from the bundle adjustment. The longitude of the control points spanned 1.7° to 7.0°. All these points were collapsed into a single elevation profile. The points were sorted by latitude values and elevations were extracted from the DEMs. There are seven DEMs that cover the control points; LO IV, LO V images 107- 106, Apollo metric merged, Apollo panoramic images 9808 and 9814 scan segments “d” and “e”, and images 9811 and 9816 scan segments “d” and “e.” The differences between the elevations for input Z control and output Z control are within the limits of the 1 km weight placed upon these points. There is strong agreement between the elevations from the output Z control and the DEMs. For last three points in the profile, the elevations for the output Z control and DEMs are lower than the elevations from the input Z control. This is in the same area as in figure 9 where the elevations from the DEMs are lower than the elevations from the Clementine Lidar points. This is at the edge of the panoramic images and additional control is probably needed in the area, which could be accomplished by using the Clementine Lidar points for Z control in the bundle adjustment.

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006

Figure 6. DEMs shown with the same color lookup table and at 50% transparency. Points from the bundle adjustment are shown with pass points being points on 2 images, tie points on 3 or more images, XY control, and XYZ control points (from ULCN 2005). The Clementine Lidar points are plotted, and groupings used for comparison with the Lidar points (Longitude 1, 3.8, 6, and 9) are shown at the bottom of the figure.

Figure 7. Elevation profile for Clementine Lidar points and DEM along longitude 9. Latitude is north, longitude is east, and elevation is measured relative to a sphere of radius 1,737.4 km.

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006

Figure 8. Elevation profile for Clementine Lidar points and DEM along longitude 6. Latitude is north, longitude is east, and elevation is measured relative to a sphere of radius 1,737.4 km.

Figure 9. Elevation profile for Clementine Lidar points and DEM along longitude 3.8. Latitude is north, longitude is east, and elevation is measured relative to a sphere of radius 1,737.4 km.

Figure 10. Elevation profile for Clementine Lidar points and DEM along longitude 1. Latitude is north, longitude is east, and elevation is measured relative to a sphere of radius 1,737.4 km.

Figure 11. Elevation profile for Z control points and DEMs. Longitude spans 1.7° - 7.0°. Latitude is north, longitude is east, and elevation is measured relative to a sphere of radius 1,737.4 km.

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 OUTPUT PRODUCTS

Based on the DEM orthorectified images, controlled mosaics and contours can be produced. These could be provided by themselves or as part of a finished map with grid and collar data. These products can be exported to Integrated Software for Imagers and Spectrometers (ISIS) (Eliason, 1997) and in formats usable by GIS and visuali- zation packages.

CONCLUSIONS

DEMs from Apollo metric and panoramic imagery did not have very many errors or artifacts. The Apollo metric DEMs are easy to edit and result in a very clean DEM. The Apollo panoramic DEMs are slightly more difficult to edit because of their large size (imagery and DEM). For this study, most of the imagery was located on the right side over Apollo metric image 0587. This resulted in most of the control being in this area. Adding additional control over Apollo metric image 0583 and in the areas were the Apollo panoramic scan segments overlap would reduce some of the differences that were seen in the elevation values. Also, for the panoramic images it would be useful to test if adding more tie and pass points in the side lap between images would reduce the errors and if that would remove y-parallax in the panoramic models. It would be useful to have more overlap between scan segments for the panoramic images. This would benefit the bundle adjustments and (in the case of increased scanning overlap) provide for more overlap between DEMs. This can be accomplished by having scan segment “c” centered on the center timing mark and having two timing marks in common between each scan segment instead of one timing mark. DEMs from the LO imagery are useful, but have more noise and artifacts that will require additional editing. As part of the bundle adjustment procedure there are options to adjust the interior orientation parameters. We have never used this option before, so this might be worthwhile to test because the interior orientation of LO images is relatively poor. Additional control on the edges of the images would reduce elevation differences in the DEMs and with Clementine Lidar points. Overall, softcopy stereomapping techniques can readily be applied to scanned lunar images to produce controlled DEMs, orthoimage mosaics, and other products that will be useful in future mission planning and scientific analysis. These products could also be used in turn to densify and improve the current global network. The full value of the legacy datasets from LO and Apollo has yet to be exploited.

ACKNOWLEDGMENT

The work reported here was supported by NASA Planetary Geology and Geophysics Program, contract number NNH04AA87I.

REFERENCES

Acton, C. H. (1999). SPICE Products Available to the Planetary Science Community. 30th Annual Lunar and Plane- tary Science Conference, March 15-29, 1999, Houston, TX, abstract no. 1233. Archinal, B. A., M. R. Rosiek, R. L. Kirk, and B. L. Redding (2006a). Completion of the Unified Lunar Control Network 2005 and Topographic Model. 37th Annual Lunar and Planetary Science Conference, March 13- 17, 2006, in League City, Texas, abstract no.2310. http://www.lpi.usra.edu/meetings/lpsc2006/pdf/2310.pdf Archinal, B. A, M. R. Rosiek, R. L. Kirk, and B. L. Redding (2006b). The Unified Lunar Control Network 2005, Including a New Lunar Topographic Model, in preparation. Cook, A. C., T. R. Watters, M. S. Robinson, P. D. Spudis, and D. B. J. Bussey (2000). Lunar polar topography de- rived from Clementine stereo imagery. Journal of Geophysical Research, Vol. 105, No. E5, pp. 12,023- 12,033, May 25.

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006 Cook, A. C., M. S. Robinson, B. Semenov, and T. R. Watters (2002). Preliminary Analysis of the Absolute Carto- graphic Accuracy of the Clementine UVVIS Mosaics, American Geophysical Union, Fall Meeting 2002, Abstract no. P22D-09. Davies, M. E., T. R. Colvin, D. L. Meyer, and S. Nelson (1994). The unified lunar control network: 1994 version. Journal of Geophysical Research, Vol... 99, No. E11, pp 23,211-23,214. November 25. Davies, M. E., and T. R. Colvin (2000). Lunar coordinates in the regions of the Apollo Landers. Journal of Geo- physical Research, Vol. 105, No. E8, pp. 20,277-20,280. Eliason, E. M. (1997). Production of Digital Image Models, Lunar Planet Sci., XXVIII: Lunar and Planetary Insti- tute, Houston, Abstract, pp. 331-332. Gaddis, L., T. Becker, L. Weller, D. Cook, J. Richie, A. Bennett, B. Redding, and J. Shinaman (2003). Reviving Lunar Orbiter: Scanning, Archiving, and Cartographic Processing at USGS. 34th Annual Lunar and Plane- tary Science Conference, March, 2003, in League City, Texas, Abstract no. 1459. Hansen, T. P. (1970). Guide to Lunar Orbiter Photographs. Langley Research Center, Hampton, VA-USA. NASA SP-242. Malin, M. and M. Ravine (1998). Clementine High Resolution Camera Mosaicking Project, Technical Report, Ma- lin Space Science Systems San Diego. NASA (1971). Apollo 15 Sim Bay Photographic Equipment and Mission Summary, Mapping Sciences Branch, Houston Texas. Neumann, G. A. (2001). Some Aspects of Processing Extraterrestrial Lidar Data: Clementine, Near, MOLA. Inter- national Archives of Photogrammetry and Remote Sensing, v. XXXIV-3W/4, Annapolis, MD, 22-24 Oct. Nozette, Stewart, et al. (1994). The Clementine Mission to the Moon: Scientific Overview. Science Vol. 266 pp. 1835 – 1839. Rosiek, M. R., R. Kirk, and E. Howington-Kraus (2002). Color-coded Topography and Shaded Relief Maps of the Lunar Hemispheres, 33rd Annual Lunar and Planetary Science Conference, March 11-15, 2002, Houston, Texas, abstract no. 1792. Smith, D. E., M. T. Zuber, G. A. Neumann, and F. G. Lemoine (1997). Topography of the Moon from the Clementine Lidar. Journal of Geophysical Research, Vol. 102, no. E1, pp. 1591-1611. USAF, United States Air Force (1968). Lunar Orbiters IV and V Camera Calibration Report, USAF/ACIC, St. Louis, Missouri. USGS (1997). Clementine Basemap Mosaic, USA_NASA_PDS_CL_30xx, NASA Planetary Data System. Weller, L., B., Redding, T. Becker, L. Gaddis, R. Sucharski, D. Soltesz, D. Cook, B. Archinal, A. Bennett, and T. McDaniel (2006). Lunar Orbiter Revived: Very High Resolution Views of the Moon. 37th Annual Lunar and Planetary Science Conference, March 13-17, 2006, in League City, Texas, abstract no.2143. http://www.lpi.usra.edu/meetings/lpsc2006/pdf/2143.pdf Wu, S. S. C. and F. J. Doyle (1990). Planetary Mapping, R. Greeley and R. Batson. Ed., Cambridge University Press, pp. 169-207.

ASPRS 2006 Annual Conference Reno, Nevada May 1-5, 2006