A new ODP DIS, which was custom-fabricated by Geotek Ltd., was installed in November 2001, prior to Leg 199. Following tests on Legs 199-201, the system was upgraded with new software in March 2002 during the port call in Valparaiso, Chile. Leg 202 was the first cruise to use the DIS routinely, and ~7 km of core was analyzed, creating ~15 GB of image data. Procedures for the reduction and use of image data were adopted and developed based on these upgrades and as needed based on a variety of tests made during Leg 202. Some problems were discovered as a result of these activities, yielding some recommendations to aid in planning and implementation of future upgrades to the DIS.
In the ODP DIS system, as many as four core sections are loaded onto stationary rails, and the camera system moves along each section in an automated sequence. The ODP DIS uses a proprietary Geoscan line-scan camera assembled by Geotek. The camera includes three 1024-pixel charge coupled device (CCD) arrays, one for each color, red, green, and blue (RGB). The camera is located above the light source and views the core through a slot in the top of the light housing. Light reflected from a core surface passes through a standard Pentax lens, and is split into three paths by a beam splitter. The three beams then pass through dichroic interference filters designed to pass blue (<515 nm), green (515-585 nm), and red (>585 nm), respectively, before falling on the CCD detectors located at the focal plane of the lens. The blue channel receives 50% of the total light collected (because the CCD detector is less efficient at short wavelengths), whereas the green and red channels each receive 25% of the light collected. Each CCD array has 8-bit precision (i.e., values from 1 to 256), so when the three color channels are combined the total color precision is 24-bit.
Two high-frequency fluorescent tubes oriented across the cores on either side of the image line illuminate the sediment cores. This lighting system is intended to flood the core with uniform light and to minimize shadows caused by microtopography on the core surface. The bright light source was generally assumed to overpower any ambient laboratory light. However, preliminary evidence during Leg 202 indicated that ambient laboratory light, which varies with position in the DIS, contaminated some images. The shipboard party therefore covered the porthole in the sediment laboratory and constructed a temporary cardboard cover for the DIS to shade the instrument from ambient light. Water on the surface of the split cores can cause specular reflections, which degrade the image by reflecting either ambient light or light from the fluorescent tubes directly to the camera.
The spectral response of the system is a function of the spectrum of the light source, which can evolve through time, the efficiency of the CCD arrays at different wavelengths, and the various optical components. Under typical conditions, the greatest efficiency of response for the Geotek system (i.e., field of >50% of peak efficiency) is ~450-505 nm for the blue channel, ~530-555 nm for the green channel, and ~605-630 nm for the red channel (Fig. F10). Spectral biases between the RGB bands are calibrated using external standards for dark current and white balance.
The housing for the light source is adjusted vertically to be as close as possible to the core surface (nominally 1 cm) so that the lighting will be as uniform and as bright as possible on the sediment surface, overpowering any ambient room light that varies depending on location in the system (e.g., from the top to the bottom of core sections) or through time (e.g., if a person walking by inadvertently shades the system from ambient light). The height of the camera assembly can be adjusted above the light housing so that the CCD array scans a width of ~9 cm and a split core section approximately fills the CCD array. The CCD array includes three reference pixels on the left side of each core section that are coated in an opaque (black) substance. These pixels are used for software adjustment of the time-variable zero value on all pixels (known as the dark current) due to thermal or electronic drift. Ideally, these calibration pixels are not affected by current in adjacent pixels, however, CCD arrays can be sensitive to pixel "bleed" effects if nearby pixels exceed saturation.
The spatial resolution of the DIS system depends on the height of the camera above the sediment, but it is nominally 100 pixels per centimeter in both the downcore and cross-core directions. Rulers are placed along the sediment cores and imaged to confirm depth calibrations.
After the geometry of the camera-core system was set, the camera was manually focused by ODP personnel (typically once per day) to maximize contrast on a sharp transition (black lines on color calibration card or sharp edge of the core liner). Given adequate depth of field (depending on aperture) the focus does not need to be reset for minor variations in the core surface. Aperture is set to maximize the dynamic range of the system for normal color variability in the cores to be measured, within the limits of needing reasonable depth of field to stay in focus. To set aperture, an RGB "scope" is used to display all 1024 pixel intensities over a relatively bright interval of core, and aperture is adjusted until one or more colors approaches but does not exceed color saturation (limited by 8-bit precision to values of 256 in each CCD). The aperture setting was entered manually into the program to allow for intensity normalization (see below). For each core section batch run, the operator monitored the intensities; if they decreased, the aperture was manually increased, and vice versa, and the new aperture setting was entered accordingly.
Color calibration of each pixel occurs at the black (dark current) level by collecting data with the lens cap on and at the white (nominally 100% reflectance) level using a white ceramic tile. The camera travels a short distance across the tile to remove the effects of any dust, which could affect individual pixels. White calibration is a function of both aperture and integration time. Calibration need not be done at the same aperture setting as data collection as long as the aperture settings are entered accurately in the software. These settings are used to adjust pixel intensities based on the theoretical relation I = 1/f2, and as a result, raw pixel intensity values can exceed the number 256. These aperture calibrations could be adjusted based on empirical analysis of grayscale standards at different aperture settings, but such features were not implemented during Leg 202. Black-and-white calibration should be done as often as possible; ideally, between each group of core sections. In practice, manual calibrations were done roughly every 12 to 24 hr, usually associated with shift changes.
The constraints of 8-bit color resolution and the typical range of sediment color variations require frequent aperture adjustments. CCD saturation levels have to be monitored constantly by the operator, aperture adjustments have to be done manually, and the new setting has to be entered manually as well. Repeat scans were often necessary to obtain images with the appropriate aperture setting after inadequate images were discovered. More automation, such as automated aperture or exposure time adjustments (based on quick lightness scans over the cores) and/or oversaturation and low-intensity alerts, would enhance the operational efficiency of the system.
Four sections at a time were placed in the DIS with the tops of sections against a rigid stop defined as zero depth in section, and a barcode label was scanned for entry of core identification information. All four sections were then scanned automatically by the system in ~20 min. Scanning generally occurred in parallel with description of other sections of the same core, and the digital images were therefore generated immediately before or after visual description and prior to analysis of full-color reflectance spectra on the AMST and intensity and direction of magnetic remanence with the cryogenic magnetometer.
When a core is cut into sections on the catwalk, ODP personnel enters the "liner length" (usually 150 cm, except for the last section, which is typically shorter) for each section into the database. Unfortunately, data processing is sometimes complicated by the fact that a core section length does not stay the same: cores keep expanding during the core analysis process as a result of continuing gas escape, clay swelling, and/or elastic rebound, depending on lithology, hydrocarbon content, and drilling depth. The expanding core pushes the liner end caps out in a bulge, breaks the end cap's acetone bond, or even extrudes above and beyond the end cap. By the time the Leg 202 cores were scanned, sections were typically 1-3 cm longer than the liner length. Since the objective was to scan the entire core, the DIS operator had to measure the "curated length" (at the time of scanning) and enter it as a measurement parameter into the instrument's program. ODP personnel subsequently entered the curated section lengths into the database.
Figure F11 illustrates the DIS image products that were created during Leg 202. The DIS saves image data in TIFF formats on a local computer. During acquisition, red, green, and blue pixel intensities can be averaged across user-defined intervals and saved as ASCII text RGB files. Because average values are saved as integers, some precision is lost in this operation. After some experimentation during Leg 202, we saved RGB summary data averaged over 0.1-mm depth increments, using 400 pixels in a stripe across the center of each core section, and then averaged these separately (using software written at sea) to intervals of 1 or 2 cm without truncating the resulting averages to integer values. Original TIFF and RGB files were archived at ODP.
The original TIFF images were compressed without loss of information using MrSID software, and these images were archived and were available online. From the MrSID files, the yeoperson created composite JPEG images of each core, which were saved as PDF images and distributed to the scientific party. These images substantially reduced the spatial resolution relative to the original TIFF images and were of an easily manageable size that proved to be very useful at sea for informal "second look" analysis and confirmation of core descriptions. Following Leg 202, the production of such "low-resolution" core summary images became an integral and direct output of the DIS system.
The first sets of Leg 202 digital core images revealed light-colored stripes across the image (3 pixels downcore) spaced very regularly at 5 mm or, in some cases, at 1 mm. The lines appeared to be an "overexposure" related to some kind of energy flux across the CCD array. These artifacts were caused by high contrast in the millimeter markings on the rulers placed along the edge of the cores. Oversaturation of image pixels adjacent to the opaque calibration pixels in the CCD array resulted in pixel "bleed" effects into the calibration pixels. The DIS interpreted this pixel bleed as thermal drift, and adjusted the image calibration accordingly, resulting in apparent overexposure. This problem was apparent when relatively low (open) aperture settings were used to image dark sediments.
Additional bright lines occurred in many images at 30-cm intervals. This artifact appears related to small variations in image integration time, as the DIS software transfers data from the camera to the computer every 30 cm and appears to stop the motor for a split second during the process. This software bug must be corrected by the manufacturer.