The Landsat processing system used by the EROS Data Center is the National Landsat Archive Production System (NLAPS). This system has replaced the EROS Digital Image Processing System (EDIPS).
A Model 107 Digital Cassette Recording System incremental (DCRSi) cassette drive is used to supply serial image data to the NLAPS. The DCRSi is a rack-mountable unit capable of recording up to 48 gigabytes on a cassette at any speed from zero to 107 megabytes per second. The built-in-Error-Correcting Code (ECC) generator and decoder provides a user Bit Error Rate (BER) of less than 1 in 1,000,000,000 bits.
The Modular Multi-Satellite Preprocessor (MMSP) receives the serial MSS-A (radiometrically corrected) or MSS-P (radiometrically and geometrically corrected) data from the DCRSi cassette drive. The serial data are routed to the Programmable Sync Module-2. The Sync Module searches for major and minor frame sync words, converts the serial data to parallel form, and passes the data over a 10-Byte per second dedicated bus (the High-Resolution Imagery bus) to the Demultiplexer/Subsampler (DMS/SS). The DMX/SS demultiplexes the data and writes to the memory on the MMSP CPU via the VME bus. When subsampling, the DMX/SS performs the data reduction function.
Data resampling begins after the completion of satellite orbit model generation/refinement and radiometric calibration. Data resampling consists of three distinct components; coarse resampling table generation, resampling table densification, and two-dimensional convolution resampling. Coarse table generation consists of calculating relationships between product line/pixel values at a set of grid-points and the corresponding line/pixel values in the raw data. Table densification consists of interpolating similar relationships between the remaining product line/pixel values and the corresponding raw data line/pixel values. Table densification also includes performing a number of corrections to the interpolation to account for various inaccuracies in the raw data. The convolution step involves using the fine table to convolve the output pixel values from the raw data.
Radiometric calibration is the process of converting raw digital numbers (DNs) observed by a sensor into physical units. The radiometric calibration of Landsat MSS data is performed in two steps:
- Absolute Calibration: the recovery of radiance (as observed by the sensor) from the raw digital number recorded. This involves modelling the characteristics of the optics and electronics of the sensor. - Relative Calibration: the removal of residual errors in the calibrated data to improve the qualitative appearance of the imagery.Geometric correction removes geometric distortions in an image based on knowledge of the satellite and sensor, and remaps the image to a regular grid in a standard map projection. This is accomplished by constructing a mapping between pixel coordinates in the image and geographic coordinates on the surface of the Earth. This mapping is referred to as the forward transformation.
The raw image contains geometric distortions induced by characteristics of the sensor. These characteristics include:
- non-linear mirror scanning velocity; - varying average mirror speed between scans; - sequential detector sampling (rather than "snapshot"); - detector offsets in the focal plane.The resulting distortions are modelled and corrected during ground processing.
Products are geometrically corrected to one of five levels;
- raw (no geometric corrections applied); - systematically geocorrected; - precision geocorrected; - precision registered; - terrain corrected. Corrections Applied for Each Geometric Correction Level Systematic Precision Precision Terrain Geocorrected Geocorrected Registered Corrected Sensor Offset x x x x Line Length Variation x x x x Scan Non-linearity x x x x Sensor Misalignment x x x x Attitude Variations x x x x Orbit Variations x x x x Panoramic Distortion x x x x Earth Curvature x x x x Earth Rotation x x x x Refine Model with GCPs x x Refine Model with TCPs x x Apply Terrain Model x A precision product uses a precision spacecraft model generated by refining the systematic model described above with data based on ground truth. For a precision product, a third step is added to the process of deriving the spacecraft model: - The systematic orbit and attitude models are refined using ground truth (Ground Control Point location) and a filter to produce a precision spacecraft model.
A precision-registered product is similar to the precision-corrected product described above, except that the precision spacecraft model is generated using image-toimage control points rather than ground control points.
A terrain-corrected product is similar to precision-corrected products and precision-registered products (the same spacecraft model is used), but during the geocorrection process, a Digital Elevation Model (DEM) is used to correct for terrain-induced parallax.
During geometric correction, the input image is resampled to a regular output grid. The resampling kernel specifies how the input pixels are sampled; how many, and how they are weighted. NLAPS employs both one-pass and multi-pass resampling with operator-selectable kernels of varying lengths including:
- Nearest-neighbor; - Bi-linear; - 4-point Cubic Convolution; - 8-point damped sinc; - 16-point sinc; - 16-point Kaiser damped sinc. The characteristics of the MSS sensor are modelled by the coordinate transformation from raw image (line, pixel) coordinates to sensor (scan angle, deployment angle) coordinates. The necessary corrections can be divided into along-scan and across-scan corrections. Nominal MSS Geometric Parameters LS 1,2,3 LS 4,5 Altitude at Equator 920 km 705.3km Scan width 185 km 185 km IFOV 79 m 83 m Along-scan pixel spacing 56 m [10] 57 m [5] [IFOV minus overlap) Average mirror across-track 5.61 m/us 5.61 m/us ground speed Mirror across-track motion 2.24 m 2.24 m between samples Nadir velocity 6.456 km/s 6.76 km/s Active scan time 32.2 ms 32.2 ms Nominal line length 3210, 3240, 3192 3240, 3233 pixels Product Framing The product framing defines the boundaries of an output product. NLAPS provides four methods of framing products: - WRS scene; - Swath; - Latitude defined portion of swath; - Latitude/longitude or center/span portion of scene or swath.The WRS identifies scenes by path and row number. A single WRS scene is nominally 184 km by 172 km. The path number identifies the nominal ground tracks over which the spacecraft relies. The row number identifies the center of a scene along any one path. On a particular path, the center point for the reference row is taken at the equator; this row is designated 60. The row immediately north of this is designated 59. Its scene center will be at a distance along the path equivalent to roughly 25 seconds of spacecraft flight time.
A swath of data is all data received from a spacecraft on a single pass from Acquisition of Signal (AOS) to Loss of Signal (LOS). The swath is specified by its WRS path number.
Latitude framing specifies a portion of a swath by the upper and lower bounding latitudes. The latitude refers to the mid-point (nadir) of the swath.
At latitudes greater than 80 degrees, ambiguities can arise in the latitude specifications due to the inclination of the orbit plane. For this reason, latitude specifications may not be used at latitudes greater than 80 degrees; products requiring such specifications must use the alternative method described below.
Certain products may be framed by specifying a pair of opposite corners in latitude and longitude or map coordinates. These corners define a rectangle that frame the product in the output imagery. This frame is constructed in the output grid defined in the product specification. The output grid can be specified as any orientation (rotation) of any of the supported map projections. Note that for a particular pair of latitude/longitude corners, changing the map projection or the orientation can drastically change the product size and shape.
A latitude and longitude product may alternatively be specified using a center/span framing. Center span products are specified by the latitude and longitude (or map coordinates) of the product center, and the span (the distance from edge to edge) in the specified output grid. The maximum spans allowed are 300 km and 500 km. This method can simplify the specification of a product, since the scene center and desired span might be more readily available than the coordinates of two corners.
NLAPS-Supported Map Projections (all GCTP map projections) (For Datums NAD27, NAD83, WGS84) Universal Transverse Mecator State Plane Coordinate Systems Albers Conical Equal-Area Lambert Conformal Conic Mercator Polar Stereographic Polyconic Equidistant Conic Transverse Mercator Stereographic Lambert Azimuthal Equal-Area Azimuthal Equidistant Gnomonic Orthographic General Vertical Near-Side Perspective Sinusoidal Equirectangular Miller Cylindrical Can der Griten Oblique Meracator Robinson Space Oblique Mercator Modified Sterographic Conformal (Alaska)The product orientation controls the angle between the map projection grid and the output image grid. NLAPS supports three standard orientations (Satellite Orientation, True North UP, Map North Up), as well as any user defined orientation.
The pixel spacing of a product is operator selectable and spans the range of 25 meters to 100 meters in increments of .25 meters. Pixel spacing is not constrained to be square.
Several factors affect the product size:
- the product framing; - the pixel size; - the orientation; - the map projection.