All Products
Search
Document Center

PolarDB:Basic concepts

Last Updated:Mar 28, 2026

Raster SQL in PolarDB for Oracle stores and processes raster data as structured objects in the database. The following terms describe the core data model.

TermDescription
raster objectA grid-based spatial data structure where space is divided into regular cells. Each cell holds an attribute value representing a real-world entity—such as elevation, temperature, or land-use category. Common raster objects include satellite images, digital elevation models (DEMs), and photographs.
cell/pixelThe smallest unit in a raster object, also called a pixel. Each cell occupies a single position in the grid. Supported data types include Byte, Short, Int, and Double.
bandA single matrix of cell values within a raster object. A raster object can have multiple bands.
chunkA tile used for internal storage of raster data. Chunk dimensions are customizable; for example, 256 x 256 x 3. Tuning chunk size affects read performance for spatial queries.
pyramidA set of progressively lower-resolution copies of a raster object, used to improve query and display performance. A pyramid can contain multiple downsampled layers: each consecutive layer is downsampled at a 2:1 scale, and Layer 0 stores the raw data.
pyramid levelA single layer within a pyramid. Layer 0 is the original full-resolution data; each higher level is a 2:1 downsampled version of the layer below.
mosaicAn operation that merges multiple raster objects into an existing raster dataset. Use mosaic to combine raster tiles or scenes that cover adjacent or overlapping areas—for example, joining multiple satellite scenes into a single continuous coverage.
interleavingThe layout used to store pixel values across bands. The three supported modes differ in how band data is organized and affect which access patterns are most efficient: BSQ (band sequential) stores each band as a complete unit—efficient for single-band analysis; BIP (band interleaved by pixel) stores all band values for each pixel together—efficient for multi-band per-pixel analysis; BIL (band interleaved by line) stores all band values for each row together—a balance between BSQ and BIP.
world spaceThe geographic coordinate space of a raster object. Coordinates in world space correspond to real-world geographic positions, using the raster's projection and spatial reference.
raster spaceThe pixel coordinate space of a raster object. The upper-left corner of the raster is used as the starting point.
metadataDescriptive information stored with a raster object, including its spatial range, projection type, and pixel type. Remote sensing platform metadata is not included.