A raster is a grid that is formed by a matrix of cells (or pixels) that are organized into rows and columns. Each cell contains attribute values that represent the information for the area within the cell. These values are also known as raster data.

Raster data can be classified into two types: thematic data and image data.
  • Thematic data: The value of each cell represents a measurement or a classification to describe the information, such as the pollutant concentration, rainfall, land ownership types, or vegetation types.
  • Image data: Image data is also called remote sensing image. It refers to a film or image taken by using ground remote sensing, aerial remote sensing, or aerospace remote sensing technologies to record the electromagnetic wave size of various ground objects. Image data includes both aerial images and remote sensing satellite images.

A raster is also called a spatio-temporal raster because raster data contains both spatial and temporal attributes. In terms of time, spatio-temporal raster data can also be used to manage time series.

DLA Ganos Raster

DLA Ganos Raster is a spatio-temporal data engine and toolset. It is used to manage and process raster data. DLA Ganos Raster allows you to use ApsaraDB for HBase Enhanced Edition (Lindorm) to store, index, query, analyze, and transmit raster data and related metadata. Raster data is stored in ApsaraDB for HBase Enhanced Edition (Lindorm) as tiles or blocks. A primary key is assigned to each tile or block. You can perform spatial and temporal queries based on the primary key. DLA Ganos Raster also allows you to integrate and analyze raster data from multiple sources, such as remote sensing, photogrammetry, and thematic maps. It also supports data service release features such as the Tile Map Service (TMS) and Web Map Tile Service (WMTS). DLA Ganos Raster can be used in fields such as location-based services, geographic image archiving, environmental monitoring and assessment, geological engineering and exploration, natural resource management, national defense, emergency response, telecommunications, transportation, urban planning, and national security.

Data model of DLA Ganos Raster


The data model of DLA Ganos Raster consists of the following elements:
  • Image: specifies a remote sensing image, for example, a Tagged Image File Format (TIFF) file.
  • Catalog: specifies a data catalog, which is similar to a database. A catalog is a logical concept. It consists of all layers and a metadata table in ApsaraDB for HBase Enhanced Edition (Lindorm). Each layer is stored in a table. The metadata of each layer is stored in a row of the metadata table.
  • Cover or coverage: specifies a dataset that consists of multiple rasters, which is the same as a mosaic dataset.
  • Layer: specifies a 2D raster data layer that consists of multiple tiles. Each tile has a row number and a column number.
  • Tile or block: specifies a tile or block that is a collection of pixels. A tile is the basic unit that is used to store raster data in a database. Each tile contains several cells. Each tile can be 256 × 256 pixels or 512 × 512 pixels.
  • Cell or pixel: specifies a pixel in the tile. It supports various data types, such as BYTE, SHORT, INT, and DOUBLE.
  • Key: specifies a key value, which uniquely identifies a tile. Valid values: SpatialKey, SpaceTimeKey, and TimeKey.
  • Pyramid: specifies a raster pyramid that is used to speed up the display of raster data. Each pyramid has different levels of raster datasets. Each level corresponds to a layer. Level 0 refers to the raw raster dataset.
  • Metadata: specifies the metadata of a raster, such as the spatial range, projection types, and pixel types. The metadata of the remote sensing platform is excluded.
  • Layout definition or layout: Defines chunking mode for tiles in a layer, the geographical range represented by each pixel, and the mapping relationship between a key and a coordinate system.
  • Layout scheme: A layout scheme consists of zoom numbers of all layers in a pyramid and its layout definitions.

Band and layer

DLA Ganos Raster uses a simple and efficient raster data model. This data model is used to manage the thematic data and remote sensing image data. An image consists of several bands that can be represented as a 2D raster layer. Each pixel of a band is represented as a cell. DLA Ganos Raster requires that all cells in each band be of the same data type and have the same projection parameters. Different bands of an image can be heterogeneous. This makes it easy for you to store and manage data. Each image has its own metadata. The metadata includes the extent, data types, projection information, row numbers, and column numbers. Raster data is stored as layers in databases. Each layer is stored and managed as a tile in DLA Ganos Raster. Tiles can be classified into two types: single-band tiles and multi-band tiles. Each multi-band tile contains multiple tiles.

The preceding figure shows the three types of relationships between the band and layer.
  • One band corresponds to one layer: For single-band raster data such as the output of the model and remote sensing image analytics results, each pixel includes only one value. If a pyramid model is not built, each band corresponds to a layer.
  • Multiple bands form a layer: If a layer of the remote sensing image is composed of RGB, the layer can be represented by a multi-band tile. In this case, the R, G, and B bands form a layer.
  • One band includes multiple layers: If you have created a pyramid model for the raster data, each band contains multiple magnification levels. Each magnification level corresponds to a layer.

Pyramid model

You can build pyramids for raster data to improve data access. A pyramid is a downsampled version of the original raster dataset. Each pyramid can contain successive downsampled layers. Each layer of the pyramid is downsampled at a scale of 2:1.

Pyramids speed up the display of the raster data by retrieving only the data at a specific resolution. The resolution varies with the display requirement. If you draw an entire dataset, you can use a pyramid to display tiles or blocks of lower resolutions. As you zoom in, levels with higher resolutions are drawn. However, the performance remains unchanged. The database automatically chooses the best-suited pyramid level for you based on your display scale. After you build pyramids for each raster dataset at a time, these pyramids are accessed each time you view the raster dataset. It requires more time to build a set of pyramids for a larger raster dataset. This also saves you more time in the future.

Layout scheme

The layout definition and layout scheme are used to define the data chunking mode of each layer. If you specify the extent and cell size (the actual spatial range defined by a cell), the layout scheme can provide the zoom levels and the layout definition of each level. DLA Ganos Raster supports two modes for data chunking: zoom and local. In zoom mode, data within the globe is chunked and encoded in compliance with the TMS standard. The tile in the upper-left corner is defined as the starting point (0,0). The coordinates increase from left to right and top to bottom. All types of raster data are chunked into tiles based on spatio-temporal grids. This makes it easy for the accumulable analytics and multivariate data fusion. In compliance with the TMS standard, tiles chunked by this mode can be rendered and then published to TMS for display. For example, you can use the OpenLayers component after publishing. However, the speed of data chunking is slow in this mode.

In addition to the zoom mode, DLA Ganos Raster also provides the local mode that is based on the local coordinate system of the image. In local mode, the upper-left corner of the extent is defined as the starting point (0,0). Then, the data is chunked into tiles of 256 × 256 pixels until these tiles completely cover the image area. The remaining pixels are filled with NoData values. The benefit of this mode is that layer 0 uses the resolution of the raw dataset, which retains the original metadata. The local mode speeds up the data chunking, and helps you update the image and perform efficient queries. In this mode, different raster data does not use a uniform method to chunk data. This makes it difficult for accumulable analytics. By default, DLA Ganos Raster uses the local mode to chunk data and create pyramids.

Coordinate system

DLA Ganos Raster supports coordinate systems defined by the Open Geospatial Consortium (OGC) Coordinate Reference System (CRS) standard. You can use the EPSG parameters to define the CRS used for raster data. Common EPSG CRSs used by DLA Ganos Raster:
  1. EPSG:4326: a commonly used projection coordinate system, which is also known as WGS 84. If this coordinate system is used, data is projected in both latitude and longitude.
  2. EPSG:3857: a Pseudo-Mercator projection coordinate system, which is also called Web Mercator coordinate system. Mainstream web mapping applications such as Google Maps use this coordinate system.

For more information about the EPSG parameters, see https://epsg.io/.

Primary key and indexes

After the raster data is chunked into tiles, you must define how they are organized so that you can create indexes for them. Tiles are stored as key-value pairs in ApsaraDB for HBase Enhanced Edition (Lindorm). The key of each tile consists of attributes such as the layer name, level, SpaceTimeKey, row number, and column number. Ganos Raster of ApsaraDB for HBase Enhanced Edition (Lindorm) provides two key models:
  • SpatialKey: the spatial primary key.

    SpatialKey uses a space-filling curve (SFC) to encode and index tiles. Ganos Raster of ApsaraDB for HBase Enhanced Edition (Lindorm) supports three types of space indexes:



  • SpaceTimeKey: the spatio-temporal primary key.

    SpaceTimeKey is a three-dimensional SFC, which integrates a time dimension based on SpatialKey.

Spatio-temporal raster of DLA Ganos Raster

In DLA Ganos Raster, a tile is the basic unit that is used to process raster data. The serverless Spark engine loads all raster data by using TileUDT.

DLA Ganos Raster supports the following raster data sources:
  • PolarDB
  • ApsaraDB for HBase Enhanced Edition (Lindorm)
  • OSS
  • HDFS