Description
Represents a task for resampling raw vision data to generate vision tiles for a single camera. The task uses vision input data and terrain tiles, which are profile tiles that are sampled at a resolution appropriate for use by the vision sampler, and which has all of its null values filled in with best-guess values. The task accepts GoWebScanVisionInputMsg objects and GoWebScanProfileTileMsg objects (for presence data) and produces GoWebScanVisionTileMsg objects.
When vision input or profile tile messages are received, they are passed immediately to the vision combiner, which retains messages until sufficient profile data is available cover the maximum potential Y-extent of the next vision message. When the vision combiner produces a frame, which includes a single vision input message and corresponding terrain tiles, the vision input is processed and then passed to the vision resampler. The vision resampler emits completed rows as they become available, and those rows are added to an output tile.
Tiles emitted by the VisionSampleTask can contain null pixels. These occur when the field of view of the camera is insufficient to cover the tile, when the speed of the system is too great, resulting in Y-gaps between images, or when input images have been dropped. The x field of the kRbg structure contains a zero value to indicate a null pixel, or a non-zero value to indicate a valid pixel.
Processing steps:
- Vision input and profile tile messages passed to vision combiner.
- Vision combiner matches vision input to a set of profile tiles.
- Vision input is reoriented to standard system orientation.
- Gains from system calibration are applied to the vision input channels.
- Vision input is demosaiced to form an RGB image.
- Vision input is added to the GoWebScanVisionResampler algorithm.
- Resampled row is assembled into emitted GoWebScanVisionTileMsg.
The documentation for this class was generated from the following file: