| 5 | | == Image processing pipelines == |
| 6 | | Image processing is done in so-called pipelines, usually sequential steps to process the image to extract the desired data. See the attached document for a writeup of the procedures used. |
| | 5 | == Terminology of !FireVision == |
| | 6 | |
| | 7 | Here are some basic definitions what we mean when we talk about !FireVision. It's mainly a writeup of how it evolved. |
| | 8 | |
| | 9 | '''Region of Interest (ROI):''' |
| | 10 | Regions in the image that are interesting in the sense that some criterion has chosen this region (see ''Classifiers'' below). |
| | 11 | '''Classifiers:''' |
| | 12 | Classifiers analyze the image for a given criterion (like color) and extract a set of regions of interest. These regions of interest are then considered to probably contain the object (although this may not hold) and they are the candidates for further processing. |
| | 13 | '''Filters:''' |
| | 14 | Filters transform images. They get an image as input and have another image as output. The processing can be constrained with ROI so that only a part of the image is processed which usually results in tremendous speedups. |
| | 15 | '''Models:''' |
| | 16 | Models transform data. You feed data and you get data. An example a position models. They usually get the position of the object in the image and from that with additional parameters calculate the position of the object relative to the robot or with the robot position the global coordinates. Models do ''not'' operate on images. |
| | 17 | '''Image processing pipelines:''' |
| | 18 | Image processing is done in so-called pipelines, usually sequential steps to process the image to extract the desired data. See the attached document for a writeup of the procedures used. They are basically the using part of the terms defined above. |
