Graphics Reference
In-Depth Information
Fig. 12.2
An illustration of an adaptive and scalable classifier
highly relevant to design and implementation of ASM systems because they enable
techniques for exploiting characteristics of the currently arriving set of image streams
as well as characteristics of the overall operating environment to dynamically opti-
mize critical trade-offs among key execution metrics, including power consumption,
communication bandwidth, knowledge extraction accuracy, and end-to-end latency.
In ASM systems for embedded computer vision, DDDAS can be employed, for
example, at network edges to systematically filter out image features that are not rel-
evant to the current operational scenario or to adjust the resolution or frequency
of captured images based on the type of object or amount of motion detected.
