Basic Approach#
Outline#
Questions#
What is the approach for finding events in dupin?
What is the basic definition for each step in the event detection scheme?
Objectives#
Explain the six step model for event detection used by dupin.
Define the basic outline of each step.
Six Step Detection#
dupin is based on a 6 step method for detecting events in point cloud data.
Generate: Generate basic features from the point cloud. Such features can be per-point, aggregate, or global.
Map: Transform the generated data into different distributions such as neighbor point averaging.
Reduce: Take the non-scalar features and create scalar features from them (e.g. maximum, minimum, spread, mean, …).
Aggregate: Perform the first three steps across the time dimension of the point cloud data.
Transform: Transform the signal using signal processing into a new feature signal (essentially the map step applied in the time direction). This step can involve feature selection or dimensionality reduction.
Detect: Given the final feature signal detect events if any in the data.
The first, fourth, and sixth steps are always necessary; however, most are useful in common use cases. This tutorial will go through all six step in the detection process.