DrowningWarning-analytics

analyzing data collected from the box (containing accelerometer + gyro sesonrs)

PreProcessing

The bach script preprocessAll.sh calls the python script, i.e. the preprocessing pipeline preprocess.py that currently does roughly this:

Telespor (telespor)

• cleans telespor export for lat, lon, and batteryvoltage
• calculates speed and engine state (on/off) from telespor data
• resmaples the data to a sampling rate that could be set via arguments and interpolates missing lat lon, and batteryvoltage values

Box of sensors (capture)

• adds absolute timestmps to the data using the recorded time of installation

• resamples (downsamples) the data to the same sampling rate as telespor measurements by averaging meaurements inside the bins

• For preparing training data, preprocess.py also accepts two timestamps (signalstart & signalend) that limits the measurements used to the time range during which the boat was actually moving

Classification

• Features: box measurements (ax, ay, az, gx, gy, gz)
• Labels: engine state (ON/OFF) or batteryvoltage (passed as a sys arg to classify.py)

Weather (used)

• AROME-arctic full model forecast
• WAM4 archive

Weather (for reference)

• list of free MET data
• Latest YR forecast (including WAM forecasts)