Pipeline for open AIS data with filtering based on vessel class

Mirjam Bayer, Tabea Fry, Sören Dethlefsen, Daniyal Kazempour

Abstract: With declining fish populations, there is a high demand for measures to monitor fishing efforts and promote maritime conservation. Due to advancing technologies, data documenting vessel activity is continuously growing. The generated data and data streams of vessels in the form of satellite-based communication such as Automatic Identification System (AIS) contain valuable information on the activities performed, yet this information is hidden inside the massive volume of data streams. To facilitate research on AIS data, we present a pipeline that extracts useful data sets from standardized AIS message streams. The pipeline is equipped to process any standardized AIS message data stream, as demonstrated on AIS messages of the Danish Marine Authority (DMA).The presented data set generator allows filtering and cleaning available AIS messages; reducing the otherwise massive volume of data and yielding a specific data set for further research like trajectory classification or fishing effort estimation. The implemented filter is based on mandatory features contained in AIS message. These features are: "Navigational status", "Ship Type" and "Cargo Type". The pipeline can output data sets of three different vessel classes given the AIS data, resulting in data sets containing either messages from merchandise ships, fishing vessels, or passenger ships. The data can optionally be cleaned and enriched by additional features such as water depth, distance to shore, or trip and anchorage annotation.