TRK_SplitByDwells takes a track, a distance threshold, and a duration threshold and returns an array of tracks. The result array contains the input track split into segments wherever there is a dwell detected based on the distance threshold and duration threshold provided. The track is split by removing the point observations between the start and end point of a dwell.
Distance threshold is the maximum distance between point observations in a track to be considered a single dwell. Duration threshold is the minimum duration of a track to be considered a single dwell.
Dwell is defined as a track that contains sequential observations with little or no movement over a certain period of time.
Tracks are linestrings that represent the change in an entity's location over time. Each vertex in the linestring has a timestamp (stored as the M-value) and the vertices are ordered sequentially.
For more information on using tracks in GeoAnalytics Engine, see the core concept topic on tracks.
| Function | Syntax |
|---|---|
| Python | split |
| SQL | TRK |
| Scala | split |
For more details, go to the GeoAnalytics Engine API reference for split_by_dwells.
Python and SQL Examples
from geoanalytics.tracks import functions as TRK
from geoanalytics.sql import functions as ST
from pyspark.sql import functions as F
data = [(1, "POINT (0 0)", "2021-10-05 10:30:00"),
(1, "POINT (5 0)", "2021-10-05 10:31:00"),
(1, "POINT (10 0)", "2021-10-05 10:32:00"),
(1, "POINT (11.2 1.1)", "2021-10-05 10:33:00"),
(1, "POINT (12.2 -0.2)", "2021-10-05 10:34:00"),
(1, "POINT (11.2 -1.2)", "2021-10-05 10:35:00"),
(1, "POINT (10.5 -0.9)", "2021-10-05 10:36:00"),
(1, "POINT (10.6 0.5)", "2021-10-05 10:37:00"),
(1, "POINT (11.5 0)", "2021-10-05 10:38:00"),
(1, "POINT (11.6 -0.5)", "2021-10-05 10:39:00"),
(1, "POINT (10.5 -0.3)", "2021-10-05 10:40:00"),
(1, "POINT (12 0)", "2021-10-05 10:41:00"),
(1, "POINT (12.5 0)", "2021-10-05 10:42:00"),
(1, "POINT (18 0)", "2021-10-05 10:43:00"),
(1, "POINT (20 0)", "2021-10-05 10:44:00")]
df = spark.createDataFrame(data, ["id", "wkt", "datetime_str"]) \
.withColumn("point", ST.point_from_text("wkt", 3857)) \
.withColumn("timestamp", F.to_timestamp("datetime_str")) \
.drop("wkt", "datetime_str")
df_trk = df.groupBy("id").agg(TRK.aggr_create_track("point", "timestamp").alias("track"))
trk_split_by_dwells = df_trk.select(TRK.split_by_dwells("track", (3, 'Meters'), (1, 'Minutes')).alias("split_by_dwells"))
axes = df_trk.st.plot("track", edgecolor="lightgrey", linewidths=10, figsize=(15, 8))
trk_split_by_dwells.select(F.explode("split_by_dwells")).st.plot(ax=axes, edgecolor="greenyellow", linewidths=3)
axes.legend(['Input track','Result track'], loc='lower right', fontsize='large', bbox_to_anchor=(1, 0))
Scala Example
Scala
import com.esri.geoanalytics.sql.{functions => ST}
import com.esri.geoanalytics.sql.{trackFunctions => TRK}
import org.apache.spark.sql.{functions => F}
import java.sql.Timestamp
case class pointRow(id: Int, pointWkt: String, timestamp: Timestamp)
val data = Seq(pointRow(1, "POINT (0 0)", Timestamp.valueOf("2021-10-05 10:30:00")),
pointRow(1, "POINT (5 0)", Timestamp.valueOf("2021-10-05 10:31:00")),
pointRow(1, "POINT (10 0)", Timestamp.valueOf("2021-10-05 10:32:00")),
pointRow(1, "POINT (11.2 1.1)", Timestamp.valueOf("2021-10-05 10:33:00")),
pointRow(1, "POINT (12.2 -0.2)", Timestamp.valueOf("2021-10-05 10:34:00")),
pointRow(1, "POINT (11.2 -1.2)", Timestamp.valueOf("2021-10-05 10:35:00")),
pointRow(1, "POINT (10.5 -0.9)", Timestamp.valueOf("2021-10-05 10:36:00")),
pointRow(1, "POINT (10.6 0.5)", Timestamp.valueOf("2021-10-05 10:37:00")),
pointRow(1, "POINT (11.5 0)", Timestamp.valueOf("2021-10-05 10:38:00")),
pointRow(1, "POINT (11.6 -0.5)", Timestamp.valueOf("2021-10-05 10:39:00")),
pointRow(1, "POINT (10.5 -0.3)", Timestamp.valueOf("2021-10-05 10:40:00")),
pointRow(1, "POINT (12 0)", Timestamp.valueOf("2021-10-05 10:41:00")),
pointRow(1, "POINT (12.5 0)", Timestamp.valueOf("2021-10-05 10:42:00")),
pointRow(1, "POINT (18 0)", Timestamp.valueOf("2021-10-05 10:43:00")),
pointRow(1, "POINT (20 0)", Timestamp.valueOf("2021-10-05 10:44:00")))
val points = spark.createDataFrame(data)
.withColumn("point", ST.pointFromText($"pointWkt", F.lit(3857)))
val tracks = points.groupBy("id").agg(TRK.aggrCreateTrack($"point", $"timestamp").alias("track"))
val splitTracks = tracks.select(TRK.splitByDwells($"track", F.lit(struct(F.lit(3).as("value"), F.lit("Meters").as("units"))),
F.lit(struct(F.lit(1).as("value"), F.lit("Minutes").as("units")))).alias("split_by_dwells"))
splitTracks.show(truncate = false)Result
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|split_by_dwells |
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|[{"hasM":true,"paths":[[[0,0,1.633455e9],[5,0,1.63345506e9],[10,0,1.63345512e9]]]}, {"hasM":true,"paths":[[[12.5,0,1.63345572e9],[18,0,1.63345578e9],[20,0,1.63345584e9]]]}]|
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------+Version table
| Release | Notes |
|---|---|
1.5.0 | Python, SQL, and Scala functions introduced |