PolarDB:Common GanosBase methods for processing trajectory sampling points

ST_removeDriftPoints — trajectory filtering

ST_removeDriftPoints removes points that deviate significantly from the expected movement pattern. It handles two types of anomalies:

• Spatial drift caused by poor GPS signal quality

• Speed anomalies caused by objects jumping between locations within a short time frame due to dense but imprecise sampling

For more information, see ST_removeDriftPoints.

Example

SELECT ST_removeDriftPoints(
    ST_SetSRID(
        ST_MakeTrajectory(
            ARRAY_AGG(ROW(traj.arrival_time::TIMESTAMP, st_x(traj.pts)::DOUBLE PRECISION, traj.lat::DOUBLE PRECISION, traj.rowid)),
            FALSE, '{"rowid"}'::cstring[]
        ), 4326
    ),
    40,   -- maximum speed threshold
    10,   -- minimum distance threshold
    '1 minute'::INTERVAL  -- minimum time interval
) FROM (
    SELECT
        time,
        ST_makepoint(lon, lat) pts,
        lat,
        rowid
    FROM point_table
    WHERE time IS NOT NULL AND lon IS NOT NULL AND lat IS NOT NULL
    ORDER BY rowid
) traj INTO trajectory_table;

ST_Split — trajectory segmentation

ST_Split divides a trajectory into multiple sub-trajectories using geometric objects. It provides three segmentation modes:

For more information, see ST_Split.

Example (remove edges longer than 10 spatial units in 2D)

WITH traj AS (
    SELECT '{"trajectory":{"version":1,"type":"STPOINT","leafcount":19,...}}'::trajectory AS a
)
SELECT ST_Split(a, '{"drop_edge.spatial_distance_2d":10}') FROM traj;

ST_Resample — trajectory resampling

ST_Resample corrects uneven point distribution across a trajectory. Uneven density typically results from noise-reduction preprocessing and can degrade the performance of local trajectory analysis.

Two resampling strategies are available:

For more information, see ST_Resample.

Example (add interpolated points so consecutive points are no more than 5 minutes apart)

SELECT ST_Resample(ST_OnlyST(traj), '{"add_point.period_lesser":"5 minute"}')
FROM table;

Trajectory simplification

Trajectory simplification (lossy compression) reduces the number of sampling points while preserving the overall movement shape. This decreases storage requirements and improves rendering performance.

Two simplification functions are available:

How the algorithms work

The Douglas-Peucker algorithm specifies a tolerance value — the maximum allowed deviation between the original and simplified trajectory. Starting from the endpoints, it iteratively identifies the point farthest from the simplified path, adds it back, and repeats until all remaining points fall within the tolerance. The algorithm operates in 2D and requires all points to lie on the same plane.

For spatiotemporal trajectories, applying the 2D Douglas-Peucker algorithm can produce incorrect results. When an object is stationary, its trajectory forms a vertical line along the time axis that collapses to a single point in 2D — causing the standard algorithm to drop one of the line's endpoints and distort the 3D structure. ST_CompressSED uses Synchronized Euclidean Distance (SED) instead of perpendicular Euclidean distance, which preserves the temporal dimension and produces accurate compression for spatiotemporal data.