Module: clustering_filters#

class kbmod.filters.clustering_filters.ClusterGridFilter(cluster_eps, pred_times)[source]#

Use a discrete grid to cluster the points. Each trajectory is fit into a bin and only the best trajectory per bin is retained.

Attributes:

bin_widthint: The width of the grid bins (in pixels).
cluster_gridTrajectoryClusterGrid: The grid of best result trajectories seen.
max_dtfloat: The maximum different between times in pred_times.

Methods

`get_filter_name`()	Get the name of the filter.
`keep_indices`(result_data)	Determine which of the results's indices to keep.

get_filter_name()[source]#

Get the name of the filter.

Returns:

str: The filter name.

keep_indices(result_data)[source]#

Determine which of the results’s indices to keep.

Parameters:

result_data: `Results`: The set of results to filter.

Returns:

list: A list of indices (int) indicating which rows to keep.

class kbmod.filters.clustering_filters.ClusterPosVelFilter(cluster_eps, cluster_v_scale=1.0, **kwargs)[source]#

Cluster the candidates using their starting position and velocities.

Methods

`get_filter_name`()	Get the name of the filter.
`keep_indices`(result_data)	Determine which of the results's indices to keep.

class kbmod.filters.clustering_filters.ClusterPredictionFilter(cluster_eps, pred_times=[0.0], **kwargs)[source]#

Cluster the candidates using their positions at specific times.

Attributes:

timeslist-like: The times at which to evaluate the trajectories (in days).

Methods

`get_filter_name`()	Get the name of the filter.
`keep_indices`(result_data)	Determine which of the results's indices to keep.

class kbmod.filters.clustering_filters.DBSCANFilter(cluster_eps, **kwargs)[source]#

Cluster the candidates using DBSCAN and only keep a single representative trajectory from each cluster.

Attributes:

cluster_epsfloat: The clustering threshold (in pixels).
cluster_typestr: The type of clustering.
cluster_argsdict: Additional arguments to pass to the clustering algorithm.

Methods

`get_filter_name`()	Get the name of the filter.
`keep_indices`(result_data)	Determine which of the results's indices to keep.

get_filter_name()[source]#

Get the name of the filter.

Returns:

str: The filter name.

keep_indices(result_data)[source]#

Determine which of the results’s indices to keep.

Parameters:

result_data: `Results`: The set of results to filter.

Returns:

list: A list of indices (int) indicating which rows to keep.

class kbmod.filters.clustering_filters.NNSweepFilter(cluster_eps, pred_times)[source]#

Filter any points that have neighboring trajectory with a higher likleihood within the threshold.

Parameters:

threshfloat: The filtering threshold to use (in pixels).
timeslist-like: The times at which to evaluate the trajectories (in days).

Methods

`get_filter_name`()	Get the name of the filter.
`keep_indices`(result_data)	Determine which of the results's indices to keep.

get_filter_name()[source]#

Get the name of the filter.

Returns:

str: The filter name.

keep_indices(result_data)[source]#

Determine which of the results’s indices to keep.

Parameters:

result_data: `Results`: The set of results to filter.

Returns:

list: A list of indices (int) indicating which rows to keep.

kbmod.filters.clustering_filters.apply_clustering(result_data, cluster_params)[source]#

This function clusters results that have similar trajectories.

Parameters:

result_data: `Results`: The set of results to filter. This data gets modified directly by the filtering.
cluster_paramsdict: Contains values concerning the image and search settings including: cluster_type, cluster_eps, times, and cluster_v_scale (optional).

Raises:

Raises a ValueError if the parameters are not valid.
Raises a TypeError if result_data is of an unsupported type.