Cluster estimation filter (#253)

* min_points_per_cluster

* config

* format

* fix

* format

* update

* update

* 'format'

* 'format'

* 'update'

* debug

* integration

* format

* main_edit

Author: szbillyang

config.yaml

@@ -76,6 +76,7 @@ cluster_estimation:
     min_new_points_to_run: 3
     max_num_components: 10
     random_state: 0
+    min_points_per_cluster: 3
 
 communications:
     timeout: 60.0  # seconds

main_2025.py

@@ -154,6 +154,7 @@ def main() -> int:
         MIN_NEW_POINTS_TO_RUN = config["cluster_estimation"]["min_new_points_to_run"]
         MAX_NUM_COMPONENTS = config["cluster_estimation"]["max_num_components"]
         RANDOM_STATE = config["cluster_estimation"]["random_state"]
+        MIN_POINTS_PER_CLUSTER = config["cluster_estimation"]["min_points_per_cluster"]
 
         COMMUNICATIONS_TIMEOUT = config["communications"]["timeout"]
         COMMUNICATIONS_WORKER_PERIOD = config["communications"]["worker_period"]
@@ -354,6 +355,7 @@ def main() -> int:
             MIN_NEW_POINTS_TO_RUN,
             MAX_NUM_COMPONENTS,
             RANDOM_STATE,
+            MIN_POINTS_PER_CLUSTER,
         ),
         input_queues=[geolocation_to_cluster_estimation_queue],
         output_queues=[cluster_estimation_to_communications_queue],

modules/cluster_estimation/cluster_estimation.py

@@ -1,6 +1,6 @@
 """
 Take in bounding box coordinates from Geolocation and use to estimate landing pad locations.
-Returns an array of classes, each containing the x coordinate, y coordinate, and spherical 
+Returns an array of classes, each containing the x coordinate, y coordinate, and spherical
 covariance of each landing pad estimation.
 """
 
@@ -14,6 +14,7 @@
 from ..common.modules.logger import logger
 
 
+# pylint: disable=too-many-instance-attributes
 class ClusterEstimation:
     """
     Estimate landing pad locations based on landing pad ground detection. Estimation
@@ -63,6 +64,7 @@ def create(
         max_num_components: int,
         random_state: int,
         local_logger: logger.Logger,
+        min_points_per_cluster: int,
     ) -> "tuple[bool, ClusterEstimation | None]":
         """
         Data requirement conditions for estimation model to run.
@@ -83,6 +85,9 @@ def create(
         local_logger: logger.Logger
             The local logger to log this object's information.
 
+        min_points_per_cluster: int
+            Minimum number of points that must be assigned to a cluster for it to be considered valid.
+
         RETURNS: The ClusterEstimation object if all conditions pass, otherwise False, None
         """
         if min_activation_threshold < max_num_components:
@@ -97,13 +102,17 @@ def create(
         if random_state < 0:
             return False, None
 
+        if min_points_per_cluster < 1:
+            return False, None
+
         return True, ClusterEstimation(
             cls.__create_key,
             min_activation_threshold,
             min_new_points_to_run,
             max_num_components,
             random_state,
             local_logger,
+            min_points_per_cluster,
         )
 
     def __init__(
@@ -114,6 +123,7 @@ def __init__(
         max_num_components: int,
         random_state: int,
         local_logger: logger.Logger,
+        min_points_per_cluster: int,
     ) -> None:
         """
         Private constructor, use create() method.
@@ -140,6 +150,7 @@ def __init__(
         self.__min_new_points_to_run = min_new_points_to_run
         self.__has_ran_once = False
         self.__logger = local_logger
+        self.__min_points_per_cluster = min_points_per_cluster
 
     def run(
         self, detections: "list[detection_in_world.DetectionInWorld]", run_override: bool
@@ -337,15 +348,16 @@ def __filter_by_points_ownership(
         # List of each point's cluster index
         cluster_assignment = self.__vgmm.predict(self.__all_points)  # type: ignore
 
-        # Find which cluster indices have points
-        clusters_with_points = np.unique(cluster_assignment)
+        # Check each cluster has enough points associated to it by index
+        unique, counts = np.unique(cluster_assignment, return_counts=True)
+        cluster_counts = dict(zip(unique, counts))
 
         # Remove empty clusters
         filtered_output: "list[tuple[np.ndarray, float, float]]" = []
         # By cluster index
         # pylint: disable-next=consider-using-enumerate
         for i in range(len(model_output)):
-            if i in clusters_with_points:
+            if cluster_counts.get(i, 0) >= self.__min_points_per_cluster:
                 filtered_output.append(model_output[i])
 
         return filtered_output

modules/cluster_estimation/cluster_estimation_worker.py

@@ -21,6 +21,7 @@ def cluster_estimation_worker(
     input_queue: queue_proxy_wrapper.QueueProxyWrapper,
     output_queue: queue_proxy_wrapper.QueueProxyWrapper,
     controller: worker_controller.WorkerController,
+    min_points_per_cluster: int,
 ) -> None:
     """
     Estimation worker process.
@@ -67,6 +68,7 @@ def cluster_estimation_worker(
         max_num_components,
         random_state,
         local_logger,
+        min_points_per_cluster,
     )
     if not result:
         local_logger.error("Worker failed to create class object", True)

tests/integration/test_cluster_estimation_worker.py

@@ -17,6 +17,7 @@
 MIN_NEW_POINTS_TO_RUN = 0
 MAX_NUM_COMPONENTS = 3
 RANDOM_STATE = 0
+MIN_POINTS_PER_CLUSTER = 3
 
 
 def check_output_results(output_queue: queue_proxy_wrapper.QueueProxyWrapper) -> None:
@@ -49,6 +50,7 @@ def test_cluster_estimation_worker() -> int:
             MIN_NEW_POINTS_TO_RUN,
             MAX_NUM_COMPONENTS,
             RANDOM_STATE,
+            MIN_POINTS_PER_CLUSTER,
             input_queue,
             output_queue,
             controller,

tests/integration/test_communications_to_ground_station.py

@@ -1,5 +1,5 @@
 """
-Test MAVLink integration test 
+Test MAVLink integration test
 """
 
 import multiprocessing as mp

tests/unit/test_cluster_detection.py

@@ -16,6 +16,7 @@
 MAX_NUM_COMPONENTS = 10
 RNG_SEED = 0
 CENTRE_BOX_SIZE = 500
+MIN_POINTS_PER_CLUSTER = 3
 
 # Test functions use test fixture signature names and access class privates
 # No enable
@@ -37,6 +38,7 @@ def cluster_model() -> cluster_estimation.ClusterEstimation:  # type: ignore
         MAX_NUM_COMPONENTS,
         RNG_SEED,
         test_logger,
+        MIN_POINTS_PER_CLUSTER,
     )
     assert result
     assert model is not None
@@ -489,3 +491,33 @@ def test_position_regular_data(
                     break
 
             assert is_match
+
+
+class TestMinimumPointsPerCluster:
+    """
+    Tests that clusters with fewer than the minimum required points are filtered out.
+    """
+
+    __STD_DEV_REG = 1
+
+    def test_outlier_is_filtered(self, cluster_model: cluster_estimation.ClusterEstimation) -> None:
+        """
+        Verify that a single outlier (cluster with only one point) is filtered out,
+        while a valid cluster with enough points is retained.
+        """
+        # Setup
+        valid_detections, valid_cluster_positions = generate_cluster_data([100], self.__STD_DEV_REG)
+        outlier_detections = generate_points_away_from_cluster(
+            num_points_to_generate=1,
+            minimum_distance_from_cluster=20,
+            cluster_positions=valid_cluster_positions,
+        )
+        generated_detections = valid_detections + outlier_detections
+
+        # Run
+        result, detections_in_world = cluster_model.run(generated_detections, False)
+
+        # Test
+        assert result
+        assert detections_in_world is not None
+        assert len(detections_in_world) == 1