Cluster estimation filter

config.yaml

@@ -74,6 +74,7 @@ cluster_estimation:
     min_new_points_to_run: 5
     max_num_components: 10
     random_state: 0
+    min_points_per_cluster: 3
 
 communications:
     timeout: 30.0  # seconds

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)
+        # Get counts for each cluster 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

@@ -20,6 +20,7 @@ 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.
@@ -64,6 +65,7 @@ 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_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