feat: Add Point Cloud Filter and 3D People Tracking Verification - #9
Open
lucumango wants to merge 1 commit into
Open
feat: Add Point Cloud Filter and 3D People Tracking Verification#9lucumango wants to merge 1 commit into
lucumango wants to merge 1 commit into
Conversation
There was a problem hiding this comment.
Pull request overview
This PR aims to improve 3D tracking stability by adding an optional Statistical Outlier Removal (SOR) filter in the point-cloud preprocessing path and introducing a pytest-based verification that simulates tracking a moving 3D torso trajectory.
Changes:
- Added optional SOR outlier filtering to
BaseTracker.normalize_data()to remove noisy radar points before downstream processing. - Added a new synthetic 3D torso tracking test intended to validate Asterios tracking behavior with noisy point clouds.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 8 comments.
| File | Description |
|---|---|
mwcore/tracking/api/base.py |
Extends normalize_data() with optional SOR filtering and updates input unpacking behavior. |
tests/test_3d_tracking.py |
Adds a synthetic-trajectory pytest intended to validate filtering and tracking behavior. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
You can also share your feedback on Copilot code review. Take the survey.
Comment on lines
+67
to
+90
| for i, center in enumerate(trajectory): | ||
| # Generate synthetic frame with noise | ||
| frame_points = generate_synthetic_torso_frame(center) | ||
|
|
||
| # We manually monkey-patch filter_outliers directly through the consume dictionary if kwargs allow, | ||
| # but since consume doesn't explicitly pass kwargs to normalize_data yet, we'll patch the instance method | ||
| # for testing purposes or assert behavior directly. | ||
| # Let's call normalize_data separately to verify SOR logic. | ||
|
|
||
| # 1. Test SOR Filter mathematically | ||
| normalized_with_filter = tracker.normalize_data(point_array=frame_points, filter_outliers=True, sor_neighbors=5, sor_std_ratio=1.0) | ||
| normalized_without_filter = tracker.normalize_data(point_array=frame_points, filter_outliers=False) | ||
|
|
||
| # Outliers should be dropped | ||
| assert len(normalized_with_filter) < len(normalized_without_filter), "SOR filter failed to drop outliers." | ||
|
|
||
| # 2. Consume data in Tracker | ||
| states = tracker.consume(point_array=normalized_with_filter) | ||
|
|
||
| if states: | ||
| # Get primary tracked state (Centroid) | ||
| tracked_states.append(states[0]) | ||
|
|
||
| # Allow a few frames for the Kalman Filter to initialize and establish a Track |
| @@ -0,0 +1,99 @@ | |||
| import numpy as np | |||
| import pytest | |||
Comment on lines
63
to
+71
| if det_obj is not None: | ||
| input_data = np.vstack( | ||
| (det_obj["x"], det_obj["y"], det_obj["z"], det_obj["doppler"], det_obj["peakVal"]) | ||
| ).T | ||
| if point_array is not None: | ||
| input_data = point_array | ||
|
|
||
| if filter_outliers and len(input_data) > sor_neighbors: | ||
| from scipy.spatial import cKDTree |
Comment on lines
19
to
+26
| def normalize_data(self, | ||
| det_obj: Optional[dict] = None, | ||
| point_array: Optional[np.ndarray] = None, | ||
| keepRadial: bool = False, | ||
| transform: bool = False): | ||
| transform: bool = False, | ||
| filter_outliers: bool = False, | ||
| sor_neighbors: int = 5, | ||
| sor_std_ratio: float = 1.0): |
Comment on lines
88
to
91
| for index in range(len(input_data)): | ||
| x, y, z, doppler, peakVal = input_data[index] | ||
| x, y, z, doppler, peakVal = input_data[index][:5] | ||
| # Compute polar coordinates | ||
| r = math.sqrt(x**2 + y**2 + z**2) |
Comment on lines
+76
to
+83
| mean_sq_distances = np.mean(distances[:, 1:], axis=1) | ||
|
|
||
| global_mean_dist = np.mean(mean_sq_distances) | ||
| global_std_dist = np.std(mean_sq_distances) | ||
|
|
||
| # Keep points whose mean neighbor distance is within the threshold | ||
| threshold = global_mean_dist + (sor_std_ratio * global_std_dist) | ||
| mask = mean_sq_distances <= threshold |
Comment on lines
+83
to
+84
| # 2. Consume data in Tracker | ||
| states = tracker.consume(point_array=normalized_with_filter) |
Comment on lines
+5
to
+24
| def generate_synthetic_torso_frame(center, num_points=20, spread=0.3): | ||
| """ | ||
| Generates a noisy cluster of points representing a human torso. | ||
| Format required by tracking base: [x, y, z, doppler, peakVal] | ||
| """ | ||
| points = np.random.normal(loc=0.0, scale=spread, size=(num_points, 5)) | ||
| # Apply center offset to x, y, z | ||
| points[:, 0] += center[0] | ||
| points[:, 1] += center[1] | ||
| points[:, 2] += center[2] | ||
| # Synthetic doppler and peakVal | ||
| points[:, 3] = np.random.uniform(-1.0, 1.0, size=num_points) | ||
| points[:, 4] = np.random.uniform(50, 100, size=num_points) | ||
|
|
||
| # Add a few extreme random noise points (outliers) to test the SOR filter | ||
| outliers = np.random.uniform(-5.0, 5.0, size=(5, 5)) | ||
| outliers[:, 3] = 0.0 | ||
| outliers[:, 4] = 10.0 | ||
|
|
||
| return np.vstack((points, outliers)) |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
This PR integrates a Statistical Outlier Removal (SOR) point cloud filter into
base.pyto reduce radar noise prior to tracking.Additionally, it adds a dedicated
pytestsuite simulating a 3D human torso trajectory, provingAsterioscan successfully track the silhouette.Resolves #8