I understand that you are trying to recognize larger shapes (such as pyramids, cylinders, planes) from clustered point cloud data, falls under the domain of point cloud segmentation and feature extraction. The PointCloud library itself focuses more on the basic manipulation and processing of point clouds, so for advanced shape recognition, you might need to employ additional techniques and libraries.
Here's a general approach you can consider to recognize larger shapes in your clustered point cloud data:
- Plane Detection: Identify and extract planes from your clustered point cloud data. The RANSAC algorithm is commonly used for plane detection. Once planes are detected, you can remove the points belonging to planes from the point cloud to focus on remaining shapes.
- Shape Recognition: Shape recognition can involve more complex algorithms and might require libraries like PCL (Point Cloud Library) or other machine learning libraries. You could use techniques like Random Forests, SVMs, or Neural Networks to classify points into different shapes based on their features.
- Feature Extraction: Extract features from the point cloud clusters that can be used for shape recognition. These features might include things like the normal vector of the points, centroid, bounding box, and more.
Remember that shape recognition from point clouds can be a complex task that might require a combination of techniques and possibly custom algorithm development. Libraries like PCL can provide a foundation for some of these tasks, but you might need to integrate multiple approaches and tools to achieve accurate and robust shape recognition results.
Refer to this documentation for more information:
