Point Cloud Active Selection via Feature Extraction and Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current manual selection and annotation of target point cloud data from three-dimensional point cloud data is time-consuming and labor-intensive, due to the disordered and irregular arrangement of the data.
Innovation Solution
A data active selection and annotation method and apparatus for point cloud, which involves inputting initial point cloud data into a feature extraction model to extract features, using a classification model to obtain classification results, and determining pseudo labels to filter to-be-annotated point cloud data based on extracted features and classification results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual filtering and annotation of target point cloud data is performed, then annotation accuracy can be ensured, but time consumption and labor cost increase significantly
Solution Approach 1:
The patent applies preliminary action by performing feature extraction and classification on unannotated point cloud data before annotation. The system extracts features from unannotated data, compares them with features from annotated data, and pre-identifies candidate target data that matches the query data. This preliminary processing reduces the volume of data requiring manual annotation, thereby reducing time consumption while maintaining annotation accuracy through automated pre-screening.
2Measurement precision
If manual filtering and annotation of target point cloud data is performed, then annotation quality can be ensured, but labor cost increases significantly
Solution Approach 1:
The patent applies self-service by enabling the system to automatically perform feature extraction, classification, and candidate identification without human intervention. The automated system extracts features from point cloud data, performs classification based on extracted features, and identifies candidate target data that matches query data characteristics. This automation eliminates the need for manual labor in the filtering and selection process, significantly reducing labor costs while maintaining annotation quality through consistent automated processing.
3Loss of time
If feature extraction and classification models are used to automatically select point cloud data, then time consumption and labor cost are reduced, but system complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the automated selection system into distinct functional modules: a feature extraction module that extracts features from point cloud data, a classification model that performs classification based on extracted features, and a candidate identification module that identifies target data matching query data. This modular segmentation manages system complexity by organizing functions into separate, manageable components while enabling automated time-efficient data selection.
4Ease of manufacture
If automated filtering based on features and classification results is implemented, then labor cost is reduced, but computational requirements increase
Solution Approach 1:
The patent applies partial action by implementing automated filtering that processes only the necessary features and classification results required for candidate identification, rather than performing exhaustive analysis on all point cloud data. The system extracts relevant features, performs classification on subsets of data, and identifies candidates based on matching criteria. This partial processing approach reduces labor cost through automation while managing computational requirements by avoiding unnecessary full-data processing.
Data Source
AI summary
A data active selection and annotation method for a point cloud includes: inputting initial point cloud data into a feature extraction model to extract a first feature of annotated point cloud data and a second feature of unannotated point cloud data, the initial point cloud data including the annotated point cloud data and the unannotated point cloud data; inputting the unannotated point cloud data into a classification model to obtain a classification result of the unannotated point cloud data; determining each piece of target point cloud data with a pseudo label identical to a real label from the unannotated point cloud data according to the classification result and the real label of the annotated point cloud data; filtering to-be-annotated point cloud data from each piece of target point cloud data according to the first feature, a second feature and the classification result of each piece of target point cloud data.