3D Point Cloud Recognition Confidence for Adjustable Detection Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Recognition processing on 3D data often has a high processing cost, particularly when changing the detection intensity of the recognition target, as it requires re-executing the processing each time.
Innovation Solution
An information processing apparatus that includes a 3D point cloud recognition unit for recognition processing and a recognition confidence calculation unit to calculate confidence for each point of the 3D point cloud, allowing extraction of point clouds based on confidence without re-executing recognition processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If recognition processing is re-executed to change detection intensity, then detection intensity can be adjusted, but processing cost increases
Solution Approach 1:
The patent extracts the confidence information from the recognition processing results and separates it as an independent output. This allows the recognition processing to be executed once and then used for multiple detection intensity levels by simply filtering based on confidence thresholds, rather than re-executing the recognition processing for each intensity level.
Solution Approach 2:
The patent performs the recognition processing and confidence calculation in advance as a preliminary action. By executing the recognition processing once and storing the confidence values, the system prepares the necessary data structures that can be efficiently queried at different detection intensities without requiring re-processing.
2Adaptability or versatility
If recognition processing is executed multiple times for different detection intensities, then detection flexibility is improved, but processing time increases
Solution Approach 1:
The patent extracts confidence values as a separate output from the recognition processing. This extraction allows the system to maintain a single recognition processing execution while enabling multiple detection intensity levels through confidence-based filtering, thereby reducing the time that would otherwise be spent on repeated processing.
Solution Approach 2:
The patent makes the recognition processing result universal by including confidence values that can be used for multiple detection intensity requirements. The same recognition processing output serves multiple purposes at different intensity levels, eliminating the need for separate processing executions for each intensity level.
3Measurement precision
If confidence is calculated for each point, then detection accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies local quality by calculating confidence values specifically for each point in the point cloud based on local region statistics. This allows the system to maintain high measurement precision for recognition confidence while managing computational complexity by focusing calculations on local point properties rather than global re-processing.
Data Source
AI summary
An information processing apparatus according to an aspect of the present disclosure includes: a three-dimensional point cloud recognition unit that executes recognition processing on a three-dimensional point cloud and gives a recognition result for each point of the three-dimensional point cloud; and a recognition confidence calculation unit that calculates a confidence of the recognition result for each point of the three-dimensional point cloud and gives the confidence for each point of the three-dimensional point cloud.