Stereo Camera Person Detection With 2D-Guided 3D Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor systems for monitoring spatial regions in outdoor or industrial environments face challenges in efficiently recognizing persons due to limited computing power and high demands on evaluation speed, particularly in distinguishing between persons and other objects.
Innovation Solution
A sensor system utilizing a stereo camera and artificial intelligence-based person detection method processes 2D data to recognize persons, determining 2D position data, which are then used to assign and determine 3D position data, reducing computational requirements and enabling fast, reliable person recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If person detection is performed on 3D data directly, then measurement precision and reliability of person recognition are improved, but computing power consumption increases and evaluation speed decreases
Solution Approach 1:
The patent segments the person detection process into two distinct stages: first performing detection on 2D data to identify candidate regions, then refining position determination using 3D data only for those specific regions. This segmentation avoids processing all 3D data comprehensively while maintaining accurate 3D position determination for detected persons.
Solution Approach 2:
The patent extracts and utilizes only the essential 2D spatial information needed for initial person detection, separating this from the full 3D data processing. By taking out just the necessary 2D projection data for detection purposes, the system achieves fast evaluation while preserving the ability to determine accurate 3D positions when needed.
2Productivity
If person detection is performed on 2D data, then computing power consumption is reduced and evaluation speed is improved, but measurement precision for 3D position determination deteriorates
Solution Approach 1:
The patent merges the advantages of both 2D and 3D data processing by combining them in a two-stage approach: using 2D data for fast initial detection and combining this with 3D data for accurate position determination. This merging allows the system to achieve both high evaluation speed and accurate 3D position measurement.
Solution Approach 2:
The patent transitions from 2D detection to 3D position determination by projecting 3D spatial data onto 2D image planes for detection, then using the detected 2D positions to retrieve corresponding 3D coordinates. This dimensionality change strategy enables efficient processing while maintaining 3D accuracy.
3Reliability
If comprehensive 3D data processing is used for person detection, then reliability of person recognition is improved, but computing power requirements increase beyond available resources in mobile applications
Solution Approach 1:
The patent applies partial action by performing complete 3D-based person detection only when necessary (when a person is detected in 2D data), rather than processing all 3D data continuously. This partial processing approach maintains high reliability for person recognition while significantly reducing computing power consumption during normal operation.
Data Source
AI summary
A sensor system for monitoring a spatial region in an outdoor region or in an industrial plant, for example for use on a manned vehicle or an autonomously driving industrial truck, includes a sensor arrangement. The sensor system is configured to generate 2D and 3D data of the spatial region. In this respect, the sensor system has a computing device that is configured to perform a person detection method on the 2D data of the spatial region in order to recognize persons in the spatial region. 2D position data are determined for a recognized person. The computing device is furthermore configured, on the basis of the 2D position data for the recognized person, to assign 3D data associated with the recognized person. The computing device is further configured to determine 3D position data for the recognized person from the 3D data associated with the recognized person.