Person Detector Plausibility Check for Functional Safety
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional person detectors do not meet the requirements for functional safety, failing to reliably diagnose errors in camera-based systems, which are crucial for safe human-robot interaction in industrial environments.
Innovation Solution
A monitoring system with a diagnostic unit that includes plausibility checks, cross-comparisons, and additional information to verify the accuracy of person detection, utilizing diverse sensors and AI models, and generating phantom images to ensure error detection in person detectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional person detectors using camera-based AI analysis are used, then person detection capability is achieved, but reliability for functional safety requirements is insufficient
Solution Approach 1:
The system segments the person detection task into multiple independent evaluation channels (first evaluation unit, second evaluation unit) that process images through different AI models or sensor types. Each channel operates independently to detect persons, and their results are cross-compared to identify errors, thereby improving reliability without requiring a single complex diagnostic system.
Solution Approach 2:
A diagnostic unit is introduced as an intermediary component that receives position indications from multiple evaluation channels and performs plausibility checks. This intermediary cross-compares results from different channels to detect errors, enabling functional safety requirements to be met while keeping individual detection channels relatively simple.
2Reliability
If multiple evaluation channels are used for cross-comparison, then error detection capability is improved, but device complexity increases
Solution Approach 1:
Instead of making all evaluation channels fully independent and complex, the system applies local quality by having channels differ in specific aspects (different AI models, different sensor types) while sharing common infrastructure. This allows error detection through cross-comparison while controlling overall system complexity.
Solution Approach 2:
The evaluation channels are designed with multi-functionality, where each channel can perform person detection using different approaches (different AI models or sensor types). This universal capability allows the system to achieve error detection through cross-comparison without requiring completely separate specialized systems for each detection method.
3Measurement precision
If plausibility checks with multiple criteria are implemented, then diagnostic accuracy is improved, but processing time increases
Solution Approach 1:
The diagnostic unit performs plausibility checks using multiple criteria (person position, number of persons, consistency across channels) but applies these checks selectively rather than exhaustively to all possible parameters. This partial action approach maintains diagnostic accuracy while limiting the time penalty to acceptable levels for functional safety applications.
Data Source
AI summary
A surveillance system having a person detector includes an imaging sensor for recording digital image files or point clouds and includes an evaluation unit that is configured to evaluate the image files or the point clouds and to output position information of a person detected in a surveillance region, wherein a diagnostics unit has a first input channel for receiving the position information, a first output for checked position information of the detected person, and has a second output for an error signal, where the diagnostics unit is provide with a checking unit for checking the output of the position information of the detected person for plausibility with a plurality of plausibility criteria and is further configured to detect an error and output the error signal at the second output, if at least one plausibility criterion is violated.