A Task-Agnostic Algebraic Integrity Metric for Event-Camera Streams Toward SOTIF-Compliant Perception using Pearson Correlation Coefficient
Researchers have developed a new task-agnostic metric to assess the integrity of event camera data streams, crucial for safety-critical perception in automated driving systems. This metric, based on the Pearson Correlation Coefficient, can be applied directly to asynchronous event streams without needing downstream task performance data. The proposed framework yields three specific metrics designed for stream integrity monitoring, adaptive region-of-interest selection, and temporal redundancy gating, addressing a gap identified in recent benchmarks. AI
IMPACT Establishes a new standard for evaluating sensor data integrity, potentially improving the safety and reliability of AI-driven perception systems in autonomous vehicles.