Evaluation of Vision Systems for Outdoor Protection of Humans and Domestic Animals

According to the new Machinery Regulation (EU) 2023/1230, autonomous mobile machinery and related products must comply with specific safety requirements designed to prevent collisions or other risks due to their movement (travelling) or hazardous moving parts. In detail: "it shall move and operate in an enclosed zone fitted with a peripheral protection system comprising guards or protective devices and/or (depending on the risk assessment) it shall be equipped with devices intended to detect any human, domestic animal or any other obstacle in its vicinity."
A promising approach to meet this requirement involves integrating advanced sensing technologies, such as vision systems, capable of detecting objects and living beings within the machine's trajectory during autonomous motion. Camera-based systems, already widely used in indoor applications such as Automated Guided Vehicles (AGVs) in warehouses, are potential candidates for outdoor autonomous operations as well.
This paper explores the useful state of the art for the investigation of the performance of cameras, evaluating their usability, reliability and effectiveness in external working environments. Indeed, outdoor conditions introduce significant challenges, including varying weather, light exposure, and surface soiling, that can degrade the functionality and accuracy of these sensing devices. This study synthesizes key requirements from diverse standards regarding AI, industrial trucks, autonomous and road vehicles to establish a framework for the definition of a vision system evaluation methodology.
The overarching goal is to support the design of safer autonomous systems for working environments by exploring robust perception solutions suitable for real construction sites.

Keywords: Machinery Regulation, Safety of machinery, Vision systems, Autonomous motion, Artificial intelligence.