In the process of industrial inspection in safety-critical environments, there is a fundamental structural tension in the balance of inspection, operations and human presence. In general, conventional inspection approaches are static, periodic or alarm-based, but they do not readily cope with the dynamic nature of risk conditions, which are subject to variability in production, congestion and emerging trends in inspection data. Therefore, the allocation of inspection resources is inefficient, either in under-inspecting critical areas or in exposing humans to unnecessarily hazardous environments. In this thesis, a risk-based inspection supervision approach is proposed, in which the process of inspection is clearly decomposed into decision, execution and interpretation within a unified system architecture.