Abstract In safety monitoring, there exists an uncertainty situation in which the sensor cannot detect whether or not the monitored object is in danger. For the uncertainty zone identified by a non-homogeneous safety monitoring system that utilizes two types of sensors with different thresholds, operators or experts are expected to judge whether the real state is safe or dangerous on the basis of additional information from a detailed inspection or other related sensors output. However, the activities for inspection performed by relevant humans may require additional cost and introduce inspection errors. The present article proposes two types of an automatic monitoring system not involving any human inspection or a human–machine (H–M) cooperative monitoring system with inspection. In order to compare the systems, an approach based on the Dempster–Shafer theory is proposed as uncertainty analysis by this theory (it is simpler than by the traditional Bayesian approach). By comparing their expected losses as a result of failed dangerous failures or failed safe failures as well as the inspection errors, the condition is determined under which H–M cooperative systems incorporating human judgements are more effective than automatic monitoring systems.