The stove is one of the most frequent sources of fire accidents in the home, many of which are caused by human error or forgetfulness, a problem which may become more serious with advanced age. An automated stove-top monitoring system could significantly increase kitchen safety. This study develops such a system, which uses a thermal camera to detect dangerous situations and behaviours, and alerts the user before a fire occurs. It is hoped that this system will serve to promote independent living among the elderly, leading to increased quality of life and decreased health care costs. The stove-top monitoring system consists of four subsystems: burner status (active/inactive), burner temperature trend, pot presence/absence, and human activity detection. Twenty-two experiments were conducted using ceramic, electric coil, and gas stove tops. Rule-based algorithms were developed to combine the outputs of the four subsystems, and to alert the user or caregiver when a dangerous situation occurs. Excellent performance was achieved for alert generation (sensitivity = 94%, positive predictive value = 83%). Furthermore, no modifications are required to the stove top, allowing this system to be retrofitted on any stove top.