Enhancing Child Safety: Computer Vision-Based Accident Detection for Infants and Toddlers

Children, who are infants and toddlers are prone to accidents when not supervised. Children may wander off and get hurt, such as falling from high places, getting burns, and more. Falls and drops are a major cause of injuries among children in the United States (US), with over 2.8 million children treated in emergency rooms each year. Parents need to always supervise the children but finding a balance with work can be challenging. The proposed system uses closed-circuit television (CCTV) and machine learning to detect two common types of accidents: cuts, and burns. The proposed system identifies children and dangerous objects, tracks movements, captures frames, and estimate the risk. The system offers a comprehensive solution to children monitoring, reducing accidents, and providing parents with peace of mind. The results show that detection accuracies are 87%, 84% and 75% for scissors, knife and both items respectively. The project findings indicate that the detection of accident risks achieves a satisfactory level of accuracy. However, when multiple dangerous objects are present in the field of view, the accuracy experiences a slight decline.