The invention provides a non-contact type emotion detection method and a device, and relates to the technical field of emotion detection. The method comprises the following steps of collecting video information containing a to-be-detected human body and a heat map / hot point area graph of the to-be-detected human body through a camera and an infrared sensor; recognizing the video information and the heat map / hot point area graph, so as to obtain the human body feature data of the to-be-detected human body; performing deep learning, engine correlation and analysis on the human body feature data, so as to obtain the nervousness degree of the to-be-detected human body; outputting the result of the nervousness degree of the to-be-detected human body through the deep learning, engine correlation and analysis. The method has the advantage that by combining intelligent sensing technique, machine vision and deep learning, engine correlation and analysis, the human body feature data can be judged in short time according to the collected video and other data, the hazard index and psychological nervousness degree of the to-be-detected human body can be finally analyzed, and the effective reference is provided for a detector to predict and prevent the hazard.