The invention relates to the technical field of sleep detection, and discloses a sleep detection method based on a front-end camera, which comprises the following steps: 1) before a system is started, adjusting and fixing the position, angle and imaging range of the front-end camera to ensure that the front-end camera is not shielded and a detection picture is clear, and setting a human body detection area according to the position of a human body, the detection area is divided into small areas for the front-end camera to perform local detection; and 2) because the monitored human body is in a lying static state under the normal condition, the monitored human body area is generally set according to the size of the lying human body. According to the sleep detection method and device based on the front-end camera, sleep and non-sleep videos are collected through the camera, sleep and non-sleep features in the videos are extracted, the model is trained to identify sleep behaviors, and the sleep behaviors and non-sleep behaviors are modeled, so that the accuracy of a detection result is ensured, and false detection caused by light change can be avoided.