The article considers a problem of detecting foreign objects on the runway. Existing automated systems that solve that issue are based on cameras and radar sensors. However, in those systems, cameras are used only for visual confirmation and are rarely used directly to perform detection. The use of video information for object detection will increase the degree of automation of such systems. The article proposes a method for detecting foreign objects in a video stream based on threshold segmentation. The method works with data from static cameras and can be used both in systems with only cameras and in complex systems as an optical component. The developed method processes a set of consecutive frames of the video stream to perform detection. Object detection is performed in two stages. At the first stage, background subtraction based on Gaussian mixture model is used. At the second stage, adaptive threshold segmentation is applied. The applicability of the developed method was investigated on the basis of experimental data. The dependences of the quality indicators of object detection on the method parameters were obtained. The advantages of the proposed approach are high speed of object detection, high accuracy of object positioning in the frame, adaptivity of segmentation threshold, as well as flexible adjustment of detection sensitivity.