A new background update algorithm for airborne camera in dynamic background

摘要

The technology of freeway vehicle detection and tracking based on airborne camera has wide applications in freeway monitoring and management. However, the image background is always changing in the aerial video so that it poses great challenges for object detection and tracking. In order to solve this problem, the paper studies background updating algorithm on basis of ViBe which focuses on the shortcomings of fixed threshold in Vibe background updating algorithm. The paper improves a new ViBe self-adaptive threshold algorithm. Distance threshold can update itself adaptively with the change of the sample value in the background model, and background learning rate can also adaptively adjust the speed of the background update with the change of the noise. Our algorithm has a better adaptability to illumination mutation, and it can quickly adapt to the background change and get accurate background model, which is helpful to the subsequent segmentation and extraction of vehicle contour.