Airport scene monitoring video target segmentation method based on ADS-B position prior

Abstract

The invention discloses an airport scene monitoring video target segmentation method based on ADS-B position prior. Airport scene priori information ADS-B is applied to airport video target segmentation, and compared with a matching-based and propagation-based method, the method explores information from more previous frames in a reasoning process. According to the positioning information provided by the ADS-B, the content of the reference frame can be continuously updated, so that the reference frame is closer to the current frame. According to the scheme, the problem of space and time dislocation is solved, and the advantages of a propagation-based method and a matching-based method are combined, so that the network can continuously correct errors in the reasoning process, and the method has robustness for a long video.

Description

technical field [0001] The invention relates to the field of target recognition, in particular to an airport scene monitoring video target segmentation method based on ADS-B position prior. Background technique [0002] As an important technology of video processing, video object segmentation under airport scene monitoring plays a vital role in airport security. Given an aircraft of interest in the video, the video object segmentation algorithm continuously separates the object pixels of interest from the background until the object disappears or the video ends. This task plays an important role in airport surface surveillance systems. [0003] Divided from the setting, we can divide video object segmentation into unsupervised video object segmentation and semi-supervised video object segmentation. Among them, the unsupervised video object segmentation relies on the network to learn the semantic information in the video and automatically separate the foreground pixels from...

AI Technical Summary

Problems solved by technology

For the airport scene surveillance video, the aircraft only occupies a small part of the screen, and the appearance of the aircraft has a strong similarity with the background pixels, so the strategy based on mask matching will be greatly disturbed by the background

In addition, for the foreground pixels, there are usually large attitude, scale and position changes between the current frame and the first frame of the aircraft, which will cause the simple matching strategy to fail

[0007] To sum up, due to the particularity of airport ground monitoring, the current mainstream video object segmentation algorithms are not suitable for this scene

Images