Foreground trajectory determination method and device based on feedback

Abstract

The invention discloses a foreground trajectory determination method and device based on feedback. The method comprises the following steps: extracting a first predetermined number of feature points from each image frame, and generating feature point trajectories by the match between image frames; screening a feature point trajectory set in which the number of continuously tracked image frames is greater than or equal to a second predetermined number and a foreground trajectory time is determined to be less than a third predetermined number from the feature point trajectories; for the feature point trajectory set, calculating a homography matrix between a current frame and an adjacent frame staring from the first frame, and re-projecting coordinates of each feature point in the feature point trajectory set in the adjacent frame by using the homography matrix; and determining, after the second predetermined number of re-projections, whether the feature point trajectories are determined foreground trajectories according to the re-projection results. By adopting the foreground trajectory determination method and device disclosed by the invention, the feature point trajectories belonging to a foreground object and subjecting to parallax interference can be determined and are determined as the foreground trajectories.

Description

technical field [0001] The invention relates to the technical fields of digital image processing and intelligent transportation, in particular to a method and device for determining a foreground trajectory for eliminating video shake. Background technique [0002] In recent years, with the vigorous development of the automobile industry, the application of traffic security monitoring equipment has become more and more popular, and the monitoring video effect brought by it has also attracted more and more attention. Mobile monitoring devices such as vehicles or handhelds inevitably shake when shooting, and the jittering monitoring video seriously affects the viewing effect and interferes with judging the driving behavior of the vehicle. Therefore, eliminating video jitter has become an urgent problem to be solved in the fields of video surveillance and intelligent transportation. [0003] Traffic videos usually contain two parts: fast-moving vehicles and relatively fixed bac...

AI Technical Summary

Problems solved by technology

However, this method cannot handle non-plane motion, and is sensitive to parallax, such as obvious changes in depth of field, and the 2D method is completely invalid.

Although the 3D method fundamentally solves the influence of parallax, it will cause overprocessing when encountering obvious plane motion, and the SFM method is complicated and time-consuming

However, this method does not actively determine whether a feature point or a feature point trajectory belongs to the background or foreground, but only to calculate a set of reasonable inter-frame homography matrix, and, due to the defects of the inter-frame homography matrix model itself, So this method will fail like other 2D algorithms when facing large parallax

Therefore, this method cannot be directly used for deshaking traffic videos that often contain rapid depth changes.

Images