Real-time video stitching method based on seam line

Abstract

The invention discloses a real-time video stitching method based on a seam line. The method comprises the steps that the first frames of two monitoring videos are read respectively as a template frame; feature points on the template frames are extracted, and feature point matching is carried out to initially screen matching point sets; an RANSAC (random sampling consistency) algorithm is used to screen out a reliable matching point set to calculate a transformation matrix; the transformation matrix is used to register two frames of images; an overlapping area, non-overlapped areas and the sizeof a final complete image are calculated according to coordinates before and after corner point transformation; the seam line is calculated according to the color intensity and geometric structure ofthe overlapping area; the overlapping area is stitched; non-overlapped areas are spliced; a spliced complete image is calculated; and subsequent frames of the videos are spliced through the acquiredtransformation matrix and the seam line, and a CUDA platform is used to accelerate the process in parallel.

Description

technical field [0001] The invention relates to the fields of computer graphics, multimedia information technology, video processing and the like, and in particular to a real-time video splicing method based on sutures. Background technique [0002] With the development of society, people enjoy the convenience of modern society, but it also brings a series of security risks. In order to eliminate hidden dangers and protect people's lives and properties, surveillance cameras all over the city have become a powerful weapon. However, roads, administrative halls and other occasions require multiple cameras to monitor at the same time, causing security personnel to pay attention to multiple surveillance videos at the same time. In addition, multiple surveillance videos are not conducive to subsequent video processing and analysis. In order to effectively solve the above problems, it is necessary to splice multiple surveillance videos into one complete video. [0003] The splici...

AI Technical Summary

Problems solved by technology

[0004] Existing image stitching methods have high requirements for input images, requiring overlapping areas of images to be on the same plane, or even requiring the entire image to be on the same plane; in addition, overlapping areas cannot have obvious depth changes

Once the conditions are not met, ghosting and blurring will occur in overlapping areas due to parallax

Blurring in the monitoring video will have a bad impact on the monitoring and is not conducive to the elimination of potential safety hazards; in addition, the splicing speed of the existing method is not satisfactory, and real-time video monitoring cannot be achieved, so that security personnel cannot eliminate potential safety hazards in time

Images