A Line Matching Method for Wide Baseline Image under Perspective Transformation Model

Abstract

The invention discloses a wide baseline image line matching method under a perspective transformation model. The steps include: 1) taking the geoid as the object plane to find out the perspective transformation relationship between the camera image plane and the object plane, That is, the perspective transformation matrix; 2) according to the perspective transformation matrix, the original image is projected onto the object plane to obtain the orthographic image after eliminating the perspective deformation effect of the original image; 3) the obtained orthographic image is Line segment matching; then inversely transform the obtained matching line segments to the original image through the perspective transformation matrix, and obtain matching line segment pairs in the original image. The number and accuracy of the line segments matched by the present invention can be greatly improved, and at the same time, the efficiency is basically not lost.

Description

technical field [0001] The present invention relates to a wide-baseline image line matching method under a perspective transformation model, in particular to a wide-baseline image line segment matching method under the perspective transformation model. Background technique [0002] Line feature matching is a common algorithm for image matching. Line segment matching has important research significance in virtual reality (VR), machine navigation, and three-dimensional structure reconstruction of objects in remote sensing mapping and measurement, and line segment matching is also the first problem to be solved. Line segment features are common in both natural environment and man-made objects. For example, houses on the ground, solar panels of artificial satellites in space, ships in the sea, etc. all have many and complex line segment features. Now, high-precision and high-efficiency line segment matching algorithms are of great value for sequence analysis, scene matching, et...

AI Technical Summary

Problems solved by technology

[0003] At present, many studies have a good line segment matching effect when dealing with conventional short baseline and small viewing angle measurement images; however, when faced with wide baseline and large viewing angle images, the matching effect drops sharply or even fails

Meltzer uses local affine invariance similar to points for matching, but it is not good at dealing with scenes with non-coplanar and complex three-dimensional structures

Li uses two adjacent line segments to construct ray-point-ray and line-node-line structural descriptors, which is time-consuming in the pyramid construction stage

Wang et al. will use line segment groups to establish local descriptors to match affine distortions and non-coplanar surfaces, but this method is time-consuming, and at the same time, there may not be enough line segments in places where the texture is not rich. fail

The key reason why line segment matching of wide baseline images is difficult is that the large angle of view leads to a huge difference in angle of view between the two images to be matched. The degree of similarity matching is very low or even impossible to match, resulting in a decline in the matching effect of line segments

Images