Image matching method and system, and computer readable medium

Abstract

The invention provides a method for matching a first image and a second image. The method comprises the following steps: S101, obtaining a feature map of the first image; s102, obtaining a feature map of the second image; s103, obtaining a feature vector at a specific position on the feature map of the first image; s104, obtaining a feature difference map according to the feature map of the second image and the feature vector; and S105, according to the feature difference image, obtaining a matching position, matched with the specific position of the first image, on the second image.

Description

technical field [0001] The present invention relates generally to the field of image processing, and more particularly to methods and systems for matching images and computer readable media. Background technique [0002] Given two or more images taken of the same scene, the position and attitude of the camera when the images are taken are known, and by finding the key points matched across the images, the position of the point in the physical world can be obtained, so that the scene can be compared from the image The various dimensions required within can be measured, which can avoid manual measurement of physical scales on site. figure 1 It shows the images taken by two cameras of the same shelf in a supermarket. The angle positions of the two cameras are different, so the positions of the same commodity or key point in the two pictures will be different. It is necessary to match the point on the left to the matching point on the right by matching the map. [0003] There ...

AI Technical Summary

Problems solved by technology

For example, the method based on the similarity of local features, because the receptive field of local features is limited, if there are multiple similar structures on the image, the correct matching point cannot be accurately found

Cannot automatically learn to improve matching results from direct historical data

The above two types of methods cannot directly learn how to match from the discrete matching points manually marked, and cannot gradually improve the matching effect as the manually marked data increases.

Images