Rigid target tracking method based on fusion of SURF feature and color feature

Abstract

The invention relates to a rigid target tracking method based on fusion of an SURF feature and a color feature. The method comprises: (1), a target area in an initial image is selected and an SURF feature in the target area is extracted and SURF feature description is established; (2), a color feature is built in each local neighborhood by using an SURF feature point as a center; (3), when a current image arrives, a preliminary target area is searched by using the color feature, an SURF feature is extracted, feature-based matching with an initial image is established, and a matching point pair is formed; and (4), calculation is carried out based on the obtained matching point pair to obtain a motion parameter, so that the target area of the current image is determined and thus target tacking is realized. Therefore, accurate description and matching of a local feature of a target area can be realized, so that robustness and stability of a target tracking effect can be guaranteed.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a method for tracking a rigid body object by combining SURF features and color features. Background technique [0002] The movement of any point on the surface of a rigid target can represent the movement of the whole, which makes it possible to use the features in the target area to describe the movement of the target. Existing rigid-body target tracking methods are dedicated to extracting some invariant features in the target region of the reference image, and quantifying and describing the extracted features, such as color features, texture features, and optical flow features. Local features refer to the invariant, reproducible, and specific features detected in the image area, which can resist complex changes such as occlusion, scale, and rotation to a certain extent, and provide quantitative descriptions of features. At present, compared with other featu...

AI Technical Summary

Problems solved by technology

[0009] In the video, one or more changes such as illumination, occlusion, viewing angle, affine, etc. often occur in the scene, which seriously interferes with the matching of local features.

The existing technology still uses the same local feature matching method as the static image, which cannot adapt to the scene with drastic changes, nor does it reflect the adaptability corresponding to the continuous change of the scene

Images