Heterogeneous twin region selection network and image matching method based on same

Abstract

The invention discloses a heterogeneous twin region selection network and an image matching method based on the same, and belongs to the field of computer vision. The system comprises a heterogeneous twin network and a regional matching network which are sequentially connected in series, wherein the heterogeneous twin network is used for extracting a feature map of a template map and a feature map of a to-be-matched map, the region matching network is used for obtaining a region matching result according to the feature map of the template map and the feature map of the to-be-matched map, the heterogeneous twin network comprises a sub-network A and a sub-network B which are connected in parallel, each sub-network comprises a feature extraction module, a feature fusion module and a maximum pooling module which are sequentially connected in series, the two sub-networks are the same in modules, and only convolution kernels of first-layer convolution of the feature extraction modules are different. The heterogeneous twin region selection network is applied to image matching, non-fixed-scale template and to-be-matched image input is achieved, image multi-layer characteristics are fully utilized, the performance of the matching method is effectively improved, and the matching success rate and speed are increased.

Description

technical field [0001] The invention belongs to the field of computer vision, and more specifically relates to a heterogeneous twin area selection network and an image matching method based on the network. Background technique [0002] Image matching refers to the process of finding images or image regions (sub-images) similar to a given scene region image in one (or a batch) of images. Usually, the image of the known scene area is called a template image, and the subimage that may correspond to it in the image to be searched is called the scene area image to be matched of the template. Image matching is to find correspondence between two or more images of the same scene from different times or different perspectives. Specific applications include target or scene recognition, solving 3D structures in multiple images, stereo correspondence, and motion tracking. [0003] At present, most image matching algorithms only use shallow artificial features, such as grayscale feature...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the defects of the prior art, the purpose of the present invention is to solve the technical problems that the prior art image matching algorithm is basically unable to adapt to the imaging viewpoint and scale change, the scene adaptability and anti-interference ability are weak, and the matching success rate is low

Images