Object re-identification method based on different granularity feature matching consistency

Abstract

The invention discloses an object re-identification method based on different granularity feature matching consistency, and the method comprises the following steps: inputting an original image and ato-be-identified image of an object, preprocessing the to-be-identified image, and obtaining the to-be-identified image without background interference through projection transformation; extracting fine-grained features of the to-be-identified image and the original image through an angular point detection method, and matching a fine-grained feature matching result of the to-be-identified image and the original image through an angular point matching method; filtering the fine and shallow textures of the to-be-recognized image and the original image through an image algorithm, extracting the coarse-grained features of the to-be-recognized image and the original image through an angular point detection method, and obtaining a coarse-grained feature matching result of the to-be-recognized image and the original image through an angular point matching method; performing comparison based on the fine-grained feature matching result and the coarse-grained feature matching result to obtain anobject re-identification result. The method can effectively improve the matching precision of the random texture object, thereby improving the re-recognition speed and efficiency of the object.

Description

technical field [0001] The invention relates to the technical field of object re-identification, in particular to an object re-identification method based on matching consistency of features of different granularities. Background technique [0002] Image target re-recognition is also called target re-recognition, which is the process of re-matching the targets captured by different camera lenses at different times, different lighting, different shooting directions, and different backgrounds. The research on image target re-identification technology is mainly concentrated in the fields of pedestrian re-identification and vehicle re-identification, and there is less research on object re-identification. [0003] In the quality inspection and production activities of enterprises such as quality inspection and sample selection, through the re-identification of the target object, it is often a very important early stage in the business process to ensure that the selected object i...

AI Technical Summary

Problems solved by technology

For textured objects, similar results can be obtained through corner detection and corner matching technology on the target object image, but the texture similarity between objects is affected by the weak constraints of the pictures in different scenes, which is easily affected by background interference, It is difficult to ensure that the matching is unique and effective due to the influence of the direction angle, the size of the brick surface image and the degree of illumination.

Images