Vehicle re-identification method, model training method and related device

Abstract

The invention provides a vehicle re-identification method, a model training method and a related device, and the method comprises the steps of inputting an image of a to-be-identified vehicle into a trained vehicle re-identification model, and obtaining a feature vector corresponding to the to-be-identified vehicle, wherein the vehicle re-identification model at least comprises an attitude information supervision sub-model, and the attitude information supervision sub-model is used for classifying the same vehicle with different attitudes according to the attitude information; and matching the feature vector with a plurality of comparison images to obtain a target image, and determining the to-be-identified vehicle as a vehicle in the target image, the target image being at least one comparison image in a comparison image library. According to the vehicle identification method provided by the invention, the target vehicle can be matched more comprehensively and accurately, so that the vehicle information of the to-be-identified vehicle can be determined.

Description

technical field [0001] The present invention relates to the technical field of vehicle re-identification, in particular to a vehicle re-identification method, a model training method and related devices. Background technique [0002] In recent years, smart cities and smart transportation have developed rapidly. Vehicles are an integral part of smart transportation, and vehicle re-identification is one of the core technologies of smart transportation. That is, given a queried vehicle image, vehicle re-ID aims to retrieve all images of the same vehicle from multiple non-overlapping cameras. [0003] Commonly used vehicle re-identification methods use models to extract vehicle features, and then perform vehicle feature matching to re-identify vehicles, but often cannot match vehicles with different angles and different attitudes, and the matching results are incomplete and inaccurate. Contents of the invention [0004] One of the objectives of the present invention is to pro...

AI Technical Summary

Problems solved by technology

[0003] The commonly used vehicle re-identification method uses the model to extract vehicle features, and then performs vehicle feature matching to re-identify the vehicle, but it often cannot match vehicles with different angles and different attitudes, and the matching results are incomplete and inaccurate.

Images