Multi-view vehicle re-identification model and method in vehicle-road cooperation scene

Abstract

The invention discloses a multi-view vehicle re-identification model and method in a vehicle-road cooperation scene, establishes a simulation data set V2I-CARLA based on the vehicle-road cooperation scene, and aims to solve the problems of large vehicle target scale difference, incapability of obtaining detail features and the like in visual information collected by a fixed platform and a mobile platform. A feature pyramid attention module (FPA) is introduced into the feature extraction module, multi-scale features of the vehicle are obtained, and the completeness of the captured vehicle features is improved. In order to solve the problem that the discrimination degree of similar detail features is insufficient, circle loss is applied to a vehicle re-identification task, the algorithm discrimination degree is enhanced, and the vehicle matching effect is improved. According to the method, high matching precision is realized in a vehicle-road collaborative simulation scene, and meanwhile, a good matching effect is also achieved on a public vehicle re-identification data set.

Description

technical field [0001] The invention belongs to the technical field of intelligent networked vehicles, and particularly designs a multi-view vehicle re-identification model and method based on vehicle-road coordination. Background technique [0002] Vehicle-road collaborative perception is an effective direction for smart cars to develop towards higher-level autonomous driving in the future. Based on vehicle-road collaboration, it can not only improve the intelligence level of bicycles, but also realize the intelligent development of roads. In the process of autonomous driving based on vehicle-road coordination, not only the self-vehicle can complete the perception of the surrounding environment, but also the road status around it or even a larger range, and the real-time dynamics of various traffic participants in advance. Based on this, scientific and reasonable decisions can be made. decision-making planning and control, thereby improving driving safety and overall traff...

AI Technical Summary

Problems solved by technology

The change of the vehicle's pose and the difference in scale under the multi-view camera make it difficult for the vehicle's detailed features to be captured by the visual sensor

In recent years, the vehicle re-identification scheme mainly focuses on the global characteristics of the vehicle's appearance, ignoring the attention to the details of the vehicle structure, resulting in more misidentification of vehicles with similar models, and the reliability of the matching effect is low.

[0006] At the same time, the existing public data sets based on vehicle re-identification are small in number and low in quality, and most of the existing public data sets are shot under one or more fixed roadside cameras, which do not meet the scene conditions of vehicle-road collaborative perception

Images