Vehicle re-identification method and system

Abstract

The invention provides a vehicle re-identification method and system, and the method comprises the steps: carrying out the angle rotation according to a three-dimensional model of each vehicle model,obtaining a picture of each angle, extracting the key point information of each vehicle model and the outer contour information of each vehicle model, and calculating a rotation angle; inputting the key point information of each vehicle model, the outer contour information of the vehicle model and the rotation angle into a vehicle re-identification model to obtain a trained vehicle re-identification model; and inputting the picture or video of the to-be-tested vehicle into the trained vehicle re-identification model, and outputting the model of the to-be-tested vehicle.

Description

technical field [0001] The disclosure belongs to the field of computer vision, and in particular relates to a vehicle re-identification method and system. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] The widespread deployment of traffic cameras provides the possibility of video analysis for applications such as logistics, transportation, and smart cities. The key issue in this analysis is to correlate targets across cameras. Although both pedestrians and vehicles are common objects in smart city applications, there has been much attention paid to pedestrian re-identification in recent years. This is because there is a large amount of annotated pedestrian data in pedestrian re-identification, and computer vision research on human faces and bodies is relatively mature. Vehicle re-ID is more challenging than pedestrian re-ID. The speci...

AI Technical Summary

Problems solved by technology

[0006] Methods based on sensors or artificially designed features are relatively complex and have a low recognition rate; methods using multi-dimensional information are sensitive to the special appearance of vehicles, but are easily affected by changes in viewing angle and illumination; methods based on metric learning have a higher recognition rate , the recognition efficiency of difficult samples is also relatively good, but the training time is relatively long; some scholars use feature learning or distance metric learning to train deep neural networks, but the effect of such methods in vehicle re-identification is far less than that of pedestrians. identify

Images