A method for retrieving suspicious vehicles based on deep car sticker detection

Abstract

The invention discloses a suspicious vehicle retrieval method based on deep vehicle sticker detection, belonging to the fields of intelligent transportation and image retrieval. The method includes: performing window detection on the bayonet vehicle image, obtaining the vehicle sticker area of ​​the vehicle window in the bayonet vehicle image; performing vehicle sticker detection on the vehicle sticker area, and the vehicle sticker detection includes the vehicle sticker information to be detected Compare with the car sticker information in the car sticker database to get the vehicle retrieval result; the car sticker information in the car sticker database is pre-stored and continuously supplemented during the detection process. The present invention retrieves the vehicle by selecting the sticker on the front window of the suspected vehicle as a unique feature of the vehicle, and can accurately and effectively retrieve the vehicle in the database through training and using a deep network model to meet the actual situation. Vehicle retrieval needs.

Description

technical field [0001] The invention relates to a suspicious vehicle retrieval method based on deep car sticker detection, belonging to the fields of intelligent transportation and image retrieval. Background technique [0002] At present, the vehicle retrieval system mainly retrieves the vehicle based on the vehicle's inherent license plate number, vehicle brand and vehicle color in the video or bayonet picture. [0003] However, in the illegal cases involving vehicles in real life, the vehicles involved are often fake (set) license plates, and the vehicle brand logo and vehicle color may even be changed. At this time, investigation based on the vehicle brand, color and license plate number will not be effective At the same time, retrieving suspicious vehicles based on vehicle make and vehicle color plays a limited role in reducing the workload. Therefore, the retrieval method based on these inherent attributes of vehicles cannot achieve the expected retrieval effect on fa...

AI Technical Summary

Problems solved by technology

[0003] However, in the illegal cases involving vehicles in real life, the vehicles involved are often fake (set) license plates, and the vehicle brand logo and vehicle color may even be changed. At this time, investigation based on the vehicle brand, color and license plate number will not be effective Any effect; meanwhile, retrieval of suspicious vehicles based on vehicle make and vehicle color is of limited use in reducing workload

Therefore, the retrieval method based on the inherent attributes of these vehicles cannot achieve the expected retrieval effect on fake (set) license plate vehicles

Images