Vehicle brand recognition method and system based on vehicle headlight image features

Abstract

The present invention relates to a vehicle brand recognition method and system based on the image features of vehicle headlights. The method comprises the steps of: acquiring a vehicle image collected by an image acquisition device, locating the vehicle headlight in the vehicle image, and obtaining the vehicle front Lamp area; extract the target feature vector of the headlight area, and identify the current headlight category according to the target feature vector and a pre-constructed multi-classifier; according to the pre-established relationship between the headlight category and the vehicle brand, A vehicle make associated with the current headlight category is identified. Adopting the scheme of the present invention can overcome the problem of high computational complexity caused by the large size of the front face area of ​​the car, and can meet the real-time application requirements; and can overcome the problem of low recognition accuracy caused by the small size of the vehicle logo problem, which increases the feasibility of large-scale practical applications, removes redundant information in the front face area of ​​the car, and improves the efficiency of vehicle brand recognition.

Description

technical field [0001] The invention relates to the technical field of monitoring, in particular to a vehicle brand recognition method and system based on vehicle headlight image features. Background technique [0002] In order to achieve traffic management, such as traffic flow investigation, traffic violation evidence collection, or fully automatic toll collection in expressways and parking lots, the technology of classifying and identifying vehicles has become more and more important. Especially in the construction of smart cities, the research on vehicle objects in road checkpoints has gradually become a hot spot in academia and industry. [0003] The identification of the vehicle brand is an important part of the extraction of the structured information of the entire vehicle. At present, there are two main ways to complete the vehicle brand identification. The first type: identify the front face area of ​​the vehicle; the second type: Extract and recognize the vehicle ...

AI Technical Summary

Problems solved by technology

The main problems of this type of method are: the size of the front face area is large, if more complex features are used to input into the classifier, it will lead to large computational complexity and low recognition efficiency; if relatively simple features are used to input into the classifier For different brands, there will be a problem that the feature distances of different brands are relatively similar, resulting in a decrease in recognition accuracy. At the same time, there is too much redundant information in the front face area of ​​the vehicle, which also makes the efficiency of the recognition algorithm low.

In the second method, the idea similar to the first method is still adopted. However, due to the small size of the vehicle logo, in the collected images, the target size corresponding to the vehicle logo is also small, and the recognition accuracy is low. In the actual bayonet application, the recognition rate cannot meet the requirements of large-scale application

Images