A fine identification method and system for a vehicle type

Abstract

The invention discloses a method and system for finely identifying vehicle types. The method includes: performing grayscale and standardization processing on an acquired original vehicle image to obtain a standardized image; calculating the gradient and direction of each pixel in the standardized image; The gradient and direction of the standardized image are extracted from the direction gradient histogram feature and local linear constraint coding to obtain the coding vector of the standardized image; according to the obtained coding vector, the weight space pyramid is used to process the standardized image after the local linear constraint coding, and the obtained The final expression vector of the vehicle image, the final expression vector of the vehicle image contains the position information and semantic information of the vehicle image; the final expression vector of the vehicle image is sent to a pre-trained linear support vector machine classifier for vehicle type identification. The invention has the advantages of high accuracy, low complexity, strong robustness and rich detail features, and can be widely used in the field of image processing.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method and system for finely identifying vehicle types. Background technique [0002] In recent years, with the increasing number of cars in our country, there are more and more car-related cases, car theft cases and various traffic violations involving the use of cars for illegal and criminal acts, and the work of vehicle management and identification has also received more and more attention. Much attention. Traditional security checkpoints, electronic police and other systems mainly obtain vehicle information by performing license plate recognition on quickly captured vehicle images. Although the current license plate recognition technology is relatively mature and popular, it is powerless in the face of license plate vehicles, blocked license plates and unlicensed vehicles. In addition, in various cases, if it is necessary to search for a certain model of vehicle based on ...

AI Technical Summary

Problems solved by technology

Moreover, most of the appearance-based discriminant methods use the SIFT feature extraction algorithm, and the SIFT feature extraction algorithm has poor description ability and slow extraction speed for rigid objects, which affects the subsequent accurate recognition. The appearance-based discriminant method usually needs to extract the vehicle's Specific areas (such as vehicle face areas, etc.), when the background of the vehicle image is complex, it will not be able to completely extract or even extract the specific area of ​​​​the vehicle, which will affect the accuracy of subsequent vehicle type fine recognition, and the robustness is low

In addition, most of the existing vehicle type fine-grained recognition methods are based on shallow features, which cannot capture the subtle differences between different vehicle types well, which will affect the accuracy of subsequent recognition.

Images