Method for carrying out fast vehicle comparison and vehicle type recognition at tollgate

Abstract

The invention relates to a method for carrying out fast vehicle comparison and vehicle type recognition at a tollgate. The method comprises the steps that firstly, a standard snapshot vehicle image is collected through a high-definition tollgate vidicon, the foreground vehicle detection is carried out on the image, and a vehicle area is extracted; the vehicle area is preprocessed, feature points are detected, a sift feature descriptor is calculated, a query tree carries out the inquiry, the rough vehicle type matching is achieved, a vehicle type data image set is obtained, accurate matching is carried out on standby vehicle type images through the sift feature descriptor again, geometric verification is carried out through geometric information, and a final vehicle type comparison result is obtained through rearrangement and is output. According to the method, the multi-scale sift feature serves as the descriptor of the image, the sift feature is the local feature of the image, invariance of image rotating, dimension zooming and luminance changing is kept, and certain stability of vision angle changing, radiation converting and noise is kept.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and in particular relates to a rapid bayonet vehicle comparison vehicle type recognition method. Background technique [0002] Intelligent Transportation System (Intelligent Transportation System, ITS) is a subject of widespread concern in the field of transportation, and vehicle identification is the basis and a key component of ITS applications. Car recognition generally includes car license plate recognition, color recognition, vehicle model recognition, car logo recognition, etc., among which the most mature one is license plate recognition. However, in the detection of public security cases, especially in the tracking of suspects, relying solely on license plate recognition, color recognition or car logo recognition is not reliable, and there are many interferences, such as decking, blocking, color cast, and changing labels; Recognition technology is particularly important, and it...

AI Technical Summary

Problems solved by technology

The advantage of this method is that the BP neural network has good generalization and fast operation speed, but the disadvantage is that the learning process is easy to fall into the local extreme point of the error function. If the initial weight coefficient is not set properly, the learning process will converge slowly. not even convergent etc.

Compared with the traditional neural network method, the car model recognition method of support vector machine has the advantages of excellent performance and simple structure. Classify vehicles, combine SVM with nearest neighbor method to improve SVM classification speed, and combine dynamic Boosting algorithm to improve classification accuracy, but the disadvantage is that the calculation amount is large, the required hardware facilities are relatively complex, and the usability is relatively high. Difference

The vehicle model recognition based on feature matching is accurate in most weather conditions, with a small amount of calculation, good robustness, and fast recognition speed, but its algorithm is easily affected by the recognition rate in complex environments

Images