Model identification method, device and equipment based on convolutional neural network, and computer readable storage medium

Abstract

The invention discloses a model identification method, device and equipment based on the convolutional neural network, and a computer readable storage medium. The method comprises steps that an acquired to-be-detected picture is pre-processed; the to-be-detected picture after pre-processing is inputted to a first preset detection model to determine whether the to-be-detected picture contains the vehicle characteristic information; if the to-be-detected picture has the vehicle characteristic information, the to-be-detected picture after pre-processing is inputted to a second preset detection model; a probability value of the to-be-detected picture corresponding to each vehicle model is acquired through calculation through the second preset detection model; the largest probability value of all the probability values is determined, and a model corresponding to the largest probability value is taken as the model of the to-be-detected picture; the first preset detection model and the secondpreset detection model are respectively acquired through utilizing preset picture data to correspondingly train the convolutional neural network. The method is advantaged in that high-precision modelidentification can be realized, and the identification process is made to be efficient and stable.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a convolutional neural network-based vehicle identification method, device, equipment and computer-readable storage medium. Background technique [0002] Vehicle model identification plays a key role in many issues such as vehicle management, vehicle violation and escape, vehicle inspection and control, and accident car damage compensation. The vehicle type has the advantage of being difficult to change, which has become a very important feature in vehicle identification. In the case of license plate recognition and image clearing processing technology, it is impossible to obtain effective vehicle information, especially in the case of vehicle damage claims, the vehicle type has a huge impact on the amount of compensation, and vehicle type recognition is used in other similar traffic monitoring and control, It also plays a very important role in many scenarios such as t...

AI Technical Summary

Problems solved by technology

[0003] Traditional car model recognition methods can only design different features such as scale-invariant feature transformation (SIFT) for different tasks, and then use classifiers such as support vector machine (SVM) or random forest (Random Forest) for training. However, it can only be applied to the scene of vehicle type and license plate recognition. For as many as thousands of models and many types are very similar, it is difficult to manually design targeted features for recognition.

In addition, although there is a vehicle identification method based on multi-sensor fusion, the principle and identification of this method are very simple, and it has the disadvantages of being sensitive to the external environment and having a high failure rate.

Images