A method for identifying the axle shape of green traffic vehicles based on convolutional neural network

Abstract

A method for identifying the axle type of green traffic vehicles based on a convolutional neural network, comprising the following steps: step 1, obtaining green traffic images; step 2, formulating criteria for judging the validity of green traffic images; step 3, using data enhancement methods to increase training The number of samples; step 4, after data enhancement, use the target detection algorithm YOLOv2 framework to detect the whole axle; step 5, classify the vehicle axle type according to the vehicle axle group type and wheel group type; step 6, use AlexNet, VGG‑ 16. ResNet-152 three kinds of convolutional neural networks are trained to classify axle types on the training set; step 7, determine the axle type of green traffic vehicles that need to be identified. The invention aims to realize the accurate recognition of the vehicle axle type in green traffic, and combines the target detection algorithm, unbalanced data set processing, etc. with the convolutional neural network model, and the images selected by the judgment standard can be used as the convolutional neural network for subsequent classification and recognition experiments The training samples of the model avoid the problem of poor training samples.

Description

technical field [0001] The invention belongs to the technical field of vehicle identification, and in particular relates to a method for identifying the axle type of green traffic vehicles based on a convolutional neural network. Background technique [0002] The image classification algorithm is divided into traditional classification algorithms according to time, including: K-neighbor algorithm, SVM support vector machine, Bayesian algorithm, etc. With the great improvement of computer computing power, deep learning algorithm has gradually become the current mainstream application. The artificial neural network simulates the working principle of neurons in the brain, and finally obtains a network model that can learn autonomously. Convolutional neural network models mainly involve AlexNet network models, VGGNet network models, and ResNet network models. The target detection algorithm based on traditional image processing is insufficient in terms of data processing capabi...

AI Technical Summary

Problems solved by technology

First of all, most studies have not established classification standards for vehicles in specific scenarios, and they all rely on classification standard documents formulated by the state, which makes the application scenarios have huge limitations

Images