Convolutional neural network generation method and vehicle series identification method

Abstract

The invention discloses a convolutional neural network generation method used for vehicle series identification of a vehicle in an image, a vehicle series identification method, a computing device anda mobile terminal. The convolutional neural network generation method comprises the steps that a first processing block is constructed, wherein the first processing block comprises a first convolution layer; a second processing block is constructed, wherein the second processing block comprises a second convolution layer and a third convolution layer and the second convolution layer and the thirdconvolution layer are sequentially connected; according to a number of first processing blocks, second processing block and pooling layers, a full connection layer and a classifier are combined to construct a convolutional neural network; the convolutional neural network uses the first processing block as input and the classifier as output; the convolutional neural network is trained according toa pre-acquired vehicle image data set, so that the output of the classifier indicates a vehicle series corresponding to the vehicle; the vehicle image data set comprises a lot of vehicle image information; and each piece of vehicle image information comprises the vehicle image and the vehicle series information of the vehicle in the corresponding image.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method for generating a convolutional neural network, a method for identifying a vehicle, a computing device, and a mobile terminal for identifying a vehicle in an image. Background technique [0002] With the rapid development of technology and economy, the types of cars on the market are becoming more and more abundant, such as the common Audi A4L, BMW 3 Series, etc., but in real life, you will often encounter cars that you don’t know or don’t know about. Vehicles. In order to be able to identify the car series of these vehicles, the mobile terminal is usually used to take pictures of the vehicles, and the pictures formed by the pictures are uploaded to the server based on cloud services. The server can use CNN (Convolutional Neural Network, Convolutional Neural Network) Based on the car series recognition method to identify the car series of the vehicle, an...

AI Technical Summary

Problems solved by technology

Due to the limited ability to describe artificially designed features, and the classifier can classify many types, the requirements for images are high, especially the shooting angle or classification types are limited, such as only for pictures and videos of the front or rear of the car. identify

This has caused application restrictions and cannot be used at will, especially in complex environments. Once the required photos cannot be taken, it cannot be recognized

Images