Fast surveillance site image vehicle retrieval method and system based on deep learning

Abstract

The invention relates to a fast surveillance site image vehicle retrieval method and system based on deep learning. According to the method, a deep neural network is adopted to extract vehicle feature information, vehicle features are extracted on the basis of the inception_resnet_v2 network, network weight sharing is realized, a large amount of repeated calculation is effectively avoided, a lossfunction thereof is a triplet loss function, and is adopted to carry out triplet sample training, and 128-dimension vectors are directly generated. In an image retrieval stage, the method adopts, toestablish an index for the features, a manner of clustering the features, and increases a query speed. According to the method, an extraction speed of image features can be accelerated, and partial illegal vehicles of fake-plate vehicles and clone vehicles can be effectively inspected, traced and arrested while fast real-time responding is carried out.

Description

technical field [0001] The present invention relates to the field of intelligent transportation, in particular to a method and system for fast retrieval of bayonet image vehicles based on deep learning. Background technique [0002] In the field of intelligent transportation, intelligent traffic monitoring is a very important development direction at present. At present, my country has deployed a large number of electronic police and checkpoint systems on urban roads. These systems can capture high-definition pictures of vehicles in real time, and identify and analyze the license plate number, as well as some vehicle model information (such as vehicle size, color, etc.). However, in the currently used bayonet monitoring system, the license plate number recognition still has a misrecognition and missed recognition rate of about 10%. More importantly, it will not be possible to identify illegal vehicles with license plates or deliberately concealed photos. Therefore, by usin...

AI Technical Summary

Problems solved by technology

[0006] Among the current vehicle retrieval methods, one is to use traditional machine learning to extract different feature vectors for identification, and the other is to use deep learning neural networks for feature extraction. Usually, traditional machine learning algorithms are not as accurate as deep learning; most of the current methods It is to use the classification model to train the network and then extract the feature layer for identification. In this way, the feature vector of the vehicle is generally large. Since the vehicle features need to be compared with all the vehicle features in the database, the bayonet pictures stored in the traffic system of the big cities often have more than The number of pictures in the billion level, this retrieval method obviously cannot meet the actual situation. In addition, the deeper the neural network layer, the higher the recognition accuracy rate is generally. However, its training is also more difficult.

Images