A vehicle model retrieval system and method based on convolutional neural network

Abstract

The invention discloses a convolutional neural network-based vehicle model search system and method. The search system comprises a traffic block port picture obtaining module, a vehicle detection module, oriented feature segmentation module, an oriented feature extraction module and a search module, which are connected in sequence, wherein the vehicle detection module is used for carrying vehicle detection on an obtained vehicle photo and capturing a vehicle picture; and the oriented feature segmentation module is used for positioning an air-inlet grille on the basis of the vehicle picture captured by the vehicle detection module, and respectively segmenting a vehicle logo picture, a left vehicle light picture, a right vehicle light picture, an air inlet grille picture and a bumper picture via a symmetry axis of the air inlet grille. The search method has the beneficial effects that the search efficiency and correctness can be greatly improved, the search time is obviously shortened when the search method is compared with the traditional search methods, the correctness can be up to 95.7%, and the requirements of intelligent traffic systems can be better satisfied.

Description

technical field [0001] The invention relates to the technical field of intelligent traffic information, in particular to a vehicle model retrieval system and method based on a convolutional neural network. Background technique [0002] In recent years, with the continuous progress of society and rapid economic development, automobiles have become an indispensable means of transportation in daily life. Under such circumstances, intelligent transportation has emerged and developed rapidly. The breakthrough of convolutional neural network in image processing provides a more effective way for the application of intelligent transportation system. Convolutional neural network is to use computer to simulate the human brain for analysis and learning, extract effective features and textures from objective images, and finally use them for retrieval and recognition. [0003] In intelligent transportation systems, vehicle model retrieval is an extremely important part. Vehicle model i...

AI Technical Summary

Problems solved by technology

[0004] In today's era when the number of vehicles is so large, each city generates tens of millions of traffic information data every day, and traditional retrieval methods such as BOG, CSS, CSR, etc., are directly comparing data in huge databases , which causes a lot of unnecessary calculations, which greatly affects the efficiency of retrieval and affects the accuracy of retrieval. However, this method solves the problem of blind calculation, thereby further improving efficiency and accuracy. Rate

Images