Font Size: a A A

The Research On Autonomous Vehicle Traffic Detection Method Based On Machine Vision

Posted on:2014-04-29Degree:MasterType:Thesis
Country:ChinaCandidate:M LiuFull Text:PDF
GTID:2252330425959880Subject:Electronics and Communications Engineering
Abstract/Summary:PDF Full Text Request
With the rapid development of people’s living standards, the vehicles play moreand more important role in people’s life. At the same time, traffic accidents are in-creasing year by year as the number of non-professional drivers growing fast, so peo-ple take more care about automobile safety. According to the United Nations, thereare3000people died in traffic accidents every day. So traffic accidents have becomethe first victim in world, and China is one of the most serious countries. In order toreduce the traffic accidents, autonomous vehicle research becomes the emergencywork. Autonomous vehicle using sensors to sense surrounding environment, and co n-trol the Steering and the speed of the vehicle to keep the vehicle more secure based onthe road information.This article do some research about the difficult problems on traffic signs recog-nition and vehicle tracking based on autonomous vehicles.This paper proposes a novel recognition algorithms about traffic signs that trafficsigns recognition using ICA-based affine invariant moments features. The obtainedtraffic signs maybe effected by affine transformation because of the road enviro n-mental factors, and the traffic signs shape are independent of each other beforetransformation in x direction and y direction. First, we can get tr affic sign’s shape thatonly effected by scale, rotating and mirror transform using independent componentsanalysis. Then recognizing traffic signs by compare the invariant moment features. Atlast, experiment by Hu moment and Zernike moment, as the result show that therecognition efficiency greatly improved, and using Zernike moment will much better.This paper also design a vehicle tracking system using particle filter trackingmethod based on color histogram feature. First, we introduced HOG features andAdaBoost classification to detect vehicle. Then described particle filter and featureextract. That extract the RGB color histogram as features, and compare features byBhattacharyya distance. At last, predicting the target by the different particles weights.We designed a vehicle tracking system based on the MFC and the opencv library, e x-periments show that tracking efficiency is good.
Keywords/Search Tags:autonomous vehicle, traffic sign recognition, independent comp o-nent analysis, moment features, vehicle tracking, particle filter, color histogram
PDF Full Text Request
Related items