Research On Image Line Feature Extraction Algorithm And ARM Implementation

In an image,the line features(including straight lines,curves)tend to form the objects' contour.So if the line features of an imag can be extracted effectively,the application of image processing has an import significance in the areas of object matching,objet recognition,machine vision,data mining,and so on.This dissertation studies the line features extraction algorithms in digital with different characteristics.The main contents of this dissertation are as follows:Firstly,a circles and ellipses detection algorithm based on multiscale edge extraction is studied when existing curves feature(circles,ellipses)with non-smooth edge in an imsge.In order to improve the detection rate of curves features,the images are detected with different scales obtained by Gaussian pyramid,then the different scale images are detected respectively by this algorithm.And,this algorithm is validated by crater images,and the detection rate can reach above 85%.At last,the proposed algorithm is compared with the results of Fornaciari' algorithm.The detection results show thatthe detection rate is increased by 40%,and the detection time is improved by more than 50%.Then,with regard to the image with multiple line features,a line and ellipses extraction algorithm based on ELSD is studied.The line segments and elliptical arc segments are detected by the methods of region growing and curve growing respectively,then using the least square method to ellipses fitting with the information of elliptical arc segments' location and tangent location.The invalid ellipses are eliminated by calculating NFA.This algorithm is used to detect traffic sign images,and the experimental results show that this algorithm has good performance the image with multiple line features and can detect the image quickly.Finally,the circle(ellipse)detecting system based on ARM platform is designed.The whole system consists of hardware and softare.In hardware part,the i.MX6 DL processor is the core,with the combination of a series of peripheral interfaces;in the software part,this system combines OpenCV(cross platform computer vision library)and QT resources to implement image processing and display under the embedded system.At last,the program of traffic sign detection is ported to th system,and realizing the traffic sign detection based on ARM platform.