Research On Fisheye Image Distortion Based On SVM

Fisheye view can be taken about 180 degrees of the scene,can even reach 220 degrees,but the problem in the advantage is obvious,with the expansion of the angle of view,fisheye image appeared in various areas of image distortion is very serious,this limits the effect of fisheye lens.This paper carried out the research on path of the fisheye image distortion based on support vector machine,to lay the foundation for the distortion of fish eye image analysis.We mainly do the following work: corner detection of the target,fitting the curve of change path by support vector machine,images feature extraction and so on.The picture characteristics of fisheye lens shooting is expansion of intermediate region and edge region compression,according to the characteristics of fisheye lens shooting,the target is designed by the correspondence between the incident light and the image point.In this subject,Harris operator was used for target corner feature extraction,regions of varying degrees of distortion are distinguished using black contour lines on the target,the extracted corner coordinates are used as the training data of SVM algorithm,distribution from the corner detection target for fisheye structure is nonlinear coordinate,support vector classification can find optimal interval of the different classified data,Gauss radial basis kernel function(RBF)is selected to realize nonlinear mapping,solving highly nonlinear problems in the sample space make the sample linear,find support vector machine classification curve.The position of the image center is gained by Hough transform.The distortion of fish eye image,in different regions for different degrees of pixel loss,loss of information content is not the same,the PCA algorithm is used to process the image,which makes the image carry the main information,and the small amount of the interference component is discarded,input the main image information to support vector for error analysis of curve.Experiments Based on MATLAB show that the distortion of the path analysis base on the support vector machine,the forecast data to the actual data has a good ability to follow,support vector machine analysis the distortion curve,it has strong adaptability to complex graphics pixel distortion.