Fast Ellipse Detection Algorithm Based On Multiple Decapitation

In the field of computer vision,detecting and identifying geometric shapes has always been an essential basic task,and circles are one of the most basic geometric shapes in nature.After perspective projection onto a two-dimensional plane,the circle will assume the shape of an ellipse.Therefore,the ellipse detection is widely used in detecting cells,irises,and round workpieces.Fast and accurate detection of ellipses in images has broad application prospects and great research value.Aiming at the problems of high error rate of arc segment extraction,low precision of ellipse detection and slow detection speed in traditional ellipse detection algorithms,this thesis proposes a fast ellipse detection algorithm based on multiple decapitation.The algorithm first performs image enhancement and edge detection on the input image to extract edge information,then uses the eight-neighbor tracking method to extract arcs,and then fits the arcs to the initial ellipse through the ellipse detection method based on multiple decapitation and removes the error detection.The main research contents and work of this article are as follows:(1)In the application scene of ellipse detection,the input image is often unclear.This thesis proposes a preprocessing method based on image enhancement.First,the image is smoothed and sharpened by operations based on points and neighborhoods.Then filter and denoise by frequency domain method to enhance the image while protecting the target edge features.Through the preprocessing step,the local features of the image can be enhanced in a targeted manner,and the differences between different features can be enlarged,so that the edges are better detected.(2)As the important information of the ellipse,the arc segment directly affects the accuracy and speed of ellipse detection.Aiming at the problem of low accuracy of the traditional arc extraction method,this thesis proposes a method of arc edge generation based on edge tracking.First,the Canny operator based on adaptive threshold is used to extract edge information from the image,and then the eight-neighbor tracking method is used to connect edge points from the edge image to form an arc.Finally,the arc segment anti-false method based on the bow-string ratio is used to filter and filter the arc segments,which can significantly remove the pseudo arcs.Through the arc segment generation method in this paper,the arc segments belonging to the ellipse can be accurately extracted from the input image.(3)Aiming at the problem of high error rate in fitting ellipse by traditional ellipse detection algorithm,this thesis designs an ellipse detection method based on multiple anti-false.First,the arcs are classified into arc triads by using the constraint method based on gradient,bump and position information,and then combined into arc triads,and then the arc triads are fitted into the initial ellipse based on the combination of geometric properties and Hough changes.Finally,a multiple anti-counterfeiting method based on the combination of local and global is used to anti-counterfeit the initial ellipse set to remove the error detection.Through the ellipse detection method with multiple anti-counterfeiting,the target ellipse can be quickly fitted from the arc,which improves the accuracy of the algorithm and reduces the calculation time.(4)This thesis cross-contrasts the proposed algorithm with other ellipse detection algorithms on public data sets.Firstly,the influence of the main threshold on the algorithm accuracy and time in this paper is analyzed,and the threshold value is determined;Secondly,the accuracy of the arc extraction method proposed by the arc generation method proposed in this paper is verified;Then,the ellipse local and global anti-false for the algorithm The influence of accuracy and speed demonstrates the ability of this method to effectively remove false ellipses;Finally,cross-comparison experiments are conducted to verify the accuracy and rapidity of the algorithm proposed in this paper.