基于OpenCV的椭圆特征识别算法改进

Improvement of Elliptic Feature Recognition Algorithm Based on Open CV

由于工业测量领域的需要,采用VC相机进行椭圆识别提取的应用越来越多。通常采取的做法是基于开源计算机视觉库OpenCV,进行特征识别与提取。但在实际应用场景,直接使用OpenCV库提取椭圆效果并不能达到预期。本次研究主要讨论现存的两个问题:其一,使用OpenCV提供的ellipse函数进行拟合椭圆时,会将矩形轮廓也进行拟合,这会对结果产生干扰;其二,在实际应用场景中,进行识别面积和点数过滤时,由于提取的轮廓并不完整,所以不能反应真实识别的面积和点数。本文针对这两个现存问题进行算法改进,以使工业领域的椭圆识别准确率更高。

Due to the needs of industrial measurement,there are more and more applications for elliptical recognition extraction using VC cameras.The usual approach is to perform feature recognition and extraction based on the open source computer vision library OpenCV.However,in the actual application scenario,directly extracting the ellipse effect using the OpenCV library does not meet expectations.This study mainly discusses two existing problems: First,when using the ellipse function provided by OpenCV to fit the ellipse,the outline of the rectangle will also be fitted,which will interfere with the result;second,in the actual application scenario.When the recognition area and the point number are filtered,since the extracted outline is not complete,the area and the number of points that are actually recognized cannot be reflected.In this paper,the algorithm is improved for these two existing problems,so that the accuracy of ellipse recognition in the industrial field is higher.