基于改进ORB算法的油水两相流油泡匹配方法

Research on the Identification and Matching of Oil-water Two-phase Flow Oil Bubbles Based on Improved ORB Algorithm

ORB算法能够快速匹配油水两相流中的油泡。相比SIFT算法和SURF算法,ORB算法拥有无可比拟的速度优势。然而,ORB算法匹配油泡的过程中,存在油泡匹配准确率低、角点提取速度慢、角点数量少等缺陷,降低了油速测量的准确度。为提升油泡匹配准确率和油速测量准确度,对ORB算法进行如下改进:角点提取方面,通过Shi-tomasi角点响应算法替代Harris角点响应算法,加快角点提取速度;通过对角点之间最小距离d值的探索优化,确定最佳d值,使Os区域,即油水两相流图像中不明显油泡所在的区域,能有效提取到角点,增加角点数量,同时优化角点分布。油泡匹配方面,在RANSAC算法基础上,融入GMS算法,进行RANSAC-GMS双重去除误匹配对,提高油泡匹配准确率,最大限度保留正确匹配对,提升正确匹配对数量。实验结果表明:改进的ORB算法在提取角点耗时上缩短至原始ORB算法的37%;角点数量是原始ORB算法的1.9倍;正确匹配对数量是原始ORB算法的2.3倍。改进的ORB算法提高了油泡匹配准确率。最后,基于改进的ORB特征点匹配的PTV算法对油速进行测量,实验结果表明油速测量准确度达94.0%,提升26.5%,能够用于实际油速测量。

The ORB algorithm can quickly match the oil bubbles in the two-phase flow of oil and water.Compared with the SIFT algorithm and the SURF algorithm,the ORB algorithm has an incomparable speed advantage.However,in the process of matching oil bubbles,the ORB algorithm has disadvantages such as low bubble matching accuracy,slow corner extraction speed,and small number of corners,which reduces the accuracy of oil velocity measurement.In order to improve the accuracy of oil bubble matching and oil velocity measurement,the ORB algorithm was improved as follows:in terms of corner point extraction,the Shi-Tomasi corner response algorithm was used to replace the Harris corner response algorithm to speed up the corner extraction.By exploring the minimum distance d value between the diagonal points,the optimal d value is determined,so that the Os region,that is,the area where the inconspicuous oil bubble is located in the oil-water two-phase flow image,can effectively extract the corner points,increase the number of corner points,and optimize the corner distribution.In terms of bubble matching,on the basis of the RANSAC algorithm,the GMS algorithm is integrated to remove the mismatched pairs by RANSAC-GMS,so as to improve the accuracy of bubble matching,retain the correct matching pairs to the greatest extent,and increase the number of correct matching pairs.The improved ORB algorithm reduces the time consumption of corner extraction to 37% of the original ORB algorithm;the number of corners is 1.9 times that of the original ORB algorithm.The number of correct matching pairs is 1.3 times that of the original ORB algorithm.The improved ORB algorithm improves the accuracy of oil bubble matching.Finally,the PTV algorithm based on the improved ORB feature point matching is used to measure the oil velocity.The experimental results show that the accuracy of oil velocity measurement is 94.0% and increased by 26.5%,which can be used for actual oil velocity measurement.