本文選題：最小平方中值　切入點(diǎn)：ORB算法　出處：《計(jì)算機(jī)應(yīng)用研究》2016年12期

【摘要】：針對(duì)傳統(tǒng)的特征提取算法在圖像匹配過(guò)程中易出現(xiàn)誤匹配現(xiàn)象,提出在ORB算法的基礎(chǔ)上融入一種最小平方上值估計(jì)法——LMedS方法。利用ORB算法的特點(diǎn)和LMedS方法去除可能存在的外點(diǎn),消除誤匹配現(xiàn)象,從而得到正確的匹配特征對(duì),使特征匹配率有很大的提高;同時(shí)采用基于非線性最小二乘進(jìn)行姿態(tài)估計(jì),通過(guò)迭代算法估算相機(jī)姿態(tài)完成虛實(shí)注冊(cè)。實(shí)驗(yàn)結(jié)果表明,該方法無(wú)論是在特征點(diǎn)匹配還是在實(shí)際場(chǎng)景中都具有很好的魯棒性,在不同尺度、部分遮擋的情況下同樣具有良好的性能,準(zhǔn)確、實(shí)時(shí)地完成了跟蹤注冊(cè)。
[Abstract]:In view of the fact that the traditional feature extraction algorithm is prone to mismatch in the process of image matching, a method of least square upper value estimation (LMedS) is proposed on the basis of ORB algorithm. The characteristics of ORB algorithm and LMedS method are used to remove the possible outer points. The false matching phenomenon is eliminated, and the correct matching feature pair is obtained, and the feature matching rate is greatly improved. At the same time, the attitude estimation based on nonlinear least squares is used. The simulation results show that the proposed method is robust in both feature point matching and real scene, and can be used in different scales. In the case of partial occlusion, it also has good performance, accurate and real-time tracking registration.
【作者單位】： 河南理工大學(xué)計(jì)算機(jī)科學(xué)與技術(shù)學(xué)院;
【基金】：河南省教育廳高等學(xué)校重點(diǎn)科研資助項(xiàng)目(15A520016) 河南理工大學(xué)博士基金資助項(xiàng)目(72515/074) 河南省高等學(xué)校礦山信息化重點(diǎn)學(xué)科開(kāi)放實(shí)驗(yàn)室基金資助項(xiàng)目(KY2015-02) 河南省科技攻關(guān)項(xiàng)目(162102310090) 河南省教育教學(xué)改革項(xiàng)目(2014)
【分類號(hào)】：TP391.41
，

本文編號(hào)：1672509