Scanning Line Based Random Sample Consensus algorithm for fast arc detection
Xiaoyu Song, Ting Jing, Shuai Yuan, Song Guo, Yuxin Li · 2016
The traditional random sample consensus (RANSAC) algorithm is capable of estimating a model with fewer data points and almost unaffected by noise. There are several drawbacks of such algorithm including detection errors, unstable threshold and massive calculation. By analyzing the spatial relations of graphics pixels, a hypothetical circle is firstly formed with three hypothetical points which are generated by three random black pixels. Secondly, based on the hypothetical circle, rectangular region and annular region are built to reduce the amount of the black pixels detected by random sample algorithm, which decreases computational burden and increases the stability of detection threshold. Then the arc detection is realized based on circle detection using the methods of scanning line and segment detection. Finally, the experimental results illustrate that the proposed method, Scanning Line Based Random Sample Consensus (SLRANSAC), can not only improve the accuracy and robustness of arc detection, but also enhance the efficiency of arc detection compared with the error vector magnitude (EVM) algorithm that is widely used.