498 / 2019-02-28 18:55:34

Cross Lines Detection Based on Gaussian Mixture Model

connected components; Gaussian Mixture Model; clustering; clusters merging

全体主题 > Algorithm

Cross lines are common in many vision perception scenes, which often provide vital information for automation systems. Effective and robust cross lines detection algorithm is still scarce. A novel algorithm based on Gaussian Mixture Model (GMM) is proposed to advance the detection of cross lines in this paper. To strengthen the cleaning of noise， the method of connected-component labeling is used in the preprocessing of the input image. We take advantage of GMM to model cross lines for their similar pattern and utilize GMM clustering to get individual lines. Because the number of clusters is pre-assigned, a cluster merging post-processing step is followed to get the whole pixel set of lines. Finally, the model of lines can be fitted. Experiments show that our algorithm gets respectable performance.