Random sample consensus

lastmod 2024-02-21created 2014-10-2012 edits

First of all, RANSAC is an acronym formed from the capital letters of RANdom SAmple Consensus. Taken literally, it means randomly drawing sample data and then selecting the one with the largest consensus.

The least squares method finds a model that minimizes ∑residual^2 with respect to the data, whereas RANSAC selects the model with the maximum consensus, that is, the model supported by the largest number of data points.

Applications of RANSAC

• Finding a specific object in an image using local feature matching
• Visual Odometry (estimating camera motion between adjacent image frames)
• Performing scene matching for location recognition
• Estimating the motion of a moving object between adjacent image frames for object tracking

See also

• Statistical algorithms
• NG RANSAC (Deep learning)

Favorite site

• Wikipedia (en) description of RANSAC
• [Recommended] Dark Pgmr :: Understanding RANSAC and its use in image processing
• [Recommended] Overview of the RANSAC Algorithm ¹
• Line Fitting Example using RANSAC(RANdom SAmple Consensus) ²
• Line fitting using RANSAC
  • Github - Multi-threaded generic RANSAC implemetation
• [Recommended] Corner Detection - #3 RANSAC(RANdom Sample Consensus) ³
• The principle of the RANSAC algorithm that even middle school students can understand
• [Recommended] Outlier removal and Robust Estimation (3) (Why you shouldn't just use OpenCV RANSAC + upgraded RANSAC)
• [Recommended] RANSAC (RANdom SAmple Consensus) concept and practice - gaussian37
• (OpenCV) Homography and image matching findHomography,drawMatches, RANSAC : Naver Blog
• (Image processing) RANSAC algorithm - Blog of screenshots of handwritten notes

References