Sift-based object matching and tracking of coal mine

摘要

Proposed a new algorithm based on Scale Invariant Feature Transform(SIFT) algorithm to suit for object matching in special environment of coal mine. New algorithm combines RANSAC with L-M nonlinear optimization algorithm after cross-matching cursorily to estimate optimization parameters, local regions of different images and angle between eigenvectors are used to reduce search scope and cost time. Experimental results show that the new algorithm has good robustness on low illumination, blur, scale change, shelter by other object and high noise condition. It can increase matching accuracy, reduce the computation for real-time processing of video surveillance and object tracking system of coal mine area.