Automatic registration method of visible light and thermal infrared image sequences

Abstract

The invention discloses an automatic registration method of visible light and thermal infrared image sequences, which solves the technical problem that the existing registration method of the visible light and thermal infrared image sequences has bad registration accuracy. A technical scheme is as follows: using the normalization light stream timing characteristic and a three-layer gradual fining similarity measurement method of an image pixel to select a candidate homologous point pair among visible light and thermal infrared image sequences; and using the movement timing characteristic of the image pixel to replace the whole characteristic of a movement target. The method effectively overcomes the problem caused by inaccurate movement detection, improves the arithmetic robustness, can be used for registration of image sequences under a crowded monitoring condition with frequency movement, improves the registration accuracy of the image sequences, and reduces an average pixel error relative to reference data to be within a range of 2 pixels from [2.03, 6.9] pixels in the prior art.

Description

Technical field [0001] The invention relates to an image sequence automatic registration method, in particular to a visible light and thermal infrared image sequence automatic registration method. Background technique [0002] Literature "Torabi A, Masse G, Bilodeau GA, et al. Feedback scheme for thermal-visible video registration, sensor fusion, and people tracking. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition-Workshop, 2010, 15~ 22." discloses a visible light-thermal infrared image sequence registration method based on moving target trajectory and moving target silhouette. This method first estimates the transformation model parameters based on RANSAC trajectory matching to maximize the overlap between the corresponding points of the trajectory and the motion foreground of the two types of sensors; then, on the basis of the registration parameters, the sum rule fusion is performed on the detection results of the two types of sensors, and bas...

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the deficiency of poor registration accuracy of existing visible light-thermal infrared image sequence registration methods, the present invention provides an automatic registration method for visible light and thermal

Images