A procedure for change detection from images acquired without a fixed camera

A novel method for change detection from sequences of terrestrial (close-range) images is illustrated and discussed. The method was developed to analyze changes over a very long period of time (years). The proposed case study is the monitoring project of a rockfall, in which the first image was acquired in 2011, i.e. 7 years ago. The method does not rely on a fixed camera, so that the same camera can be removed and used for other applications. A procedure able to recover the alignment of the different images is required before running a image-to-image comparison for change detection. In the second step of the process, normalized multispectral cross correlation analysis with additional vegetation filtering is used to determine variations in the scene.