Automatic Co-registration of Satellite Time Series via Least Squares Adjustment

Abstract

Image-to-image co-registration is a fundamental task for data processing of satellite time series. This paper presents a new multi-image co-registration algorithm that simultaneously uses multi-image corresponding points in the whole multi-temporal sequence. Image co-registration parameters are then computed on the basis of a global adjustment. The implemented algorithm provides sub-pixel accuracy similar to that achievable with interactive measurements, but it is also able to register images which do not directly share corresponding features with the reference. Results for a (i) synthetic dataset and a (ii) real complex multi-temporal series made up of 13 Landsat images collected over a period of 30 years are illustrated and discussed. The obtained results showed that the implemented algorithm is atmospheric resistant and quite robust against land cover changes, cloud cover, and snow.

Sliver Removal in Object-Based Change Detection from VHR Satellite Images

“…The tiles of both images are independently matched by using initial georeferencing parameters as approximate values. The implementation is derived from a solution for close-range image orientation described in Barazzetti et al (2013), which was adapted to handle medium-resolution satellite images (Barazzetti et al, 2014).…”

“…Metric accuracy has recently reached 30 cm with the launch of WorldView-3 in August 2014. Although the metric resolution is still not comparable with aerial images (VVHR, i.e., very, very high resolution), the continuous acquisition of VHR images is more straightforward than an aerial block (Barazzetti et al, 2014), which is expensive and requires a flight plan. Some actual information on aerial cameras and rules for image acquisition can be found in Kraus (2007).…”

Robust Automated Image Co-Registration of Optical Multi-Sensor Time Series Data: Database Generation for Multi-Temporal Landslide Detection

“…Liu and Chen [35] have co-registered multi-temporal Formosat-2 Level 1A images (8-m resolution), achieving a RMSE of approximately 1.5 pixels in flat terrain and 2.2 pixels in mountainous areas. Barazetti et al [36] automatically co-registered 13 Landsat TM datasets acquired over a 30-year period mainly by correcting sub-pixel translation errors, resulting in a relative accuracy of sub-pixel RMSE values. Although, accuracy requirements for co-registration depend on the targets and methods used for change detection [12,37], in general, accuracies (RMSE) of less than one pixel are considered suitable for subsequent change detection [38].…”

Automatic registration of multi-source medium resolution satellite data

“…On the other hand, this feature provides 2N equations. The first multi-image mapping function proposed in this implementation was the similarity transformation (Barazzetti et al, 2013;2014a;2014b):…”

“…The study was carried out by using spectral band 1 for Terra/ASTER (0.520-0.600 µm), band 2 for EO-1/ALI (0.525-0.625 µm) and band TM2 for Landsat-5/TM (0.520-0.600 µm). No atmospheric correction was applied beforehand (Barazzetti et al, 2013;2014a;2014b Table 2. The multi-source data set available for the case study.…”