A critical review on multi-sensor and multi-platform remote sensing data fusion approaches: current status and prospects

Numerous remote sensing (RS) systems currently collect data about Earth and its environments. However, each system provides limited data in terms of spatial resolution, spectral information, and other parameters. Given technological constraints, combining data from diverse sources can effectively enhance RS solutions through data enrichment. Many studies have investigated the fusion of RS data acquired through different sensors and platforms. This paper provides a comprehensive review of research on multi-platform and -sensor RS data fusion, encompassing visible-light images, multi/hyper-spectral images, RADAR images, LiDAR point clouds, thermal images, spectrometry samples, and geophysical data. An analysis of over 950 papers revealed that feature-level fusion of multi-sensor RS data was the most commonly employed technique, surpassing pixel- and decision-level approaches. Moreover, satellite data fusion was more prevalent than the fusion of data acquired from manned and unmanned aerial vehicles. The integration of multi-sensor RS data initially gained traction in applications such as precision agriculture before expanding to land use and land cover mapping. This paper addresses previously overlooked issues and presents a framework to facilitate the seamless fusion of multi-platform and multi-sensor RS data. Guidelines for this fusion include ensuring the data have the same acquisition time, spatial co-registration, true orthorectification, consistent spatial resolution or information content, radiometric consistency, and wavelength of spectral band coverage.