Seamless Mosaicking of UAV-Based Push-Broom Hyperspectral Images for Environment Monitoring

This paper proposes a systematic image mosaicking methodology to produce hyperspectral for environment monitoring using an emerging UAV-based push-broom imager. The suitability of alternative methods in each step is assessed by experiments urban scape, river course and forest study area. First, the strips were acquired sequentially stitching UAV images scanning along flight line. Next, direct geo-referencing was applied strip get initial geo-rectified result. Then, with ground control points, curved surface spline function used transform improve their geometrical accuracy. To further remove displacement between pairs strips, improved phase correlation (IPC) SIFT RANSAC-based method (SR) registration. Finally, weighted average best fusion spectral differences seamless mosaic. Experiment results showed that as GCPs‘ number increases, mosaicked image‘s accuracy increases. In registration, there exists obvious edge information can be accurately extracted from scape area; comparative achieved IPC less time cost. However, objects complex texture like forest, edges prone inaccurate result failure method, only SR good fusion, all three areas. Whereas, useful eliminating line areas but failed area due heterogeneity different objects. For applications, proposed provides practical solution seamlessly mosaic high fidelity.