The hyperspectral image has spatial resolution and inter-spectral resolution, which can visually display the information of the ground object. It is of great significance to the typical targets of hyperspectral data mining. With the development of drone technology, it is possible to detect targets with airborne hyper spectrometers, thereby greatly improving the perception ability of unmanned aerial vehicles. For this reason, we have carried out research on target data mining based on the advantages of hyperspectral detection of ground object attributes and the strong flexibility of UAVs. First, on the basis of acquiring hyper spectral images, normalize the images, construct an edge extraction model, and introduce the idea of clustering to find spatially similar regions. Then a Dynamic Time Warping model is constructed to extract the features between the spectra, and finally, the DEC algorithm is improved, and a deep network is used to achieve typical target clustering. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.