To address the issues of low detection rate and high false alarm rate caused by complex background during sub-pixel aerial aircraft detection in hyperspectral remote sensing image， an aerial aircraft detection method was proposed based on contrails cloud proposal. Firstly， a hyperspectral semantic segmentation model was used to search for the contrails cloud， and regions of interest（ROIs） of aircraft were proposed to reduce invalid search ranges and suppress false alarms based on the contrails cloud. Secondly， an endmember extraction algorithm based on dictionary learning and semi-blind non-negative matrix factorization was proposed to improve the accuracy of aircraft endmember extraction for hyperspectral subpixels. Finally， verification experiments were carried out on the hyperspectral remote sensing image dataset of Gaofen-5 satellite. The results demonstrated that the algorithm proposed in this paper can effectively suppress false alarms in complex scenes， and significantly improve the detection rate and detection accuracy of sub-pixel aerial vehicles.