3D modeling of unmanned aerial vehicle night images based on improved retinex

Unmanned aerial vehicle oblique photogrammetry technology cannot achieve good image modeling under poor lighting conditions, and existing algorithms are prone to image detail loss, edge blurring, and other issues when enhancing nighttime images. This article proposes a single scale Retinex algorithm based on bilateral filtering to address this issue. Firstly, the algorithm incorporates reflection components into the real image to reduce images with low lighting and high nighttime light brightness. Compared with the traditional Retinex algorithm, which uses Gaussian function, this paper adopts bilateral filtering function instead. Finally, in order to reduce the color influence of nighttime street lights, color distortion correction is carried out, and the modeling of nighttime unmanned aerial vehicle images is finally achieved.