A shadow removal method for UAV remote sensing images based on deep learning

Abstract

The invention discloses a method for removing shadows of UAV remote sensing images based on deep learning. First, the UAV is used for data collection, and the data is subjected to radiation normalization and geometric registration processing to construct a UAV shadow database; and then Based on the shadow database, the conditional generative adversarial network 1 is used to learn the shadow removal relationship between sample pairs, so as to realize the preliminary removal of shadows; considering the difference in radiation before and after shadow removal, a radiation normalized database for non-shaded areas is constructed, and in Based on this database, conditional generative adversarial network 2 is trained; finally, based on this relationship, radiation normalization is performed on the preliminary results of shadow removal to obtain the final shadow removal results. Considering the flexibility of UAV data acquisition, the present invention collects and constructs a shadow image data set, and utilizes deep learning theory to dig deep into the transformation relationship between sample pairs to obtain the optimal shadow removal result. It has high accuracy, fast calculation efficiency, easy implementation, strong scalability and high practical value.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, and relates to a method for removing shadows, in particular to a method for removing shadows from UAV remote sensing images based on deep learning. Background technique [0002] Shadows widely exist in high-resolution remote sensing images, especially in urban areas with dense buildings, causing brightness loss of local information and directly affecting the accuracy of remote sensing interpretation. Therefore, in order to improve the utilization efficiency of remote sensing images, it is very necessary to remove shadows in high-resolution remote sensing images. [0003] Existing methods can be mainly divided into two categories: local matching method and global optimization method. The local matching method is better for the case of a single type of object inside the shadow, but because it is more sensitive to sample selection, and for the case of complex objects inside ...

AI Technical Summary

Problems solved by technology

The local matching method is better for the case of a single type of object inside the shadow, but because it is more sensitive to sample selection, and for the case of complex objects inside the shadow, the artifacts in the compensation results are obvious, and it is easy to produce serious color loss. Partial

The global optimization method obtains the global optimal solution through iterative optimization, which can often obtain better overall correction results, but for complex shadows covering multiple surface types, it often leads to color distortion and blurred details

Images