Mejoramiento de fotografías aéreas tomadas por drones, implementando el modelo preentrenado de inteligencia artificial vgg16

Abstract

Photo editing is a professional field that requires a high degree of skill and expertise, and even experienced photographers can make mistakes, such as overexposing an image or removing important details. Today, AI (Artificial Intelligence) is emerging as a tool with great potential to optimize various fields. Its ability to process information at high speed opens a world of opportunities to improve efficiency and quality, enriching the photo editing workflow. The above raises the question of what potential is applicable in the field of topography, more specifically in photogrammetry. Therefore, this document describes the research on these artificial intelligence technologies and how they are applied to give way to the development of a code whose main objective is to improve the radiometric resolution of images captured by drones, using the pre-trained artificial intelligence model VGG16. This model analyzes the histogram of the reference image and performs photo editing, modifying the color of the images and matching the tones and contrast of the reference image in less time than it takes to perform this same activity using more conventional and manual techniques. The research work is divided into 4 phases ranging from the collection of the necessary information for the development of the algorithm, the analysis of the characteristics of the inputs and resources used, the structuring of the algorithm which is done in the Python language and whose operation is based on the adjustment by equalization of histograms and correction of contrasts through the use of convolutional networks (CNN), the implementation of this algorithm and finally the analysis of the products results of the execution of the code in which improvements in color, brightness and shadows are evident in a significant way where the final result is equitable in visual terms achieving harmony in the images in shorter periods of time than those that take adjustments of the same type using a manual methodology, which allows to reduce the time by 69.06% between adjustments, evidencing that it really is a subject that is not in foray, since it has not yet been applied currently. In conclusion, it is a process that must be implemented and continue to be developed since it improves the quality of the images, as well as their efficiency.