Improvement of images obtained from unmanned aerial vehicle with deep learning method

The development of technology and the camera has added a different dimension to the images obtained after the photography process. Images, which were initially obtained in black and white, began to be obtained in color with the developments over time. As the importance given by human beings to image quality increases day by day, many studies have been carried out on improving image quality. As a result of the studies carried out to increase image quality, the lenses in cameras were improved and different image enhancement techniques emerged. In different lighting conditions, the development of the lens in cameras has not been sufficient to improve image quality. Therefore, traditional image enhancement techniques such as histogram, contrast, brightness and histogram equalization have been used. Improvements have been made based on traditional image enhancement techniques. Unmanned aerial vehicles (UAVs) are remote-controlled and autonomously controlled aerial vehicles designed to be used in various fields such as reconnaissance, surveillance, search and rescue, military and entertainment. The images obtained during missions carried out using UAVs in low lighting conditions must be clear and understandable. Therefore, in this study to improve and improve image quality, the improvement of low-light images was analyzed using the deep learning method. Image improvement was made in low light using Convolutional Neural Networks, one of the artificial neural networks that form the basis of deep learning. The study was conducted using the Python programming language in the Google Colab environment. The main purpose of the study is to improve and develop the low-light images obtained from the UAV with the deep learning method. In this study based on Zero DCE-Net, quality measurement techniques were used to evaluate and compare image quality. In the literature, the superiority of the Zero DCE-Net study compared to other studies has been seen. In this study, according to the data obtained using quality measurement techniques with the developed model, a more successful result was obtained compared to Zero DCE-Net and other low-light image improvement methods. According to the results of the study, it may contribute to other studies on improving the image quality in UAV technologies.