Presenting a new algorithm in order to improve the radiometric of UAV photogrammetry images based on color, light and contrast

The radiometric quality of the images is a critical factor that significantly affects the fitting and creation of 3D models.Factors such as the type of camera used, the camera setting parameters, the shooting time and the weather conditions have a significant effect on the radiometric quality of the images.In addition, taking pictures in the last hours of the day can lead to reduced light and image brightness, thus affecting color and contrast.The aim of this study is to identify the inappropriate images in order to increase the radiometric quality of a UAV data set in terms of color, contrast and brightness.In this regard, a pre-processing method using the non-reference method has been used to identify the quality.The division of the images into good or bad quality categories is achieved using a threshold based on a fuzzy system.Image enhancement has been done using the convolutional neural network.The analysis of the sample images shows that the quality of the images has increased by 86% in terms of color, 56% in brightness and 53% compared to the original mode.The evaluation performed after the image improvement process shows that compared to the raw data processing mode, the quality of the generated orthophoto has increased, the errors related to 3D modeling have decreased, and the density of pointclould has increased.