The Rescarch Of Aerial Image Quality Evaluation Based On Reduced Redundancy

In the actual production application,UAV is often restricted by altitude and phase amplitude.In the aerial photography area,there are more images and higher overlaps,resulting in a large amount of data redundancy,which results in the long running time of stitching images on conventional computers.The image mosaic efficiency and quality can not be seriously affected by large area stitching,and some phenomena such as blur,double shadow or abnormal exposure caused by aerial photography.In response to the above problems,based on the land right confirmation project of a county in Gansu province and the measured UAV image as the basic data,this paper analyzes and studies the reduction of UAV image data redundancy and the selection of high quality stitched images.And the result is as follows:1.The statistical method of overlap degree of UAV image is studied.After analyzing the shortcomings of the existing statistical methods,a new statistical method of overlap degree based on image matching is proposed in this paper.This method can realize multi-sorties,large-batch image overlap automatic,high-precision statistics.2.In the study of image redundancy interval,firstly,the information of image overlap degree of different vehicles and intervals is calculated,and these data are compared and analyzed.Finally,the relationship between image overlap degree and image redundancy interval is obtained.3.The experimental analysis without reference quality evaluation method,through review a large number of domestic and foreign literature,summed up six commonly used non-reference quality assessment algorithms:BIQI algorithm,BRISQUE algorithm,NIQE algorithm and SSEQ algorithm and SSIM algorithm,are summarized.Then the above six evaluation methods are tested by LIVE image database.The score of various image quality evaluation methods and the original subjective score of the database were analyzed by Spearman rank order correlation coefficient.Analysis.Finally,the correlation between BRISQUE algorithm and subjective score of human eyes is the highest.Therefore,BRISQUE algorithm is selected as the method of image quality evaluation without reference in this paper,and it is applied to the experiment of reducing redundancy.4.The image splicing experiments are divided into three kinds:the splicing of the original image,the stitching of the fixed interval screening image,the stitching of the image after the fixed interval based on the image quality evaluation method,and the comparison and analysis of the accuracy of the control points,the quality of the image and the time limitation of the stitching,and the final fixation of the three images.The interval method is the least used when compared with the original stitching,which is 48.42%higher than that of the original stitching.The image quality based on the image quality evaluation is the best and the error value in the image plane position meets the requirements of the standard,and the stitching of the method is 37.63%higher than that of the original stitching.Synthesizing the above experimental analysis,it is concluded that the method of image quality evaluation in fixed interval is used to screen the original image,which can improve the quality and efficiency of image stitching obviously.