UAV Image Splicing Parallel Computing Technology Research

The unmanned aerial vehicle(UAV)remote sensing is an important technology that can supply real-time data acquisition and emergency commands and rapid responses from a natural disaster(such as earthquake relief,etc).It not only has the characteristics of low cost,simple operation,flexible and so on,but also it has a lot of images those have high resolution and strong real-time and other advantages.At the same time,the remote sensing image of UAV has the characteristics of small format,a large amount and multi-overlapping.It is very important to realize fast and efficiently UAV sequence image splicing with real-time images for the emergency command.When dealing with UAV's sequence of images,there is no doubt that the error of image deformation will continue to increase,with the number of stitching images increasing.In dealing with UAV sequence images mosaic,the next images depend on the previously registered stitching images,which will produce accumulation of errors gradually.Therefore,in the process of completing the splicing,we not only need to ensure the efficiency of image splicing,but also the quality of image stitching.Only in this way,can we continue to play the advantages of UAV remote sensing,so that UAV stitching images are becoming more and more widely used,and play a more and more important role.With the popularity of multi-core computers,the use of multi-core parallel programming has become an irresistible trend.It not only has a super computing power,but also has a high operating efficiency.In order to further liberate human and material resources,we need to give full play to the use of multi-core computer performance to improve the quality and efficiency of UAV sequence image stitching.In this paper,first,the UAV images are preprocessed.It is mainly about Laplacian sharpening,enhancing the edge of the image,using Wallis filter to uniform light and color,according to the block gray information,according to the overlap to intercept the region of interest.Second,we present a Harris feature extraction based on uniform grid,then use the Mahalanobis distance to extract the tie points.At the same time,the nuclear line and unique multi-level constraints are used to implement from coarse match to fine match.Finally,the RANSAC algorithm is used to register the images,finishing the splicing images.On this basis,we consider the sources of the quality of UAV-image stitching,the multi-flight UAV images are spliced by multi-level groups,and the mosaic errors of different chunks are controlled.In this context,multi-core parallel programming has been successfully introduced.That is to ensure the quality of image stitching,but also the efficiency.In order to improve the multi-core computer running speed ratio,we propose to dynamically create threads and create a thread pool.Through the experiment,it is found that the extracted feature points are evenly distributedby the designed idea in this paper,and the accuracy of matching is improved,compared with Harris feature extraction algorithm.Through the experiment it is found that parallel splicing can effectively control the accumulation of error propagation,compared with the serialized splicing.At the same time,the running time is greatly reduced.To a certain degree,the results validate the feasibility and effectiveness of the Harris feature extraction based on uniform grid and multi-level blocks parallelization programming algorithm.