Research On Splicing Algorithm Based On UAV Aerial Image

In recent years,unmanned aerial vehicles(UAV)have been widely used in land dynamic monitoring,disaster emergency and sea area monitoring because of their high flexibility,convenience,high efficiency and low cost.It has become a powerful complement to satellite remote sensing.However,compared with traditional aerial surveying,UAV aerial photography can not have a macroscopic understanding of the whole aerial photography area because of its limited visual angle and other shortcomings.Therefore,it is necessary to quickly stitch aerial video or a large number of UAV remote sensing image sequences to obtain a wide range of panoramic images.Therefore,how to fast registration and automatic mosaic of massive data or highdefinition video from low-altitude remote sensing is an urgent technical problem.In this paper,the geometric correction,distortion correction and image enhancement of aerial images are studied theoretically.The image pre-processing experiments were carried out.Three evaluation indexes,namely registration rate of feature points,extraction stability of feature points and detection speed of feature points,are proposed to compare Harris,SIFT(Scale-invariant feature transform),SURF(Speeded-Up Robust Features),ORB(Oriented FAST and Rotated BRIEF)feature extraction algorithms.Finally,SURF features are selected for image frame registration.Finally,SURF features are selected for image frame registration.The traditional SURF features are improved.Firstly,feature detection is performed by using FAST(Features from Accelerated Segment Test)in feature selection.The SURF description is calculated on the basis of FAST detection.The uniqueness of feature description is improved by adding color information of feature points to the original SURF feature descriptor and normalizing it.Finally,the overlapping regions between images are estimated by the flight parameters of UAV.The improved SURF algorithm is used to extract the overlapping regions and combined with RANSAC(Random Sample Consen).Compared with the traditional SURF,the registration speed is improved by about 800 ms,and the matching accuracy is improved by 8%.Finally,the geometric transformation matrix between images is obtained by using the refined matching point pairs and the improved weighted average fusion strategy is used to fuse the images.A relatively good stitching effect is achieved.Then,according to the characteristics of UAV aerial images,a mosaic method based on geographic coordinates is proposed.This method uses the spatial coordinates between images for registration.The main idea is to use the matching pairs of the same feature points to correct the spatial coordinates of the image in turn,and convert the offset of the spatial coordinates into the pixel coordinate offset of the image for projection.Coordinate offset and fine-tuning are less than 100 pixels.The result of the final stitching experiment not only has the geographic coordinate information,but also can solve the problem of registration error accumulation in feature-based stitching.The new method can quickly stitch the image,and the overall visual effect is better,especially in the aerial area as the sea area.It is practical to use the spatial coordinates of images under the condition of less effective regions.Finally,according to the above research results,combined with the actual application of the subject,using C++ programming language to achieve image preprocessing,registration,fusion and other functions modules,and for the UAV aerial video,the design and construction of aerial video mosaic system,through this system,can play aerial video while completing the mosaic.The system is stable and reliable,and has good maneuverability and interactivity.