Research On Key Technology Of Airborne Image Mosaicking

The aerial images with wide field of view and high resolution are often required for many applications,which are the basis of remote sensing interpretation and analysis.However,due to the limitation of factors such as the size of aerial camera sensor,the field of view,flight altitude and so on,there is a contradiction between the field of view and resolution,when obtaining aerial images using unmanned aerial vehicle(UAV)equipped airborne photoelectric platform.The common method is to ensure a certain overlapping coefficient for a region to carry out continuous image acquisition,and then to synthesize a panoramic image by image mosaic technology.UAV is small and light,easy to be affected by unstable airflow,and the external environment is complex and changeable.These will affect the quality of image and increase the difficulty of image mosaicking.At present,some technologies of aerial image mosaic are still immature.How to achieve fast and high-precision mosaicking is still a major challenge.Therefore,the in-depth study of image mosaic technology has important research significance and engineering value.In this paper,we expound the basic theory and development trend of image mosaic technology.Around the mosaicking process,the various segments involved in image mosaicking are researched systematically.Focusing on how to improve the accuracy and speed of registration,we study and realize aerial images registration.In order to improve the accuracy of image stitching,some new image mosaicking methods are studied,and the mosaic of multiple aerial images is realized.Based on the engineering requirements of airborne photoelectric imaging system,the image mosaicking hardware system is designed,which can realize the real-time mosaic of aerial images in a specific scene.The main contents and contributions of this paper are summarized as follows:1.Considering the shortcomings of the traditional SURF algorithm,a fast image registration algorithm based on the SURF features of the concentric circles neighborhood is proposed(CSURF).Firstly,the feature points are extracted using Hessian matrix.Then,the feature descriptor for each keypoint in the circular neighborhood is constructed using Haar wavelet response;meanwhile,the normalized gray values difference and second-order gradient of this region are computed.Finally,RANSAC algorithm is applied to eliminate false matches.This method not only performs faster than SURF algorithm,but also fully employs the image gray information and details to acquire higher accuracy.The proposed method has strong robustness and stability for blur,illumination difference,angle rotation,scale change,and viewpoint change.2.In order to realize the fast registration of aerial images and improve the registration accuracy,a novel image registration method based on the BRISK feature of the directed line segment is proposed(DLBRISK).In this method,firstly BRISK algorithm is used to match in order to acquire rough point matching.Secondly,construct directed line segments,and describe them with BRISK feature.Finally,the nearest line segment matching method and probability statistics model are applied to precisely match feature points.The algorithm effectively eliminates the mismatching point pairs using the texture information of the image,and is robust to the problems such as illumination difference,angle rotation,low resolution and scale change which often appear in image mosaic.Especially for low resolution images,the algorithm can effectively reduce the mismatch rate and provide more accurate transformation model for image mosaicking.3.In view of the ghost effect and geometric dislocation in the process of image mosaic,a seamline detection method based on parametric kernel graph cuts(PKGC)segmentation is proposed.A two-level optimization strategy is applied to determine the seamline.Object-level optimization is executed firstly.Background regions(BRs)and obvious regions(ORs)are extracted based on the results of parametric kernel graph cuts segmentation.The global cost map which consists of color difference,a multi-angle multi-scale morphological gradient(MAMSMG)constraint,and texture difference is weighted by BRs.Finally,the seamline is determined in the weighted cost from the start point to the end point.Dijkstra’s shortest path algorithm is adopted for pixel-level optimization to determine the positions of seamline.The method effectively extracts the positions of obvious objects by using image segmentation,avoids the seamline passing through the obvious area,and the energy function contains comprehensive information,which is beneficial to the detection of more reasonable seamline.At the same time,a new seamline detection strategy is designed for multi-image overlapping regions,and multi-image mosaicking is realized.4.Considering the actual needs of the project,a hardware system of image mosaic based on FPGA is designed in this paper,including visible light digital camera,target location unit,image mosaic unit,memory and input / output unit,and the design of the system schematic diagram and the PCB circuit design are completed by the module.The system realizes the mosaicking of aerial images directly according to the geographic information of aerial images,and the mosaic effect is good.The processing time can meet the requirements of quasi real time in the engineering.At the same time,according to the processing of aerial image with geographic information on PC,hierarchical registration and mosaicking method is used.And before each level of the image processing,the approximate overlap area is determine by using the image geographic information to save time cost and improve the accuracy of the mosaic image.In summary,the related theories of aerial image mosaic technology are discussed and analyzed.The key technologies such as image registration,seamline detection and real-time processing have been studied,and achieved some results.The relevant achievements of this paper provide a theoretical basis for the further study of aerial image mosaicking,which can be used for reference to improve the accuracy and speed of the mosaic image.