Research On Automatic Seamline Detection Model And Its Application For Urban Orthoimage Mosaicking

Image mosaicking is an essential step in production of large-scale digital orthophoto map(DOM).There are many obvious ground objects such as buildings,trees,and cars in urban DOM.And it is necessary to prevent the seamline from passing through these areas with large projection difference to avoid dislocations and fractures in the final mosaic.At present,in the actual production the automatic detection of the seamline is still the key problem that restricts the automatic production of large-scale DOM.Therefore,this study proposed an automatic model to detect the optimal seamline for urban orthoimage mosaicking.The processing flow could be divided into two parts: global initial seamline network generation and initial seamline network optimization.Focusing on the concept and the practical application research of the proposed model,the key work and innovations of this study were as follows:(1)This study improved the method for generating the seamline network based on Voronoi diagrams.Firstly,based on the central point set of DOM effective regions,the Voronoi diagrams was applied to generate a globally consistent seamline network,determining the effective contribution region of each orthoimage for the final mosaic.Then,the corresponding digital surface model(DSM)data was first identified as ground regions and obstacle regions(e.g.,buildings,trees,and cars)by threshold segmentation.Furthermore,the mathematical morphology method was used to make the edge of obstacles more prominent.After that,vertices of the mosaic polygon were relocated with the processed DSM to achieve a seamless photomosaic without holes,and ensured those vertices were located on the ground area.(2)This study proposed the method for detecting the optimal seamline based on DSM and improved jump point search(JPS)algorithm.When the initial seamline network was generated and adjusted,the optimization of each seamline segment was performed.First,the search region was determined according to the starting and finishing points of the seamline segment.And the corresponding DSM processed image was obtained as the search cost image.Then,the JPS algorithm was improved and employed to search for key jump points in the cost image.Meanwhile,the jump points would be evaluated according to an optimized function,finally generating a minimum cost path as the optimal seamline.Furthermore,the search strategy was modified to avoid search failure when the cost map was completely blocked by obstacles in the search direction.Last,the image mosaic and fusion were performed according to the optimized seamline network,in order to eliminate visible splice marks and achieve a smooth transition between images.(3)The proposed model was verified by an experiment with the open dataset of Vaihingen,Germany.And application research was carried out based on the urban DOM dataset of a region in Guilin,China.The comparison methods were the Dijkstra's algorithm,Inpho OrthoVista and ArcGIS.Results demonstrated that the seamlines obtained by the proposed model could effectively avoid crossing obvious ground objects such as buildings,trees,etc.The practical application showed that the number of seamlines crossing obvious obstacles obtained by the proposed model was only 38%-55% of that obtained by other three algorithms.Furthermore,the efficiency and resource consumption were greatly improved since the improved JPS algorithm skipped many image pixels without them being explicitly evaluated.The time consuming of the proposed model and other three methods were 11s?121s?376s and 215 s,respectively,proving the superiority of the proposed mode.In general,the proposed automatic model in this study was helpful for detecting the optimal seamline for massive and large-scale urban orthoimage mosaicking.