A Method For Semi-Automatic Extraction Of Buildings

As aerospace technology becomes more and more mature, a variety of high-resolution satellite sensors have been launched. Faced with a large number of remote sensing image data, extract buildings from the high resolution image in the areas of production, mainly need the operator’s manual identification and painted hook, it is a labor-intensive and tedious work, On the other hand, Automatic extraction methods have not yet been achieved thus far for the high difficulty and low precision. On the current level of technology, the use of semi-automatic extraction of human-computer interaction is more realistic. In this paper, several automatic building extraction algorithms and conventional semi-automatic extraction algorithms were studied, like GLC tree algorithm and pixel shape index(PSI) algorithm. the optimal path of seed filling algorithm, region growing algorithm and improved regional filling scan line algorithm. Finally, an was proposed, an improved seed filling algorithm which was used for semi-automatic extraction was proposed,and it was based on the studies of the semi-automatic extraction algorithm.In this paper, GLC tree algorithm and the rule sets were used for automatic building extraction. After the comparison of the simplicity and accuracy between two algorithms, the result of GLC tree was used as the final result for automatic building extraction. Then, the improved seed filling algorithm was worked for misclassification and leakage of the result, Semi-automatic extraction vector data as reference data was involved in the second automatic building extraction, to ensure the accuracy of building extraction. The problem of cumbersome and low efficiency in production were solved by the improved strategy of building semi-automatic extraction.Ultimately, an interactive workflow and semi-automatic building extraction plug which was based on the intelligent interpretation platform FeatureStation GeoEX were worked.