Research On Detection Of Airport Shelter Based On Multi-Feature In Optical Remote Sensing Images

Target detection for the airport shelters in remote sensing images is one of the significant aspects in target detection and recognition. The detection and location of the airport shelters, which have important strategic meaning and tactical values, can provide important target parameters for the precision attacking of guided weapons. At present, as the processing of the targets like the airport shelters is mainly depended on human interpretation, the speed and precision of processing can't fundamentally satisfy the requirements for the reconnaissance in battlefield and the processing of intelligence. Automatic detection of airport shelters, which resolves the key problem very well, can improve the speed and precision of target detection. However, automatic target detection is quite difficult because the shelter targets have less counterpart pixels in images and are hard to describe. Beginning from the analysis of airport shelter features in single source remote sensing images, the detection methods are investigated and good results are acquired.Firstly, the features of airport shelter targets are analyzed, and the position features and shape features are chosen for target discrimination.Secondly, the extraction methods for the position features of airport shelters are investigated. With the study of airport runway, taxiway and shelter combination target, it is find that the position feature can be extracted by the End Road Feature (ERF) in airport areas. The ERF extraction methods in larger remote sensing images are studied, experiments based on the images of high resolution satellites are carried out and the effect of extraction is analyzed.Thirdly, the matching methods for the shape features of airport shelters are investigated. Contour Sequence Moments (CSM) and Fourier Descriptors (FD) are chosen to describe the shape features. The theories and methods for matching shape features by CSM and FD are studied, and the detection effects of the two matching methods are compared. Simulation experiments are designed for both methods to validate and analyze the validity.Finally, the multi-feature detection methods for airport shelter targets based on serial structure is proposed by analyzing the characterization of position features and shape features in detection process. The position of the shelter targets can be finely detected with position features, while high false alarms ratios will emerge. By contrast, shape features are superior to the former in discrimination. Therefore, fused discrimination methods for targets which fuse the position features and the shape features are processed using the serial detection structure. The proposed method with high detection ratio removes the false alarm effectively. The performance of the method is validated using high resolution satellite-borne remote sensing images of airport.