Object detection with multiple features in remote sensing image

A method is presented for the localization of manmade objects in remote sensing images with complex scenes. The methodology was based on mean shift clustering in high dimension by extracting multiple features of manmade objects and forming them into a feature vector. By clustering the feature vectors, the manmade objects were segmented from the complex scenes. Results obtained on the detection of architectures in the image show that this methodology has good robustness and autonomy.