Object Recognition for High-resolution Remotely Sensed Imagery Based on Energy in Frequency Domain

An object recognition method for high-resolution remotely sensed imagery based on energy in frequency domain was proposed.Firstly,the pre-processed remote sensing images of typical objects were transformed from spatial domain into frequency domain by using the two-dimensional fast Fourier transform processing.Then the selected coefficients were composed feature vectors and sent into SVM(support vector machine) for training.Finally,SVM was used for recognition for test samples of typical objects,and the effect of feature window length on the object recognition rate has been investigated.The experimental results show that each object sample achieves comparatively high correct recognition rate when the width of feature window is 6,and the overall recognition rate is up to 88.96%.