Classification Method Based on Feature between the Spectral Using High Spatial Resolution Remote Sensing Images
The traditional image classification method can only use the feature at the spectral such as gray and texture,can' t take full use of the features between the spectral.In order to overcome this insufficient,a method of fusion features between the spectral is proposed in the paper.Image was transformed using PCA method.The first two principal components was chosen as image after transform.Image segmentation was employed to get image objects from which training sample objects were selected.Gray histogram and joint gray histogram were used to describe features at the spectral and features between the spectral.Histogram distance was measured by G statistic.Inverse distance weighting method was used to calculate the probability at the spectral and probability between the spectral which were weighted to build joint probability.Image classification result was obtained by maximum joint probability.The experimental results on the QuickBird image verify the effectiveness of the proposed method.The overall classification accuracy and kappa coefficient have been reached 90％ and 86.7％.
feature in the spectralfeature between the spectralPCAhistogramG statistic
NETL
NSTL
万方数据
维普
