For the reconnaissance problem of the spatial information’s low utilizatio in hyper-spectral sensing, a novel anomaly detection algorithm based on spectral domain-spatial joint characteristics is proposed. At first, each neighborhood pixel is given weights through the cosine of the spectral gradient angle and the spatial feature is gotten by adding the weighted neighborhood pixels together. The spectral domain-spatial joint characteristic is gotten by adding the weighted spectral feature and the spatial feature together. And then, the hyper-spectral data with the spectral domain-spatial joint characteristics is carried out principal component analysis to extract the main components to carry on anomaly detection. At end, from the binary image and the receiver operating characteristic curve (ROC) of the anomaly detection, it can be seen that the proposed algorithm has superiority and it could improve the detection effect.