Hyperspectral anomaly detection based on laplace of gaussian operator

Hyperspectral remote sensing images contain not only spatial information, but also abundant spectral information, which are widely used in the field of space-spectrum joint target detection. Unlike other target detection algorithms, the anomaly detection doesn’t require any prior knowledge, and can effectively identify the pixels that stand out from the cluttered backgrounds in high spectral images. At the same time, compared with the background objects, the abnormal target is composed of sub-pixels and has distinctive spectral characteristics. In this paper, a new anomaly target detection algorithm based on Laplace of Gaussian (LoG) operator is proposed to solve the problem that spatial information is not fully utilized and the real-time detection capability is not strong. Firstly, the algorithm uses the LoG operator to obtain the target detection results under different bands with analyzing the spatial characteristics of the anomaly, combined with the blob detection theory which is widely used in the field of the image recognition field. The results are finished by the spatial filtering, which highlights the anomaly and effectively suppress the background. Then, a Boolean map-based fusion approach and morphological expansion theory is used to synthesize the detection results of different bands. In the end, the real AVIRIS Imagery and HYDICE Imagery are used for simulation, and the results show that the algorithm is with strong robustness, high detection probability and low false alarm rate.