Hyperspectral Target Detection for Multiple Target Spectra
Abstract— Target detection is an important application in hyperspectral imaging. Conventional detectors for hyperspectral target detection requires that there is only one target spectrum from a spectral library for detecting a man-made object. The conventional detectors are derived assuming that target and background pixels are from different multivariate normal distributions. This paper proposes a target detection model for detecting a man-made object from multiple target pixels. The model is derived without requiring the assumption of multivariate normal distributions for the pixels. Experimental results using multiple target pixels from a hyperspectral image show that the model can work well in target detection.
Index Terms— target detection, hyperspectral imaging, remote sensing.
Department of Mathematical Sciences, Susquehanna University, USA
Cite: Edisanter Lo, "Hyperspectral Target Detection for Multiple Target Spectra," Proceedings of 2018 the 8th International Workshop on Computer Science and Engineering, pp. 775-779, Bangkok, 28-30 June, 2018.