Target spectral signature is modeled firstly based on the thermal radiation theory and a multispectral background suppression approach is given. An experimentally justified assumption is made that the probability density functions (PDFs) of the feature vector can be modeled as Gaussian random process, and then a new unifying radiation intensity and radiation spectral signature (URIS) detector is developed. Finally, performance analyses based on a set of multispectral imagery and receiver operating characteristic (ROC) curves are presented. According to the experimental results, the URIS method can successfully detect dim point target in rather low signal-to-noise condition.