The use of multimodal or heterogeneous sensors for surveillance greatly increases the diversity of information available from a given region of interest. Since the underlying scene is the same for all the sensors, the data across the sensors are inherently dependent. The nature of this dependence can be quite complex and quantifying it is a challenging task, especially in the case of heterogeneous sensing. We consider the problem of fusion for the detection of dependent, heterogeneous signals and design a detector using a copula-based framework. Past applications using the copula based approach have mostly been limited to the bivariate (2 sensor) case. We will address copula construction and model selection issues for the multivariate case.