Information Theoretic Measures for Change Detection in Urban Sensing Applications

In through-the-wall radar imaging and surveillance applications, it is important for the imaging system to be able to automatically quantify and detect the changes in the imaged scene without the need for operator interpretation. In previous work [1], we considered two information theoretic measures, entropy and divergence, for this purpose. Preliminary analysis of these measures revealed that they can provide reliable notifications of changes in the scene. In this paper, we expand on this work by introducing two different classes of measures, namely, complexity and difference measures. Complexity measures, which includes entropy, quantify the amount of activity in the given scene. Difference measures, on the other hand, are effective at detecting the changes in the imaged scene. Our results, based on experimental data, show that the ratio of the norms is the most sensitive complexity measure and is useful for discriminating between populated and unpopulated scenes, whereas the Jensen-Renyi divergence measure is the most sensitive difference measure and can be applied for change detection in the scene.