Rectangular structure extraction by morphological decomposition and primitives grouping

In this paper, we present a method for automatic detection of rectangular structures. This method is based, first on linear and directional morphological transformations to detect linear structures, second on the notion of primitive grouping to detect the rectangular structures. We applied this method to a problem of numbered label recognition. Tests have been made on images of an experimental ant-heap in which ants have a numbered label on their back. Ethologists take pictures of this ant-heap every twenty minutes and analyse it to have a better understanding of the behaviour of the ants. The aim is to help them to analyse this information. Moreover, these images are very noisy, and the tests show the robustness of the method against noise