Surveillance mission systems acquire data, which can then be interpreted by humans or by automatic methods to meet various military or civilian needs. In order to meet ever-increasing needs, these systems must evolve, in particular through the use of new types of sensors and imagery. Hyperspectral imaging is used to estimate the presence and position of chemical compositions in images, making it possible to discriminate and locate materials over long distances. Over the last few decades, hyperspectral imaging has become much more accessible, making it a prime candidate for the evolution of surveillance mission systems. In this thesis, we focus on the problems of detecting, discriminating and recognising materials in hyperspectral images acquired under a variety of acquisition conditions on the ground or at low altitude. We have acquired two large databases of hyperspectral images of land, vehicles and materials of interest to surveillance mission systems, under a wide range of illumination conditions. We have proposed several methods for reducing the constraints associated with the calibration of the hyperspectral images acquired. Finally, we have proposed several methods for processing and analysing hyperspectral images acquired under a variety of conditions for the detection and discrimination of materials.