Image fusion provides an attractive technical solution for surveillance applications, offering improved task performance for imaging systems. Through the fusion process, information from different spectral bands is combined to form a single video stream which gives an improved scene understanding together with an enhanced target detection and recognition capability. In this paper, ground-based surveillance requirements are used to define a tri-band image fusion camera comprising LWIR, NIR, and Visible spectral bands. The design approach used addresses both the performance and commercialisation challenges encountered in the development of an image fusion camera. These challenges are discussed in the context of earlier image fusion systems where useful lessons were learnt. Establishing an effective processing architecture is critical to image interpretation, and the functional design is presented. The tri-band camera design allows the user to view different image streams including enhanced single-band image data as well as both dual and tri-band fused imagery. Such flexibility enables the selection of the best imagery for specific scenarios and viewing conditions. The physical characteristics are a major constraint for handheld camera designs where the size, weight, and power limitations dictate both the choice of sensors as well as the processor card. For the design presented here, power consumption and latency figures are minimised using relatively simple arithmetical fusion algorithms which are combined with an adaptive colour weight map for regional optimisation. Example results are presented to illustrate the various technical challenges and trade-offs undertaken within this development programme.
|