Single sensor passive millimetre wave (PMMW) imaging systems typically suffer from long acquisition times, which
inhibit their utilization in some security applications (i.e. concealed weapons detection systems) where real time
operation is required. This is inherent to the physical principle behind the system, where the achievable temperature
resolution and the integration time are inversely proportional to each other. The longer the sensor can collect (and
integrate) energy radiated by the scene at each position, the finer the temperature resolution becomes. Reducing the
integration time without degrading the temperature resolution can be achieved by increasing the bandwidth of the
radiometer, but this is possible only up to certain limits.
We propose to reduce the acquisition time in such a single sensor, by combining the recent theory of compressive
sensing with special sparse trajectories, designed to achieve the level of incoherence that the theory of compressive
sensing requires to successfully recover the image. Another alternative proposal is to sample all the pixels in the scene,
but for each pixel, the integration time is randomly selected from a pool of discrete possible values. The radiometric
image is then reconstructed by combining the images obtained from the individual application of compressive sensing to
each group of pixels having the same integration time.
For demonstration purposes, a single pixel PMMW imaging simulator has been implemented in Matlab/Simulink, which
includes a configurable radiometric scene generator. The paper presents results from simulated radiometric scenes at
140GHz acquired with the proposed sparse trajectories and recovered using compressive sensing. The results show that
savings in acquisition times between 50% and 70% are possible while maintaining the required temperature resolution.