Hidden object detection based on probabilistic fuzzy fusion and fisher vectors in passive millimeter-wave images

Millimeter-wave radiometric imaging technology holds enormous potential for hidden-object detection in the application of security inspection. Existing hidden-object detection methods based on pixel or regional processing have limited detection performance on the low signal-to-clutter ratio or the targets with multiple sizes or irregular shapes. This paper proposes a new detection method combining the probabilistic fuzzy fusion and Fisher vectors processing for hidden object detection in multiple polarization passive millimeter-wave images. Experimental results employing the real data collected from Multiple Polarization Scanning Imaging Radiometer demonstrate that the proposed method improves the detection performance and the extraction accuracy of object contours compared with the state-of-the-art methods.