Abstract:
Shape variance of target images, image overlapping for adjacent targets, and weak scattering target detection are critical challenges of human target detection for throug...Show MoreMetadata
Abstract:
Shape variance of target images, image overlapping for adjacent targets, and weak scattering target detection are critical challenges of human target detection for through-the-wall radar imaging. In this letter, an adaptive target detection method is proposed based on fully convolutional network (FCN). The downsampling-upsampling structure is employed to extract multiscale features. The attention mechanism is integrated with the FCN for weak scattering target detection. Exploiting both the intensity and geometrical features of the target image, the proposed algorithm could overcome the abovementioned challenges and achieve better detection performance compared with the state-of-the-art methods. The proposed algorithm is evaluated via simulation and experimental tests.
Published in: IEEE Geoscience and Remote Sensing Letters ( Volume: 18, Issue: 9, September 2021)