Fast 3D Reconstruction of Space Targets From ISAR Image Sequences based on Neural Network

Traditional 2D ISAR image suffer from an inevitable loss of target imformation. It can not reflect target's altitude information and attitude information. To solve this problem, an ISAR 3D reconstruction network based on Instant NGP, named Instant RaNGP is proposed in this paper. Firstly, ISAR imaging principle and projection plane are analysed. The network input rays are constucted accoding to the projection relationship. Second the input rays are fed into the network for training, and the network is accelerated using multiresolution hash encoding as well as ray marching strategies. Finally, the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$3\mathrm{D}$</tex> point cloud of the target can be retrieved from the learned volume density. Compared to the existing state-of-the-art ISAR <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$3\mathrm{D}$</tex> reconstruction methods RaNeRF, it takes less time to train. Experimental results validate the effectiveness of the proposed method.