Improved CNN for E2E SAR ATR


Our member proposed an improved convolutional neural network (CNN) for end-to-end synthetic aperture radar (SAR) automatic target recognition (ATR).

For more information on the CNN, it was presented at the 2018 IEICE General Conference held at Tokyo Denki University on March 21.

*IEICE: Institute of Electronics, Information and Communication Engineers

VersNet-v2: Improved CNN for E2E SAR ATR

As with VersNet (verification support network), the VersNet-v2 inputs a SAR image of arbitrary sizes with multiple classes and multiple targets, and outputs a SAR ATR image representing the position, class, and pose of each detected target.

To improve posture estimation performance, VersNet-v2 estimates the score of the combined class of the target class and pose class, and maps it to the target class and pose class in postprocessing.