Enhancing Near-Field Holographic Imaging with Predicted Object Position

In current near-field holographic imaging, a circular region is scanned using an array of transmitting and receiving antennas over a narrow frequency band. This makes the data acquisition system slow, complex, bulky, and costly. Reducing the number of receiver antennas and using a narrower frequency band can significantly reduce the cost and complexity of the data acquisition system. To do so, we propose a method that uses prior knowledge about the object position, obtained by applying a neural network algorithm, called convolutional neural network (CNN), to the scattered field responses. This prior knowledge is then used to add a new regularization term to the cost function that is minimized in near-field holography.