Polsar Image Classification based on Optimal Feature and Convolution Neural Network

PolSAR Image Classification Framework With POA Align and Cyclic Channel Attention

“…To facilitate information interaction in the entire image space, combining a deep neural network-based framework with channel attention holds promise. At present, the dominant framework still relies on convolutional operations, such as CNN [36], [37], FCN [38]- [40], and advanced complexvalued [41] and 3-D networks [42], which have demonstrated commendable performance. However, although the design of multi-level attempts to incrementally enlarge the receptive field to improve global modeling capacity, resource overconsumption also appears.…”