NSANet: Noise-Seeking Attention Network

Abstract

Light detection and ranging technology has remained popular in capturing natural and built environments for numerous applications. The recent technological advancements in electrooptical engineering have aided in obtaining laser returns at a higher pulse-repetition frequency (PRF), which considerably increased the density of the 3-D point cloud. Conventional techniques with lower PRF had a single pulse-in-air zone, large enough to avoid a mismatch among pulse pairs at the receiver. New multiple-pulse-in-air technology guarantees various windows of operational ranges for a single flight line and no blind zones. The disadvantage of the technology is the projection of atmospheric returns closer to the same pulse-in-air zone of adjacent terrain points likely to intersect with objects of interest. These noise properties compromise the perceived quality of the scene and encourage the development of new noise-filtering neural networks, as the existing filters are significantly ineffective. We propose a novel dual-attention noise-filtering neural network called noise-seeking attention network (NSANet) that fuses physical priors and local spatial attention to filter noise. Our research's fusion module is motivated by two psychological theories of feature integration and attention engagement to prove the role of attention in computer vision at the encoding and decoding phase. The presented results of NSANet show the benefit of attentional engagement theory and a performance boost of 7.30% on recall and 4.10% on F1-score compared to the state-of-the-art noise-filtering deep convolutional neural networks.

Keywords

• Attention
• computer vision
• noise-filtering neural network
• physical priors
• point cloud
• systematic noise