Slim and Efficient Neural Network Design for Resource-Constrained SAR Target Recognition

Abstract

Deep convolutional neural networks (CNN) have been recently applied to synthetic aperture radar (SAR) for automatic target recognition (ATR) and have achieved state-of-the-art results with significantly improved recognition performance. However, the training period of deep CNN is long, and the size of the network is huge, sometimes reaching hundreds of megabytes. These two factors of deep CNN hinders its practical implementation and deployment in real-time SAR platforms that are typically resource-constrained. To address this challenge, this paper presents three strategies of network compression and acceleration to decrease computing and memory resource dependencies while maintaining a competitive accuracy. First, we introduce a new weight-based network pruning and adaptive architecture squeezing method to reduce the network storage and the time of inference and training process, meanwhile maintain a balance between compression ratio and classification accuracy. Then we employ weight quantization and coding to compress the network storage space. Due to the fact that the amount of calculation is mainly reflected in the convolution layer, a fast approach for pruned convolutional layers is proposed to reduce the number of multiplication by exploiting the sparsity in the activation inputs and weights. Experimental results show that the convolutional neural networks for SAR-ATR can be compressed by 40 × without loss of accuracy, and the number of multiplication can be reduced by 15 × . Combining these strategies, we can easily load the network in resource-constrained platforms, speed up the inference process to get the results in real-time or even retrain a more suitable network with new image data in a specific situation.

Attention Graph Convolution Network for Image Segmentation in Big SAR Imagery Data

“…C indicates the number of the convolution kernels and is equal to the dimension of the output features of nodes. Similarly, Θ (1) ∈ R C ×F represents the parameter matrix of the second convolutional layer, and F is the dimension of each node's feature in the output. The second convolutional layer outputs…”

“…is the f th element of y i . Parameters → a , Θ (0) and Θ (1) are then updated using the Adam Stochastic Gradient Descent (SGD) optimizer.…”

“…SAR image segmentation is a foundation for many high-level interpretation tasks. Accurate segmentation can greatly reduce the difficulty of subsequent advanced tasks (e.g., target recognition [1,2], target detection [3], change detection [4], etc.). With the development of Synthetic Aperture Radar (SAR) imaging sensors, the resolution of SAR images has increased considerably [5,6], which also brings the explosion of SAR data, termed big SAR imagery data.…”

GPU-Based Soil Parameter Parallel Inversion for PolSAR Data

“…With the increase in the number of global satellites, the application of ground exploration has become increasingly common [6][7][8]. But low resolution is also a major problem for exploration [9,10].…”

A Novel Semi-Supervised Convolutional Neural Network Method for Synthetic Aperture Radar Image Recognition

“…Synthetic Aperture Radar (SAR) has been widely used due to its high resolution and penetrating ability [1][2][3]. SAR image automatic target recognition technology (SAR-ATR) is one of the research hotspots in the field of image cognitive learning [4,5]. Based on the cognitive system, humans are able to recognize targets quickly and accurately.…”