Self-Matching CAM: A Novel Accurate Visual Explanation of CNNs for SAR Image Interpretation

Zhenpeng Feng,Hongbing Ji,Mingzhe Zhu,Ljubiša Stanković

doi:10.3390/rs13091772

Zhenpeng Feng, Hongbing Ji + Show 2 more

Open Access

https://doi.org/10.3390/rs13091772

Copy DOI

Journal: Remote Sensing	Publication Date: May 1, 2021
Citations: 31	License type: CC BY 4.0

Affiliation: Xidian University, University of Montenegro

Abstract

Synthetic aperture radar (SAR) image interpretation has long been an important but challenging task in SAR imaging processing. Generally, SAR image interpretation comprises complex procedures including filtering, feature extraction, image segmentation, and target recognition, which greatly reduce the efficiency of data processing. In an era of deep learning, numerous automatic target recognition methods have been proposed based on convolutional neural networks (CNNs) due to their strong capabilities for data abstraction and mining. In contrast to general methods, CNNs own an end-to-end structure where complex data preprocessing is not needed, thus the efficiency can be improved dramatically once a CNN is well trained. However, the recognition mechanism of a CNN is unclear, which hinders its application in many scenarios. In this paper, Self-Matching class activation mapping (CAM) is proposed to visualize what a CNN learns from SAR images to make a decision. Self-Matching CAM assigns a pixel-wise weight matrix to feature maps of different channels by matching them with the input SAR image. By using Self-Matching CAM, the detailed information of the target can be well preserved in an accurate visual explanation heatmap of a CNN for SAR image interpretation. Numerous experiments on a benchmark dataset (MSTAR) verify the validity of Self-Matching CAM.

Highlights

Synthetic aperture radar (SAR) can produce high-resolution radar images in various extreme weather conditions, such as precipitation, dust, mist, etc., which makes it widely applied in many fields, like topographic mapping, urban planning, traffic monitoring, electronic reconnaissance, etc. [1,2,3,4]
It is worth noting that the original SAR images are gray-scale; to avoid modification of the parameters of AlexNet, all the SAR images are transformed into pseudo-RGB images
Note that the gist of this paper is to probe into this convolutional neural networks (CNNs) to understand what information hidden in the input works on correct classification, but not the relationship between class activation mapping (CAM) effects and complex parameter tuning

Summary

Introduction

Synthetic aperture radar (SAR) can produce high-resolution radar images in various extreme weather conditions, such as precipitation, dust, mist, etc., which makes it widely applied in many fields, like topographic mapping, urban planning, traffic monitoring, electronic reconnaissance, etc. [1,2,3,4]. Note that traditional target recognition technology is composed of multiple individual steps [8,9,10]. Such complex procedures will reduce processing efficiency and make it difficult to realize real-time applications. Deep learning algorithms can allay the aforementioned limitations greatly because deep networks own an end-to-end structure without complex preprocessing operations [11,12]. Such an end-to-end structure can Remote Sens.

Objectives

Methods

Results

Discussion

Conclusion