Compared to other recognition tasks, SAR ATR tasks require a stronger adaptability to background environments. This is because as the maneuverability of targets improves, there may be a wider variety of terrains encountered. Traditional SAR ATR algorithms assume that the background in both training and testing datasets follows the same distribution. This assumption can lead to a significant decrease in recognition performance when targets appear in new terrains or environments. We propose a SAR ATR causal inference framework to eliminate background-related biases. From a causal inference perspective, we review the ATR task, aiming to obtain unbiased predictions from predictions that include background biases. Our framework does not impose specific constraints on the model’s implementation nor does it introduce any new training parameters. Therefore, it has broad application prospects and is expected to facilitate the widespread use of deep neural network models in SAR target recognitio...

In synthetic aperture radar (SAR) automatic target recognition (ATR) tasks, deep learning-based methods usually work with the assumption that the training and test target samples are independent and identically distributed. However, the performance of the deep model degrades dramatically when there exists a large distribution variation between training and test data. The collected target samples include not only the target entity but also the target's complicated surrounding environment. So it is difficult to accurately identify targets when they appear in a new background. In this article, we propose a causal inference framework for SAR ATR by removing the background-related bias. This framework can handle more challenging recognition scenarios, SAR background out of distribution (o.o.d) recognition task. First, the SAR ATR task is modeled as a causal graph from a causal inference perspective, and this graph clearly explains the sources of background-related bias in traditional deep m...