Radio frequency (RF) fingerprinting is a challenging and important technique in individual identification of wireless devices. Recent work has used deep learning-based classifiers on ADS-B signal without missing aircraft ID information. However, traditional methods are difficult to obtain well performance accuracy for classical deep learning methods to recognize RF signals. This letter proposes a Gaussian Low-pass Channel Attention Convolution Network (GLCA-Net), where a Gaussian Low-pass Channel Attention module (GLCAM) is designed to extract fingerprint features with low frequency. Particularly, in GLCAM, we design a Frequency-Convolutional Global Average Pooling (F-ConvGAP) module to help channel attention mechanism learn channel weights in frequency domain. Experimental results on the datasets of large-scale real-world ADS-B signals show that our method can achieve an accuracy of 92.08%, which is 6.21% higher than Convolutional Neural Networks.

Automatic dependent surveillance-broadcast (ADS-B) has been widely used due to its low cost and high precision. The deep learning methods for ADS-B signal classification have achieved a high performance. However, recent studies have shown that deep learning networks are very sensitive and vulnerable to small noise. We propose an ADS-B signal poisoning method based on U-Net. This method can generate poisoned signals. We assign one of ADS-B signal classification networks as the attacked network and another one as the protected network. When poisoned signals are fed into these two well-performed classification networks, the poisoned signal will recognized incorrectly by the attacked network while classified correctly by the protected network. We further propose an Attack-Protect-Similar loss to achieve “triple-win” in leading attacked network poor performance, protected network well performance and the poisoned signals similar to unpoisoned signals. Experimental results show attacked network classifies poisoned signals with a 1.55% classification accuracy, while the protected network classifies rate is still maintained at 99.38%.

Infrared camera array has been widely used in computer vision, remote sensing and computational Imaging. Due to the lack of a infrared camera array imaging model and a high-quality calibration object, high-quality calibration of the intrinsic and extrinsic parameters of the infrared array camera is a challenge. This letter proposes a calibration method for Infrared camera array. Firstly, we propose an infrared camera array imaging model which contains intrinsic, extrinsic and distortion parameters of all sub-cameras of the infrared camera array. Then, we propose an objective function for infrared camera array calibration and use a non-linear algorithm to solve all parameters in infrared camera array calibration model. To get stable infrared checkboard images, we design a constant temperature heating checkboard. Experiment results show our method can efficiently solve intrinsic, extrinsic and distortion parameters with a low reprojection error and an excellent refocusing performance.