Summary With the development of deep‐learning technology, how to prevent signal modulation from being correctly classified by deep‐learning‐based intruders becomes a challenging issue. Adversarial attack provides an ideal solution as deep‐learning models are proved to be vulnerable to intentionally designed perturbations. However, applying adversarial attacks to communication systems faces several practical problems such as shift‐invariant, imperceptibility, and bandwidth compatibility. To this end, a shift‐invariant universal adversarial attack approach is proposed in this work for misleading deep‐learning‐based modulation classifiers used by intruders. Specifically, this work first introduces a convolutional neural network (CNN)‐based UAP (universal adversarial perturbation) generation model that contains an finite impulse response (FIR) filter layer to control the bandwidth of the output perturbation. Besides, this work proposes a circular shift scheme that simulates the random signal cropping in the inference phase and thus ensures the shift‐invariant property of adversarial perturbations. In addition, this work designs a composite loss function that improves the imperceptibility of the adversarial perturbation in both time and frequency domains without decreasing the effectiveness of the adversarial attack. Experimental results demonstrate the effectiveness of the proposed approach, achieving about 50 % accuracy drop on the target model when the perturbation‐to‐signal ratio (PSR) is −10 dB. Furthermore, extensive experiments are conducted to validate the shift‐invariant, imperceptibility, bandwidth compatibility, and transferability of the proposed approach for modulation classification tasks.

Estimating the parameters of Frequency-Hopping (FH) signals using deep learning has attracted keen attentions in recent years. However, it is also a challenging issue as it has inadequate generalization capacity and requires a large number of annotations for network training. To overcome these limitations, we first introduce a deep learning based time-frequency ridge detection and feature generation framework which can detect the time-frequency ridges and generate separable and discriminative features for model generalization. Then, we propose a few-shot learning strategy according to the temporal relationship between adjacent frames, aiming at reducing the dependence on the number of annotations for network training. Extensive experiments demonstrate that our proposed approach can robustly estimate the parameters of multiple superimposed FH signals under noisy electromagnetic environments with only a few annotations.

随着频谱设备和电磁技术的广泛使用,电磁环境变得日益复杂,对国家电磁环境安全形成巨大的挑战.在复杂环境下进行电磁信号识别,关乎对电磁环境的态势感知,但是在实际应用中往往缺少质量高、数量多的电磁信号样本训练网络.针对这一问题,提出了基于数据增强和集成学习的小样本电磁信号识别技术,对切片处理、时间序列增强和对抗训练增强等技术进行了研究.通过实验发现,这三种数据增强方法均能提高小样本信号的识别准确率,最后在对抗训练增强下,集成学习取得了最好的识别效果,达到了96.5％的识别准确率.

The topological design and optimization of metaporous materials is one of the key challenges in the field of sound absorption. Limited by the expensive computational cost, it is particularly disadvantaged when instantaneous multiple designs are required. In recent years, an increasing number of research fields are harnessing machine learning approaches thanks to their experience-free manner and outstanding efficiency. Generative Adversarial Networks (GANs), as a type of machine learning algorithms, enjoy the special benefit of powerful generative capability, making them brilliantly suitable for designing purposes. Additionally, it can fully explore the data distribution space with enormous computational power and create brand new designs. In this work, GANs are newly employed for the topological design of metaporous materials for sound absorption. Trained with numerically prepared data, they successfully propose designs with high-standard broadband absorption performance, verified by simulation and experiment. The designing process is dramatically accelerated by hundreds of times using GANs (100 designs in 4.372 s). This allows GANs to easily provide more structures and configurations, and achieve instantaneous multiple solutions, giving designers more choices to satisfy various constraints such as mass or porosity. In addition, GANs are demonstrated remarkably capable of generating creative configurations and rich local features. This work proposes a new designing principle, illustrates the value of machine learning in guiding the designing and optimizing process in the mechanical world, and opens new possibilities for the future of AI-materials interdisciplinary research.

Airborne sound absorption coefficient is the premise for investigating the sound absorption performance or mechanism of metaporous materials. The common numerical evaluation approach is FEM which is relatively computationally costly particularly when processing complex structures or a large batch of data. Rapidly developing deep learning algorithms, on the other hand, show a promising trend in the data-driven manner to learn and predict material parameters efficiently and precisely. We propose SAP-net based on deep convolutional neural network to predict the sound absorption coefficient at a specific frequency of an input image representing the topological structure of metaporous materials. Trained with FEM-prepared data for six frequency points, SAP-net demonstrates outstanding evaluation speed of 0.007 s/image and brilliant prediction accuracy with mean absolute errors all smaller than 0.019 (the smallest 0.008 at f = 1000 Hz). Meanwhile, the fact that SAP-net remains accurate when predicting for images that are essentially different from those in the training data shows its capability of learning and capturing the underlying physical mechanism linking the topological structure to the sound absorption performance. In conclusion, SAP-net provides an extraordinarily fast and accurate approach for the investigation of sound absorption performance, which is expected to accelerate the examination and design process of materials.

Learning monocular depth in a self-supervised manner is desirable for numerous applications ranging from autonomous driving, robotics to augmented reality. However, the current challenges lie in the problems of scale ambiguity, dynamic scene and hardware limitations. To this end, a self-supervised approach is proposed in this work for monocular depth learning and estimation. Specifically, we first introduce a self-supervised depth learning framework that learns a part of the camera intrinsics and the stereo extrinsics, which ensures the absoluteness of the predicted depth while achieving enhanced performance (i.e., lower error between the predicted depth and the ground-truth) on depth estimation. Besides, we further improve the accuracy and the efficiency (i.e., shorter inference time and lower weight/activation footprints) via a specially-designed network that exploits multi-scale context across multi-level feature maps. In addition, we propose a quantization scheme for our depth estimation networks. The scheme allows the network inference to be carried out using INT4-INT8 arithmetic while keeping a high performance on depth estimation. Extensive experiments on KITTI and Make3D datasets demonstrate that our approach substantially boosts the performance compared to the existing state-of-the-art methods on monocular depth estimation.

Object detection using FMCW (Frequency-modulated continuous wave) radars is of massive importance for the advanced driver assistance systems. However, it is exceptionally challenging due to the diversity of the electromagnetic environment and the existence of the class imbalance in the radar data space. In this paper, we propose a cascaded object detection network to achieve accurate object detection using FMCW radars. Consisting of a ROI generation stage and a final detection stage, the proposed cascaded network can tackle the problem of the class imbalance and detect objects from the range-Doppler or range-velocity space effectively. Besides, we propose a range-velocity regression procedure to improve the performance of the range-velocity localization. Extensive simulation experiments demonstrate that our proposed approach can robustly detect objects from noisy electromagnetic environments with a high localization accuracy.

This work addresses camera selection, the task of predicting which camera should be "on air" from multiple candidate cameras for soccer broadcast. The task is challenging because of the scarcity of learning data with all candidate views. Meanwhile, broadcast videos are freely available on the Internet (e.g. Youtube). However, these videos only record the selected camera views, omitting the other candidate views. To overcome this problem, we first introduce a random survival forest (RSF) method to impute the incomplete data effectively. Then, we propose a spatial-appearance heatmap to describe foreground objects (e.g. players and balls) in an image. To evaluate the performance of our system, we collect the largest-ever dataset for soccer broadcasting camera selection. It has one main game which has all candidate views and twelve auxiliary games which only have the broadcast view. Our method significantly outperforms state-of-the-art methods on this challenging dataset. Further analysis suggests that the improvement in performance is indeed from the extra information from auxiliary games.

•We propose an end-to-end framework to jointly detect players and classify their team membership.•In the proposal network, we design a multi-branch network to effectively generate candidate image patches with different aspect ratios.•In the classification network, we design a cascaded CNN and a joint learning objective so that the learned model can quickly reject negative examples and output accurate team membership.•Our trained model is very compact (less than 100KB). It saves 1000 time in memory compared with previous methods without sacrificing performance. It is also very efficient in testing (about 10 fps for images of 1280 × 720 with un-optimized Matlab code).

Object detection is important in car sharing services. Accuracy, efficiency, and low memory consumption are desirable for object detection in car sharing services. This paper presents a network system that satisfies all these requirements. Our approach first divides the object detection task into multiple simpler local regression tasks. Then, we propose the generalized Haar filter-based convolutional neural network to reduce the consumption of memory and computing resource. To achieve real-time performance, we introduce a sparse window generation strategy to reduce the number of input image patches without sacrificing accuracy. We perform experiments on both vehicle and pedestrian data sets. Experimental results demonstrate that our approach can accurately detect objects under challenging conditions. Note to Practitioners—Object detection is an important part of intelligent vehicle technologies, which play an important role in car sharing services. Object detection provides metadata for collision avoidance, self-driving systems, and driver-assistance systems, which can result in better safety and consumer experiences in car sharing services. Although deep learning has achieved an excellent performance in object detection, they consume a large amount of storage and computing resource, which makes them difficult to be deployed for car sharing services. This paper suggests a novel approach which is based on the generalized Haar filter and the local regression strategy. Our approach is accurate, efficient, and light. The experimental results verify the effectiveness of the proposed approach in car sharing services.

Vision-based object detection is essential for a multitude of robotic applications. However, it is also a challenging job due to the diversity of the environments in which such applications are required to operate, and the strict constraints that apply to many robot systems in terms of run-time, power and space. To meet these special requirements of robotic applications, we propose an efficient deep network for vision-based object detection. More specifically, for a given image captured by a robot mount camera, we first introduce a novel proposal layer to efficiently generate potential object bounding-boxes. The proposal layer consists of efficient on-line convolutions and effective off-line optimization. Afterwards, we construct a robust detection layer which contains a multiple population genetic algorithm-based convolutional neural network (MPGA-based CNN) module and a TLD-based multi-frame fusion procedure. Unlike most deep learning based approaches, which rely on GPU, all of the on-line processes in our system are able to run efficiently without GPU support. We perform several experiments to validate each component of our proposed object detection approach and compare the approach with some recently published state-of-the-art object detection algorithms on widely used datasets. The experimental results demonstrate that the proposed network exhibits high efficiency and robustness in object detection tasks.

Vision-based object detection is one of the fundamental functions in numerous traffic scene applications such as self-driving vehicle systems and advance driver assistance systems (ADAS). Meanwhile, it also poses to be a demanding task due to the diversity of traffic scenes and resource limitations of the platforms for traffic scene applications. To address these issues, we present a generalized Haar filter based CNN (Convolutional Neural Network) which is suitable for the object detection tasks in traffic scenes. In this approach, we first decompose an object detection task into multiple local regression tasks. Thereafter, we handle these local regression tasks using several light and efficient networks which simultaneously output the bounding boxes, categories and confidence scores of detected objects. To reduce the consumption of storage and computing resources, the weights of these deep networks are constrained to the form of generalized Haar filters. Finally, we carry out various experiments to evaluate the performance of our proposed approach in traffic scene datasets. Experimental results demonstrate that our object detection system is light and effective in comparison with the state-of-the-art.

Vision based player detection is important in sports applications. Accuracy, efficiency, and low memory consumption are desirable for real-time tasks such as intelligent broadcasting and automatic event classification. In this paper, we present a cascaded convolutional neural network (CNN) that satisfies all three of these requirements. Our method first trains a binary (player/non-player) classification network from labeled image patches. Then, our method efficiently applies the network to a whole image in testing. We conducted experiments on basketball and soccer games. Experimental results demonstrate that our method can accurately detect players under challenging conditions such as varying illumination, highly dynamic camera movements and motion blur. Comparing with conventional CNNs, our approach achieves state-of-the-art accuracy on both games with 1000x fewer parameters (i.e., it is light}.

Vision-based object detection is one of the fundamental functions in numerous traffic scene applications such as self-driving vehicle systems and advance driver assistance systems (ADAS). However, it is also a challenging task due to the diversity of traffic scene and the storage, power and computing source limitations of the platforms for traffic scene applications. This paper presents a generalized Haar filter based deep network which is suitable for the object detection tasks in traffic scene. In this approach, we first decompose a object detection task into several easier local regression tasks. Then, we handle the local regression tasks by using several tiny deep networks which simultaneously output the bounding boxes, categories and confidence scores of detected objects. To reduce the consumption of storage and computing resources, the weights of the deep networks are constrained to the form of generalized Haar filter in training phase. Additionally, we introduce the strategy of sparse windows generation to improve the efficiency of the algorithm. Finally, we perform several experiments to validate the performance of our proposed approach. Experimental results demonstrate that the proposed approach is both efficient and effective in traffic scene compared with the state-of-the-art.

Road detection is an essential component of field robot navigation systems. Vision sensors play an important role in road detection for their great potential in environmental perception. In this paper, we propose a hierarchical vision sensor-based method for robust road detection in challenging road scenes. More specifically, for a given road image captured by an on-board vision sensor, we introduce a multiple population genetic algorithm (MPGA)-based approach for efficient road vanishing point detection. Superpixel-level seeds are then selected in an unsupervised way using a clustering strategy. Then, according to the GrowCut framework, the seeds proliferate and iteratively try to occupy their neighbors. After convergence, the initial road segment is obtained. Finally, in order to achieve a globally-consistent road segment, the initial road segment is refined using the conditional random field (CRF) framework, which integrates high-level information into road detection. We perform several experiments to evaluate the common performance, scale sensitivity and noise sensitivity of the proposed method. The experimental results demonstrate that the proposed method exhibits high robustness compared to the state of the art.

Road detection is a crucial problem for autonomous navigation system (ANS) and advance driver-assistance system (ADAS). In this paper, we propose a hierarchical road detection method for robust road detection in challenging scenarios. Given an on-board road image, we first train a Gaussian mixture model (GMM) to obtain road probability density map (RPDM), and next oversegment the image into superpixels. Based on RPDM and superpixels, initial seeds are selected in an unsupervised way, and the seed superpixels iteratively try to occupy their neighbors according to GrowCut framework, the road segment is obtained after convergency. Finally, we refine the road segment with a conditional random field (CRF), which enforces the shape prior on the road segmentation task. Experiments on two challenging databases demonstrate that the proposed method exhibits high robustness compared with the state-of-the-art.

This paper presents a hierarchical model for robust image representation. We first introduce multi-level sparse coding algorithm and normalized max pooling strategy which are designed to obtain meaningful sparse codes and robust pooled codes, respectively. With the sparse codes and pooled codes, a hierarchical architecture is built and more robust features are extracted at the second layer. The proposed method has been evaluated on two widely used datasets: Caltech-101 and Caltech-256, and experimental results demonstrate that the proposed method is both effective and robust in image representation compared with the state-of-the-art.

In this paper, we propose a robust method for the fast estimation of the road vanishing point in challenging scenarios based on multiple population genetic algorithm (MPGA). The proposed method consists of three main parts: the first part is a searching method which searches for vanishing point candidates by multiple population genetic algorithm, the second part is a locally tangent-based voting method which obtains the value of the fitness function for the first part, and the last part is a local dominant texture orientation estimation method which applies Gabor filter banks to estimate the local dominant orientation of the vanishing point candidate and its voters for the second part. The proposed method has been tested on a vanishing point dataset which contains over 200 various road images. The experimental results demonstrate that the proposed method is both efficient and effective in detecting vanishing point compared with some state-of-the-art methods.