基于宏微观因素的概率级别的车辆事故预测

车辆事故预测是避免道路车辆事故发生的重要研究课题.以往的研究使用的事故数据集只包含地理情况、环境情况、交通情况等宏观因素,或者只包含车辆行为和驾驶员行为等微观因素.因为很难收集到同时包含2类因素的事故数据集,很少有研究将这2类因素结合起来,然而车辆事故往往是两者共同作用的结果.此外,在收集到的数据中没有可以用于预测的事故发生概率标签,所以目前多数的研究关注点只是在于事故是否发生而不能得到准确的概率值.然而在实际应用场景下,驾驶员需要的是不同级别的危险预警信号,而这种信号正是应该由事故概率值决定的.2019年发布的事故宏观因素数据集OSU(Ohio State University)与宏观因素数据集FARS(fatality analysis reporting system)和微观因素数据集SHRP2(strategic highway research program 2)都具有一些相同的特征,为它们的融合提供了机遇.因此,首先得到了一个同时包含宏观和微观因素的数据集,其中事故数据(正样本)融合自OSU、FARS数据集,以及与SHRP2分布相同的数据集Sim-SHRP2(simulated strategic highway research program 2),而安全驾驶数据(负样本)则由自己驾驶汽车获得.然后,针对收集到的数据中没有概率标签的问题,还设计了一个概率级别的无监督深度学习框架来预测准确的概率值,该框架使用迭代的方式为数据集生成准确的概率标签,并使用这些概率标签来进行训练.实验结果表明,该框架可以使用所得到的数据集来灵敏而准确地预测车辆事故.     

基于深度卷积神经网络的视觉里程计研究

视觉里程计利用视频信息来估计相机运动的位姿参数,实现对智能体的定位。传统视觉里程计方法需要特征提取、特征匹配/跟踪、外点剔除、运动估计、优化等流程,解算非常复杂,因此,提出了基于卷积神经网络的方法来实现端对端的单目视觉里程计。借助卷积神经网络对彩色图片自动学习提取图像帧间变化的全局特征,将用于分类的卷积神经网络转化为帧间时序特征网络,通过三层全连接层输出相机的帧间相对位姿参数。在KITTI数据集上的实验结果表明,提出的Deep-CNN-VO模型可以较准确地估计车辆的运动轨迹,证明了方法的可行性。在简化了复杂模型的基础上,与传统的视觉里程计系统相比,该模型的精度也有所提高。     

PET-Derived Radiomics and Artificial Intelligence in Breast Cancer: A Systematic Review

Breast cancer (BC) is a heterogeneous malignancy that still represents the second cause of cancer-related death among women worldwide. Due to the heterogeneity of BC, the correct identification of valuable biomarkers able to predict tumor biology and the best treatment approaches are still far from clear. Although molecular imaging with positron emission tomography/computed tomography (PET/CT) has improved the characterization of BC, these methods are not free from drawbacks. In recent years, radiomics and artificial intelligence (AI) have been playing an important role in the detection of several features normally unseen by the human eye in medical images. The present review provides a summary of the current status of radiomics and AI in different clinical settings of BC. A systematic search of PubMed, Web of Science and Scopus was conducted, including all articles published in English that explored radiomics and AI analyses of PET/CT images in BC. Several studies have demonstrated the potential role of such new features for the staging and prognosis as well as the assessment of biological characteristics. Radiomics and AI features appear to be promising in different clinical settings of BC, although larger prospective trials are needed to confirm and to standardize this evidence.     

ETLi: Efficiently annotated traffic LiDAR dataset using incremental and suggestive annotation

Autonomous driving requires a computerized perception of the environment for safety and machine-learning evaluation. Recognizing semantic information is difficult, as the objective is to instantly recognize and distinguish items in the environment. Training a model with real-time semantic capability and high reliability requires extensive and specialized datasets. However, generalized datasets are unavailable and are typically difficult to construct for specific tasks. Hence, a light detection and ranging semantic dataset suitable for semantic simultaneous localization and mapping and specialized for autonomous driving is proposed. This dataset is provided in a form that can be easily used by users familiar with existing two-dimensional image datasets, and it contains various weather and light conditions collected from a complex and diverse practical setting. An incremental and suggestive annotation routine is proposed to improve annotation efficiency. A model is trained to simultaneously predict segmentation labels and suggest class-representative frames. Experimental results demonstrate that the proposed algorithm yields a more efficient dataset than uniformly sampled datasets.     

PartitionTuner: An operator scheduler for deep-learning compilers supporting multiple heterogeneous processing units

Recently, embedded systems, such as mobile platforms, have multiple processing units that can operate in parallel, such as centralized processing units (CPUs) and neural processing units (NPUs). We can use deep-learning compilers to generate machine code optimized for these embedded systems from a deep neural network (DNN). However, the deep-learning compilers proposed so far generate codes that sequentially execute DNN operators on a single processing unit or parallel codes for graphic processing units (GPUs). In this study, we propose PartitionTuner, an operator scheduler for deep-learning compilers that supports multiple heterogeneous PUs including CPUs and NPUs. PartitionTuner can generate an operator-scheduling plan that uses all available PUs simultaneously to minimize overall DNN inference time. Operator scheduling is based on the analysis of DNN architecture and the performance profiles of individual and group operators measured on heterogeneous processing units. By the experiments for seven DNNs, PartitionTuner generates scheduling plans that perform 5.03% better than a static type-based operator-scheduling technique for SqueezeNet. In addition, PartitionTuner outperforms recent profiling-based operator-scheduling techniques for ResNet50, ResNet18, and SqueezeNet by 7.18%, 5.36%, and 2.73%, respectively.     

On flow regime transition in trickle bed: Development of a novel deep-learning-assisted image analysis method

An image analysis method was developed based on deep-learning algorithms to extract phase fractions quantitatively in a rectangular trickle bed, and the average identification error was lower than 5%. Furthermore, the flow regime transition in the trickle bed was studied. In trickle-to-pulse flow transition, the trickle flow could be further classified into the stable trickle flow and accelerated one. The SD of liquid fractions and the peak width at half-height of the probability density curve of liquid fractions were close to zero in stable trickle flow, increased rapidly in accelerated trickle flow, and remained approximately constant in pulse flow. In bubble-to-pulse flow transition, dispersed bubbles in bubble flow induced the outliers outside the upper boundary of the boxplot of gas fraction, while alternative appearance of gas-rich zone and liquid-rich zone in pulse flow induced outliers outside both the upper and lower boundaries of the boxplot of gas fraction.     

焊接领域激光结构光视觉传感技术的研究及应用

为保证自动焊接的质量,提高焊缝识别的准确性和适应性,提出一种焊缝特征提取的关键点检测方法.基于卷积神经网络设计了焊缝特征提取网络,该网络通过卷积、池化操作提取焊缝特征.将来自深层的特征图进行上采样,最后将深层特征图和浅层特征图相融合,提高焊缝特征提取精度.输出焊缝图像的热力图来预测焊缝特征点位置,实现多种坡口焊缝的识别定位,且不需要非极大值抑制算法,提升了特征提取速度.采集不同的焊缝特征图像,进行网络模型训练.结果表明,焊缝特征点定位均方根误差为0.187 mm,网络模型在焊缝特征点识别任务中检测精度较高,而且适应性和泛化性较强,满足自动焊接的要求.

     

基于NVIDIA Jetson TX2的道路场景分割

图像语义分割是计算机视觉领域重要研究方向之一,其中基于深度学习的语义分割相较于传统分割算法更为高效可靠,可应用于交通监控、自动驾驶等领域的场景理解阶段.但复杂的分割网络在嵌入式平台上的推理速度较低,难以进行实际应用.因此针对交通监控、无人驾驶等应用背景,在嵌入式平台NVIDIA Jetson TX2上,采用基于深度卷积编解码器结构的图像分割网络,对道路场景进行语义分割,并基于NVIDIA的推理加速器TensorRT2,完成网络模型简化、网络自定义层添加与CUDA并行优化,实现了对网络推理阶段的加速.实验结果表明,加速引擎在TX2上的推理速度约为原模型的10倍,为复杂分割网络在嵌入式平台上的应用提供了支持.     

生成对抗网络辅助学习的舰船目标精细识别

针对近岸舰船目标细粒度识别的难题,提出了一种利用生成对抗网络辅助学习的任意方向细粒度舰船目标识别框架。通过训练能模仿舰船目标区域的抽象深度特征的生成网络引入生成样本,来辅助分类子网络学习样本空间的流形分布,从而增强细粒度的类别间判别能力。在细粒度类别的近岸舰船数据集上,引入生成对抗网络后的算法识别准确率得到较大提升,平均识别精度提升了2%。消融实验结果表明,利用生成样本辅助训练分类子网络可以有效地提升舰船目标的细粒度识别精度。     

Shift-invariant universal adversarial attacks to avoid deep-learning-based modulation classification

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.