Search

Abstract In civil aviation industry, runway overrun is a typical landing safety incident concerned by both airlines and authorities. Among various contributing factors to the runway overrun incident, long landing plays an important role. However, existing studies for long landing prediction mainly depend on classic machine learning methods and handcrafted features. As a result, they usually require much expert knowledge and provide unsatisfactory results. To address these problems, this paper proposes an innovative deep sequence‐to‐sequence model which utilizes QAR (Quick Access Recorder) data for accurate long landing pre‐ diction. Specifically, to cope with the high heterogeneity of QAR dataset, a data pre‐processing procedure is first proposed which includes data cleaning, interpolation and normalization steps. Second, to avoid the noises incurred by too many QAR parameters and relieve the reliance on expert experience, the GBDT (gradient boosting decision trees) model is employed to choose the most relevant parameters as features. Then a CNN‐LSTM and TG‐attention encoder‐decoder architecture is proposed to accurately predict future aircraft ground speed and radio height sequences, based on which the touchdown distance can be finally calculated. Experimental results on a large QAR dataset with 44,176 A321 flights validate effectiveness of the proposed method.

Accurate fast hand detection and gesture recognition for hand understanding are still challenging tasks that are influenced by the diversity of hands and the complexity of the scene in color images. To address the above problem, we propose a novel SqueezeNet and fusion network-based fully convolutional network (SF-FCNet) to accurately and quickly perform hand detection and gesture recognition in color images. First, we introduce the first 17-layer structure in the lightweight SqueezeNet as the hand feature extraction network to accelerate the detection and recognition speed by greatly compressing the network parameters. Second, a precise hand prediction fusion network is designed by adding a residual structure to the deconvolutional network to integrate high- and low-level features of hands, and hand detection and gesture recognition are performed on a single convolutional layer at multiple scales to improve the precision and reduce the computational costs. The verification results on the Oxford hand dataset show that SF-FCNet can reach a precision of 84.1% and a speed of 32 FPS. The experimental results show that SF-FCNet can substantially enhance the precision and speed of hand detection and gesture recognition on three benchmark datasets and has a strong generalization ability on a homemade test set.

The issue of employee turnover is always critical for companies, and accurate predictions can help them prepare in time. Most past studies on employee turnover have focused on analyzing impact factors or using simple network centrality measures. In this paper, we study the problem from a completely new perspective by modeling users' historical job records as a dynamic bipartite graph. Specifically, we propose a bipartite graph embedding method with temporal information called dynamic bipartite graph embedding (DBGE) to learn the vector representation of employees and companies. Our approach not only considers the relations between employees and companies but also incorporates temporal information embedded in consecutive work records. We first define the Horary Random Walk on a bipartite graph to generate a sequence for each vertex in chronological order. Then, we employ the skip-gram model to obtain a temporal low-dimensional vector representation for each vertex and apply machine learning methods to predict employee turnover behavior by combining embedded features with employees' basic information. Experiments on a real-world dataset collected from one of China's largest online professional social networks show that the features learned through DBGE can significantly improve turnover prediction performance. Moreover, experiments on public Amazon and Taobao datasets show that our approach achieves better performance in the link prediction and visualization task than other graph embedding methods that do not consider temporal information.

Learning based hashing has been widely used in approximate nearest neighbor search for image retrieval. However, most of the existing hashing methods are designed to learn only simplex feature similarity while ignored the location similarity among multiple objects, thus cannot work well on multi-label image retrieval tasks. In this paper, we propose a novel supervised hashing method which fusions the two kinds of similarities together. First, we leverage an adjacency matrix to record the relative location relationship among multiple objects. Second, by incorporating matrix discretization difference and image label difference, we re-define the pairwise image similarity in a more meticulous way. Third, to learn more distinguishable hash codes, we leverage an attention sub-network to identify the approximate regions of the objects in an image so that the extracted features can mainly focus on the foreground objects and ignore the background clutter. The loss function in our method consists of a multi-categories classification loss which is used to learn the attention sub-network and a hash loss with a scaled sigmoid function which is used to learn the efficient hash codes. Experiment results show that our proposed method is effective in preserving high-level similarities and outperforms the baseline methods in multi-label image retrieval.

The problem of influence maximization (IM) has been extensively studied in recent years and has many practical applications such as social advertising and viral marketing. Given the network and diffusion model, IM aims to find an influential set of seed nodes so that targeting them as diffusion sources will trigger the maximum cascade of influenced individuals. The largest challenge of the IM problem is its NP-hardness, and most of the existing approaches are with polynomial time complexity, making themselves unscalable to very large networks. To address this issue, in this paper, we propose LAIM: a linear time iterative approach for efficient IM on large-scale networks. Our framework has two steps: 1) influence approximation and 2) seed set selection. In the first step, we propose an iterative algorithm to compute the local influence of a node based on a recursive formula and use the local influence to approximate its global influence. In the second step, the k influential seed nodes are mined based on the approximated influence in the first step. Based on our model, we theoretically prove that the proposed approach has linear time and space complexity. We further accelerate our algorithm with simple modifications and propose its fast version. Experimental results on eight real-world large-scale networks exhibit the superiority of our approach over the state-of-the-art methods in terms of both effectiveness and efficiency.

Real-time 3D reconstruction is one of the current popular research directions of computer vision, and it has become the core technology in the fields of virtual reality, industrialized automatic systems, and mobile robot path planning. Currently, there are three main problems in the real-time 3D reconstruction field. Firstly, it is expensive. It requires more varied sensors, so it is less convenient. Secondly, the reconstruction speed is slow, and the 3D model cannot be established accurately in real time. Thirdly, the reconstruction error is large, which cannot meet the requirements of scenes with accuracy. For this reason, we propose a real-time 3D reconstruction method based on monocular vision in this paper. Firstly, a single RGB-D camera is used to collect visual information in real time, and the YOLACT++ network is used to identify and segment the visual information to extract part of the important visual information. Secondly, we combine the three stages of depth recovery, depth optimization, and deep fusion to propose a three-dimensional position estimation method based on deep learning for joint coding of visual information. It can reduce the depth error caused by the depth measurement process, and the accurate 3D point values of the segmented image can be obtained directly. Finally, we propose a method based on the limited outlier adjustment of the cluster center distance to optimize the three-dimensional point values obtained above. It improves the real-time reconstruction accuracy and obtains the three-dimensional model of the object in real time. Experimental results show that this method only needs a single RGB-D camera, which is not only low cost and convenient to use, but also significantly improves the speed and accuracy of 3D reconstruction.

Indoor autonomous navigation refers to the perception and exploration abilities of mobile agents in unknown indoor environments with the help of various sensors. It is the basic and one of the most important functions of mobile agents. In spite of the high performance of the single-sensor navigation method, multi-sensor fusion methods still potentially improve the perception and navigation abilities of mobile agents. This work summarizes the multi-sensor fusion methods for mobile agents’ navigation by: (1) analyzing and comparing the advantages and disadvantages of a single sensor in the task of navigation; (2) introducing the mainstream technologies of multi-sensor fusion methods, including various combinations of sensors and several widely recognized multi-modal sensor datasets. Finally, we discuss the possible technique trends of multi-sensor fusion methods, especially its technique challenges in practical navigation environments.

In recent years, increasing image data comes from various sensors, and object detection plays a vital role in image understanding. For object detection in complex scenes, more detailed information in the image should be obtained to improve the accuracy of detection task. In this paper, we propose an object detection algorithm by jointing semantic segmentation (SSOD) for images. First, we construct a feature extraction network that integrates the hourglass structure network with the attention mechanism layer to extract and fuse multi-scale features to generate high-level features with rich semantic information. Second, the semantic segmentation task is used as an auxiliary task to allow the algorithm to perform multi-task learning. Finally, multi-scale features are used to predict the location and category of the object. The experimental results show that our algorithm substantially enhances object detection performance and consistently outperforms other three comparison algorithms, and the detection speed can reach real-time, which can be used for real-time detection.

In recent years, increasing human data comes from image sensors. In this paper, a novel approach combining convolutional pose machines (CPMs) with GoogLeNet is proposed for human pose estimation using image sensor data. The first stage of the CPMs directly generates a response map of each human skeleton’s key points from images, in which we introduce some layers from the GoogLeNet. On the one hand, the improved model uses deeper network layers and more complex network structures to enhance the ability of low level feature extraction. On the other hand, the improved model applies a fine-tuning strategy, which benefits the estimation accuracy. Moreover, we introduce the inception structure to greatly reduce parameters of the model, which reduces the convergence time significantly. Extensive experiments on several datasets show that the improved model outperforms most mainstream models in accuracy and training time. The prediction efficiency of the improved model is improved by 1.023 times compared with the CPMs. At the same time, the training time of the improved model is reduced 3.414 times. This paper presents a new idea for future research.