Intelligent Transportation Applications Research Articles

A fast fused part-based model with new deep feature for pedestrian detection and security monitoring

In recent years, pedestrian detection based on computer vision has been widely used in intelligent transportation, security monitoring, assistance driving and other related applications. However, one of the remaining open challenges is that pedestrians are partially obscured and their posture changes. To address this problem, deformable part model (DPM) uses a mixture of part filters to capture variation in view point and appearance and achieves success for challenging datasets. Nevertheless, the expensive computation cost of DPM limits its ability in the real-time application. This study propose a fast fused part-based model (FFPM) for pedestrian detection to detect the pedestrians efficiently and accurately in the crowded environment. The first step of the proposed method trains six Adaboost classifiers with Haar-like feature for different body parts (e.g., head, shoulders, and knees) to build the response feature maps. These six response feature maps are combined with full-body model to produce spatial deep features. The second step of the proposed method uses the deep features as an input to support vector machine (SVM) to detect pedestrian. A variety of strategies is introduced in the proposed model, including part-based to full-body method, spatial filtering, and multi-ratios combination. Experiment results show that the proposed FFPM method improves the computation speed of DPM and maintains the performance in detection.

An ISVD and SFFSD-based vehicle ego-positioning method and its application on indoor parking guidance

Applications in intelligent transportation systems may not function properly due to loss of positioning precision or reliability in certain GPS-denied environment, e.g., underground parking lots, street surrounded by tall buildings, tunnels, etc. To overcome these challenges, this study proposes a new video-based vehicle ego-positioning method for intelligent transportation applications. Our method follows the procedure of finding feature correspondences from consecutive frames and minimizing its re-projection error. Firstly, we propose an image singular value decomposition algorithm to mitigate the impact of picture blurring. Secondly, a statistic filter of feature space displacement and a circle matching method are proposed to screen or prune potential matched features so as to remove the outliers caused by the mismatching. Thirdly, experiments are conducted on a benchmark dataset KITTI and real outdoor and indoor image data with blur, weak texture or illumination changes to verify the validation of the proposed algorithm. Lastly, the proposed algorithm is successfully applied to indoor parking guidance. The results demonstrate that compared with other state-of-the-art approaches, the proposed video-based vehicle ego-positioning scheme has more precise and reliable performance with a lower cost, which provides an appropriate way for indoor parking guidance.

LaIF: A Lane-Level Self-Positioning Scheme for Vehicles in GNSS-Denied Environments

Vehicle self-positioning is of significant importance for intelligent transportation applications. However, accurate positioning (e.g., with lane-level accuracy) is very difficult to obtain due to the lack of measurements with high confidence, especially in an environment without full access to a global navigation satellite system (GNSS). In this paper, a novel information fusion algorithm based on a particle filter is proposed to achieve lane-level tracking accuracy under a GNSS-denied environment. We consider the use of both coarse-scale and fine-scale signal measurements for positioning. Time-of-arrival measurements using the radio frequency signals from known transmitters or roadside units, and acceleration or gyroscope measurements from an inertial measurement unit (IMU) allow us to form a coarse estimate of the vehicle position using an extended Kalman filter. Subsequently, fine-scale measurements, including lane-change detection, radar ranging from the known obstacles (e.g., guardrails), and information from a high-resolution digital map, are incorporated to refine the position estimates. A probabilistic model is introduced to characterize the lane changing behaviors, and a multi-hypothesis model is formulated for the radar range measurements to robustly weigh the particles and refine the tracking results. Moreover, a decision fusion mechanism is proposed to achieve a higher reliability in the lane-change detection as compared to each individual detector using IMU and visual (if available) information. The posterior Cramer–Rao lower bound is also derived to provide a theoretical performance guideline. The performance of the proposed tracking framework is verified by simulations and real measured IMU data in a four-lane highway.

Separable reverse connected network for efficient multi-scale vehicle detection

Vehicle detection is involved in a wide range of intelligent transportation and smart city applications, and the demand of fast and accurate detection of vehicles is increasing. In this article, we propose a convolutional neural network-based framework, called separable reverse connected network, for multi-scale vehicles detection. In this network, reverse connected structure enriches the semantic context information of previous layers, while separable convolution is introduced for sparse representation of heavy feature maps generated from subnetworks. Further, we use multi-scale training scheme, online hard example mining, and model compression technique to accelerate the training process as well as reduce the parameters. Experimental results on Pascal Visual Object Classes (VOC) 2007 + 2012 and MicroSoft Common Objects in COntext (MS COCO) 2014 demonstrate the proposed method yields state-of-the-art performance. Moreover, by separable convolution and model compression, the network of two-stage detector is accelerated by about two times with little loss of detection accuracy.

Tightly-Coupled Vehicle Positioning Method at Intersections Aided by UWB

Reliable and precise vehicle positioning is essential for most intelligent transportation applications as well as autonomous driving. Due to satellite signal blocking, it can be challenging to achieve continuous lane-level positioning in GPS-denied environments such as urban canyons and crossroads. In this paper, a positioning strategy utilizing ultra-wide band (UWB) and low-cost onboard sensors is proposed, aimed at tracking vehicles in typical urban scenarios (such as intersections). UWB tech offers the potential of achieving high ranging accuracy through its ability to resolve multipath and penetrate obstacles. However, not line of sight (NLOS) propagation still has a high occurrence in intricate urban intersections and may significantly deteriorate positioning accuracy. Hence, we present an autoregressive integrated moving average (ARIMA) model to first address the NLOS problem. Then, we propose a tightly-coupled multi sensor fusion algorithm, in which the fuzzy calibration logic (FCL) is designed and introduced to adaptively adjust the dependence on each received UWB measurement to effectively mitigate NLOS and multipath interferences. At last, the proposed strategy is evaluated through experiments. Ground test results validate that this low-cost approach has the potential to achieve accurate, reliable and continuous localization, regardless of the GPS working statue.

AUTOMATIC ROAD STRUCTURE DETECTION AND VECTORIZATION USING MLS POINT CLOUDS

Abstract. Accurate three-dimensional road structures and models are of great significance to intelligent transportation applications, such as vehicle navigation, inventory evaluation, construction quality control, self-driving vehicles and so on. This paper proposes an efficient and robust method to automatically extract structured road curbs from mobile laser scanning (MLS) data. The proposed method mainly consists of three steps: efficient supervoxel generation, road curbs detection and driving free space estimation. First, supervoxels are generated by assigning ground points with similar geometrical characteristics into the same group. Second, supervoxels with higher local projection density and height difference are identified and clustered as initial road curbs, which are continuous vertical curb facets. The continuous facades consisting of lots of scanned points on the road shoulder can be modeled as multi-dimensional boundary models depending on the requirements of the application, such as vector lines with or without height, micro-facades, etc. Finally, driving free space is obtained due to the road limits can be defined by road boundary in most scenarios. The proposed method is tested on two complex datasets acquired by an Alpha3D mobile laser scanning system from the urban area of Shanghai, China. Experimental results show that the road boundaries and driving free space can be accurately and efficiently extracted, which also demonstrates the superiority of the proposed method.

Investigating long‐term vehicle speed prediction based on BP‐LSTM algorithms

Vehicle speed prediction is quite essential for many intelligent vehicular and transportation applications. Accurate on-road vehicle speed prediction is challenging because individual vehicle speed is affected by many factors related to driver-vehicle-road-traffic system, e.g. the traffic conditions, vehicle type, and driver's behavior, in either a deterministic or stochastic way. Also machine learning makes vehicle speed predictions more accessible by exploring the potential relationship between the vehicle speed and its main factors based on the historical driving data in the context of vehicular networks. This study proposes a novel data-driven vehicle speed prediction method based on back propagation-long short-term memory (BP-LSTM) algorithms for long-term individual vehicle speed prediction along the planned route. Also Pearson correlation coefficient is adopted to analyse the correlation of driver-vehicle-road-traffic historical characteristic parameters for the enhancement of the model's computing efficiency. Finally, a real natural driving data in Nanjing is used to evaluate the prediction performance with a result that the proposed vehicle speed prediction method outperforms other ones in terms of prediction accuracy. Moreover, based on the predicted vehicle speed, this work studies and analyses its effectiveness in two scenarios of energy consumption prediction and travel time prediction.

On Link Stability Metric and Fuzzy Quantification for Service Selection in Mobile Vehicular Cloud

Vehicular cloud (VC) is a promising environment, where intelligent transport applications can be developed relying on mobile vehicles, which can be both cloud users and cloud service providers. It enables vehicles that have sufficient resources to act as mobile cloud servers by offering a variety of services to users’ vehicles. In this context, to consume a cloud service on the move, a user vehicle must first identify the most stable vehicles, relative to his/her motion, which are able to provide the service, and then select the most suitable service according to his/her preferences, while both provider vehicles and their services are described by attributes or quality constraints. Therefore, we introduce a generic relative motion model, as a generic link stability metric, upon which vehicles can form a stable cloud, and we address the VC service selection by using linguistic quantifiers and fuzzy quantified propositions, to define our flexible quantified service selection (FQSS) scheme, which aggregates efficiently both user preferences and service constraints and ranks service providers from the most to the least satisfactory. To break ties among the top-ranked service providers, we make use of our parameters for ranking refinement, called least satisfactory proportion (lsp) and greatest satisfactory proportion (gsp). The simulation results show that our link stability achieves generic motion, by modeling a wider range of vehicle motion types, and our FQSS scheme allows a good successful service consumption rate while reducing latency.

Visible Light Communication System Based on Software Defined Radio: Performance Study of Intelligent Transportation and Indoor Applications

In this paper, our first attempt at visible light communication system, based on software defined radio (SDR) and implemented in LabVIEW is introduced. This paper mainly focuses on two most commonly used types of LED lights, ceiling lights and LED car lamps/tail-lights. The primary focus of this study is to determine the basic parameters of real implementation of visible light communication (VLC) system, such as transmit speed, communication errors (bit-error ratio, error vector magnitude, energy per bit to noise power spectral density ratio) and highest reachable distance. This work focuses on testing various multistate quadrature amplitude modulation (M-QAM). We have used Skoda Octavia III tail-light and Phillips indoor ceiling light as transmitters and SI PIN Thorlabs photodetector as receiver. Testing method for each light was different. When testing ceiling light, we have focused on reachable distance for each M-QAM variant. On the other side, Octavia tail-light was tested in variable nature conditions (such as thermal turbulence, rain, fog) simulated in special testing box. This work will present our solution, measured parameters and possible weak spots, which will be adjusted in the future.

Tracking control of a convoy of autonomous robotic cars with a prescribed performance

This paper studies the tracking control of a convoy of multiple autonomous cars or car-like robots in the planar motion without any collision with a prescribed performance that has not been addressed sufficiently in the literature yet. Toward this end, a coordinate transformation is initially employed to transform the posture errors between each two successive cars in a group from the earth-fixed frame into their relative distance and angle errors that should satisfy some predefined transient and steady-state constraints. By employing the prescribed performance technique, a nonlinear transformation is used to transform the constrained relative distance and angle errors and to obtain an unconstrained kinematic error equation. Then, a novel kinematic controller is designed to satisfy the prescribed transient and steady-state performance specifications without any collision and any controller singularity. Next, the Dynamic Surface Control technique is effectively used to simplify the convoy controller design at the dynamic level by using a first-order filter. An adaptive neural network is used to maintain the control system robustness against modelling errors and external disturbances. The Lyapunov theory proves the stability of the convoy control system and, finally, simulation examples verify the controller performance for multiple autonomous cars in intelligent transportation applications.

A Comprehensive Study of the Effect of Spatial Resolution and Color of Digital Images on Vehicle Classification

Vehicle-type classification is considered a core module for many intelligent transportation applications, such as speed monitoring, smart parking systems, and traffic analysis. In this paper, many vision-based classification techniques were presented relying only on a digital camera without the need for any extra hardware components. Dimension and color are two important characteristics of any digital image that affect the cost of the digital camera used in the image acquisition. In this paper, we present a comprehensive study of the effect of these two characteristics on the vehicle classification process in terms of accuracy and performance. We apply a set of different state-of-the-art image classifiers to the BIT-Vehicle and LabelMe data sets. Each data set is downscaled into different scales to generate a variety of spatial resolutions of each data set. Besides, we examine the effect of color by converting each color version to a gray-scale one. At last, we draw a valid conclusion in regards to the impact of these two characteristics (i.e., dimension and color) on the classification accuracy and performance of the image classification methods using more than 46 000 individual experiments. Experimental results show that there is no significant influence of both color and spatial resolutions of the vehicle images on the classification results obtained by most state-of-the-art image classification methods. However, there is a correlation between the spatial resolution and the processing time required by most image classification methods. Our findings can play an important role in saving not only money, but also time for vehicle-type classification systems.

Vehicular Communications: A Network Layer Perspective

Vehicular communications, referring to information exchange among vehicles, pedestrians, and infrastructures, have become very popular and been widely studied recently due to its great potential to support intelligent transportation and various safety applications. Via vehicular communications, manually driving vehicles and autonomous vehicles can collect useful information to improve traffic safety and support infotainment services. In this paper, we provide a comprehensive overview of recent research on enabling efficient vehicular communications from the network layer perspective. First, we introduce general applications and unique characteristics of vehicular networks and the corresponding classifications. Based on different driving patterns of vehicles, we divide vehicular networks into two categories, i.e., manually driving vehicular networks and automated driving vehicular networks, and then discuss the available communication techniques, network structures, routing protocols, and handoff strategies applied in these vehicular networks. Finally, we identify the challenges confronted by the current vehicular communications and present the corresponding research opportunities.

A Novel Method for Road Intersection Construction From Vehicle Trajectory Data

Road intersection is an important part of the road network, and it plays a key role in many intelligent transportation applications. Currently, the generation of road maps required expensive field surveying and labor-intensive mapping. With more and more public vehicles equipped with positioning devices, massive trajectory data are available. These trajectory data provides a new way to generate and update urban road networks in real time. Although numerous methods are currently available to reconstruct road maps from GPS trajectories, most of these methods focus on the generation of road segments (i.e., road centerlines or road lanes). Road intersections are simply represented by nodes, the generation of road intersections with detailed structures still remains a challenge. Therefore, this paper proposes a novel method for constructing road intersections from vehicle GPS trajectories. First, the boundaries of road intersections are detected by turning angles clustering. Then, the entrances and exits of each intersection are identified based on a newly developed method. Finally, the detailed geometric structures of intersections are reconstructed based on the detected entrances and exits, and turning rules are further extracted and assigned to each intersection. The experiments on real-world trajectory data show that the proposed method can generate the detailed geometric structures of road intersections as well as the turning rule semantic information successfully.

Clustering of Remote Sensing Data Based on K-Nearest Neighbors Sampling With Non-Evenly Division

Data computation and traffic is the key step to rapid analysis and intelligent transportation application based on remote sensing data. To tackle the low computing efficiency and high storage cost in the analysis of remote sensing and improve the computational performance quickly, this paper proposed a new processing method of remote sensing data based on k-nearest neighbors (KNN) sampling with non-evenly division. In the method, we first sort and preprocess the original dataset in terms of any size of one-dimension and segment the sample dataset by non-evenly division. Then the samples with the range of boundary width are reserved, and a new local unsampled mapping table is reconstructed. Next, we traverse the subset and compute the distance matrix by Euclidean distance and the local density with descending order, and further determine whether the sample belongs to boundary sample in accordance with distance matrix and local density. We then construct the sampling dataset and combine again and achieve the processing result via adding the entire unsampled mapping table to the sample dataset. Finally, the current study is tested and verified by the simulation data and true traffic jam prediction case. Our experiments present that the proposed method not only can record precisely the correspondence relations between samples and unsampled data by the KNN sampling with non-evenly division and ensure the accuracy of clustering results, but also significantly reduce the data traffic and effectively improve the memory utilization. The result reveals that the proposed method can potentially contribute to the data analysis of remote sensing data and prediction of traffic jam with large scale and high real-time performance.

Enhancing data delivery in vehicular networks using dual-radio architecture

Vehicular ad-hoc networks (VANETs) have drawn increasing attention from both academia and industry. With the proliferation of sensor and wireless technologies, a large amount of sensory data need to be transmitted among vehicles for various intelligent transportation applications such as autonomous vehicles. The conventional WiFi-based VANETs cannot perform well due to its short communication range. To this end, we introduce a hybrid dual-radio architecture consisting of a 2.4 GHz WiFi and a 433 MHz Eagle Eye (EE) radio characterized by long-range, low-rate, and low-cost communication. Through this EE radio, a mobile node can “see” more nodes in a farther distance to improve the quality of next relay selection for packet delivery using WiFi. Collaborating between WiFi and EE radios, we propose a novel dual-radio routing protocol that significantly alleviates the delivery delay in a distributed manner. To evaluate the performance, we build a dual-radio prototype to verify its feasibility and efficiency. Furthermore, extensive simulations conducted using 4000+ taxis’ trajectories in Shanghai demonstrate that the proposed dual-radio architecture and protocol can reduce up to 50% delivery delay in VANETs.

Data Feature Analysis of Non-Scanning Multi Target Millimeter-Wave Radar in Traffic Flow Detection Applications.

The millimeter-wave radar has been widely used in traffic applications. However, little research has been done to install the millimeter-wave radar on the top of a road for detecting road traffic flow at a downward looking direction. In this paper, the vehicle parameters, including the distance, angle and radar cross-section energy, are collected by practical experiments in the aforementioned application scenario. The data features are analyzed from the dimensions of single parameter sampling characteristics and multi-parameter relationships. Further, the correlations of different parameter series are given using the grey correlation analysis method. For millimeter-wave radar used in the traffic flow detection, our work can definitely provide significant support for further intelligent transportation applications, such as vehicle trajectory tracking, traffic flow estimation and traffic event identification.

Co-Domain Embedding Using Deep Quadruplet Networks for Unseen Traffic Sign Recognition

Recent advances in visual recognition show overarching success by virtue of large amounts of supervised data. However, the acquisition of a large supervised dataset is often challenging. This is also true for intelligent transportation applications, i.e., traffic sign recognition. For example, a model trained with data of one country may not be easily generalized to another country without much data. We propose a novel feature embedding scheme for unseen class classification when the representative class template is given. Traffic signs, unlike other objects, have official images. We perform co-domain embedding using a quadruple relationship from real and synthetic domains. Our quadruplet network fully utilizes the explicit pairwise similarity relationships among samples from different domains. We validate our method on three datasets with two experiments involving one-shot classification and feature generalization. The results show that the proposed method outperforms competing approaches on both seen and unseen classes.

A hybrid error modeling for MEMS IMU in integrated GPS/INS navigation system

Continuity of accurate navigational data for intelligent transportation applications has been widely provided by utilizing low-cost navigation systems through integrating GPS with micro-electro-mechanical-system (MEMS) inertial sensors. To achieve the required accuracy, augmentation of Kalman filter (KF) with nonlinear error modeling techniques such as fast orthogonal search (FOS) was introduced to enhance the navigational solution by estimating and eliminating a great part of both linear and nonlinear errors of azimuth angle sensed by MEMS gyro. Although this augmented approach enhanced the overall navigational accuracy to some extent, it still suffers from some drawbacks that diverge the system accuracy during GPS long outage periods. These drawbacks stem from the wide-variational behavior and high nonlinearities of the errors in MEMS gyros which make it difficult to depend on the non-adaptive linear error model provided by KF to model the two types of MEMS azimuth errors.In this paper we tried to minimize the effect of uncertainties associated with the KF azimuth prediction during the absence of GPS by introducing a hybrid error model which employs support vector machine (SVM) to model the KF output and FOS, based on autoregressive (AR) concept, to model the nonlinear azimuth errors. The performance of the proposed hybrid SVM-FOS approach is evaluated for GPS/ RISS (Reduced inertial sensor system integrated system) and the results were compared with the conventional KF and augmented KF-FOS approaches.

Trust-based security adaptation mechanism for Vehicular Sensor Networks

Vehicular Sensor Networks (VSNs) are foreseen as a promising technology that can provide safe and reliable road travel in smart cities. By establishing pervasive connectivity among vehicles and infrastructure units on the road, various intelligent transport applications could be realized. Efficient network security for VSNs thus becomes a key challenge to reliably implement these applications. However, robust security techniques incur high security overhead and processing delays which significantly impact Quality of Service (QoS) particularly in a dense traffic scenario. In this paper, we propose a trust-based security adaptation mechanism to improve the QoS of safety applications in VSNs. The trust level is calculated using connectivity duration, security level and centrality metrics of nearby vehicles. The simulation results we have obtained with our proposed research shows an improvement of 25–65% in terms of safety awareness and 33–53% improvement in terms of packet inter-arrival time of safety applications in VSNs.

Reconstructing Activity Location Sequences From Incomplete Check-In Data: A Semi-Markov Continuous-Time Bayesian Network Model

Geo-location data from the check-ins made in online social media offers us information, in new ways, to understand activity-location choices of a large number of people. However, one of the major challenges of using check-in data is that it has missing activities, since users share their activities voluntarily. In this paper, we present a probabilistic modeling approach to reconstruct user activity-location sequences from this incomplete activity participation information. Specifically, we answer the question of how to predict an individual’s next activity, its duration and location given the incomplete trajectory data. The model describes the dynamics of individual activity participation behavior evolving over continuous time. A semi-Markov modeling approach is used to capture the stochastic processes involved in the activity generation mechanism. We present a particle-based Markov chain Monte Carlo sampler to run inference over the model. We further develop an expectation-maximization algorithm to learn the unknown parameters of the model from incomplete trajectory data. Finally, the method is applied to synthetically generated activity-location sequences and a data set of Foursquare check-ins of the users from New York City. Our experiments show that this method can successfully extract the true transition and duration distributions given the incomplete trajectory information. The proposed approach can help building many intelligent transportation applications using check-in data.