Internet of Vehicles (IoV) is a distributed network of connected cars, roadside infrastructure, wireless communication networks, and central cloud platforms. Wireless recommendations play an important role in the IoV network, for example, recommending appropriate routes, recommending driving strategies, and recommending content. In this paper, we review some of the key techniques in recommendations and discuss what are the opportunities and challenges to deploy these wireless recommendations in the IoV.

/pdf/wireless-recommendations-for-internet-of-vehicles-recent-2cvoiwxk9g.pdf

Wireless recommendations for internet of vehicles: Recent advances, challenges, and opportunities

High Definition map (HD Map) is an important part of autonomous driving vehicle. Most conventional method to generate HD map requires expensive system and postprocessing of observed data. In this paper, we propose automatic HD map generating algorithm using just monocular camera without further human labors. The proposed algorithm detects road lane from image and classifies the type of road lane at pixel-level with Fully Convolutional Network (FCN) which outperforms the other semantic segmentation methods. The segmentation results are used to extract lane features, and the features are used for loop-closure detection. Final map is generated with graph-based Simultaneous Localization and Mapping (SLAM) algorithm. The experiment is done with monocular camera mounted on mobile vehicle. In this paper, final map generated by proposed method is compared with aerial view data. The results show that the proposed method can generate reliable map that is comparable to real roads even only the low-cost sensor is used.

Road Lane Semantic Segmentation for High Definition Map

As the rise of the intelligent vehicle applications in recent years, the development of onboard vision systems for advanced driving assistance has become a popular research topic. This paper presents a real-time system using a monocular camera mounted on the rear of a vehicle to perform overtaking detection for safe lane change operations. In this work, the possible overtaking vehicle is first located based on motion cues. The candidate is then identified using Convolutional Neural Network (CNN) and tracked for behavior analysis in a short period of time. We also propose an algorithm to solve the issue of repetitive patterns which is commonly appeared in the highway driving. A series of experiments are carried out with real scene video sequences recorded by a dashcam. The performance evaluation has demonstrated the effectiveness of the

Overtaking Vehicle Detection Techniques based on Optical Flow and Convolutional Neural Network.

High-precision indoor localization plays a vital role in various places. In recent years, visual inertial odometry (VIO) system has achieved outstanding progress in the field of indoor localization. However, it is easily affected by poor lighting and featureless environments. For this problem, we propose an indoor localization algorithm based on VIO system and three-dimensional (3D) map matching. The 3D map matching is to add height matching on the basis of previous two-dimensional (2D) matching so that the algorithm has more universal applicability. Firstly, the conditional random field model is established. Secondly, an indoor three-dimensional digital map is used as a priori information. Thirdly, the pose and position information output by the VIO system are used as the observation information of the conditional random field (CRF). Finally, the optimal states sequence is obtained and employed as the feedback information to correct the trajectory of VIO system. Experimental results show that our algorithm can effectively improve the positioning accuracy of VIO system in the indoor area of poor lighting and featureless.

Indoor Localization Based on VIO System and Three-Dimensional Map Matching

For robots in unknown environments, the visual-inertial simultaneous localization and mapping (SLAM) incorporates the multisensor data and enables a globally consistent trajectory tracking and mapping. Given the scene structures in the SLAM problem, the high-quality lines can serve as primitives and provide additional constraints for camera pose estimation. This study is devoted to a pipeline design of dynamically fused point-line visual-inertial navigation system (VINS). The design builds on the heritage of VINS-Mono but employs an ensemble pose estimation strategy to integrate point and line features. As far as we know, it is the first optimization-based monocular VINS method that gives an insight into the quantitative relationship between point and line feature numbers. Additionally, the measured pixels for line feature-length determination and two hidden parameters in line segment detector (LSD) extractor are investigated, guaranteeing the line feature extraction optimization in both quality and efficiency. The benchmark dataset and real-world tests are conducted. In terms of challenging scene tests in the EuRoc dataset, results are compared against ROVIO, VINS-Mono, and PL-VINS frameworks, revealing our design is substantially more accurate and robust than others with both locally and globally consistent pose estimates. The engineered algorithm in the camera “Mynteye” further enables the Bulldog-CX Robot platform verification. The results concerning the drifts, time taken for line feature extraction, and real-world trajectory consistency show our design outperforms the point-based VINS. The strong adaptability of our dynamically fused point-line VINS to the broad range of indoor and outdoor scenes is simultaneously verified.

Visual-Inertial Simultaneous Localization and Mapping: Dynamically Fused Point-Line Feature Extraction and Engineered Robotic Applications

This paper presents an early research on the route recommendation system based on street view images. It consists of three primary functions. (1) Data collection: The user selects a path using a graphical user interface and download all street view images for the path. (2) Road surface extraction: The vanishing point in the image is obtained according to the road surface region segmentation by local growing at a super-pixel level. (3) Road surface identification: The images of brick and asphalt road surfaces are collected for training, and then used to identify the road surface type for route recommendation. The objective is to find the most comfortable path, which is not necessarily to be the shortest one. The experimental results show that our pre-selected groups of route well suit the original expectation.

Road surface detection and recognition for route recommendation

SLAM (simultaneous localization and mapping) is commonly considered as a crucial component to achieving autonomous robot navigation. Currently, most of the existing visual 3D SLAM systems are still not robust enough. Image blur, variation of illumination, and low-texture scenes may lead to registration failures. To let the visual odometry (VO) deal with these problems, the workflow of traditional approaches becomes bulky and complicated. On the other hand, the advancement of deep learning brings new opportunities. In this paper, we use a deep network model to predict complex camera motion. It is different from previous supervised learning VO researches and requires no camera trajectories which are diﬃcult to obtain. Using the image input and IMU output as an end-to-end training pair makes data collection cost-effective. The optical flow structure also makes the system independent of the appearance of training sets. The experimental results show that the proposed architecture has faster training convergence, and the model parameters are also significantly reduced. Our method is able to remain certain robustness under image blur, illumination changes, and low-texture scenes. It can correctly predict the new EuRoC dataset, which is more challenging than the KITTI dataset.

Tse-An Jacky Liu

Papers

Road surface detection and recognition for route recommendation

InertialNet: Toward Robust SLAM via Visual Inertial Measurement