A monocular vision based rear vehicle detection and tracking system is presented for Lane Change Assist (LCA), which does not need road boundary and lane information. Our algorithm extracts regions of interest (ROI) using the shadow underneath a vehicle, and accurately localizes vehicle regions in ROI by vehicle features such as symmetry, edge and shadow underneath vehicles. The algorithm realizes vehicle verification by combining knowledge-based and learning-based methods. During vehicle tracking, templates are dynamically created on-line, tracking window is adaptively adjusted with motion estimation, and confidence is determined for tracked vehicle. The algorithm was tested under various traffic scenes at different daytime, the result illustrated good performance.

Detection and tracking of lane marking is essential for driving safety and intelligent vehicle. In this paper, an algorithm is presented which allows detection and tracking of multiple lane markings. Edge points cue is used to detect the lane marking and a road orientation estimation method is used to delete the edge lines which are impossible attribute to lane markings. In order to select the candidate lane marking, a confidence measures method is proposed. Then a finite-state machine decides whether or not a lane marking is really detected by fusion multi-frame detection results. Specifically, a particle filter is used to predict the future values of the lane marking model parameters, based on past observations. With particle filtering and confidence measures method, lane markings on various road scenes are detected and tracked. Experimental in different conditions, including illumination, weather and road, demonstrates its effectiveness and robustness. The algorithm runs in real-time at rates of about 30 Hz.

Lane detection has been one of the most important functions in the autonomous driving perception module. Most of the current research require complex post-processing and curve fitting processes before they can be used by subsequent regulation modules. In this paper, we propose the LLFormer algorithm combining CNN and Transformer structure, which is the first attempt to perform end-to-end lane detection based on laser point cloud and output its cubic polynomial coefficients. In addition, this paper modifies the structure of the conventional transformer and proposes the Generating Lane Query (GLQ) module. The output of encoder is plugged into GLQ for initialization of lane query in decoder, preserving the uniqueness of each frame of point cloud data. We test the performance of the proposed algorithm in the public dataset K-Lane, and the results show that the accuracy of the proposed LLFormer is close to the existing SOTA algorithm. The number of model parameters of LLFormer is only 9.01M, and the amount of operations is only 0.19GFLOPs, which are 1/26 and 1/2937 of the existing SOTA algorithm, respectively. The frequency of inference calculation is 35.9FPS, which can fully meet the real-time requirements for industrial deployment.

This paper presents a fast LiDAR-inertial odometry (LIO) that is robust to aggressive motion. To achieve robust tracking in aggressive motion scenes, we exploit the continuous scanning property of LiDAR to adaptively divide the full scan into multiple partial scans (named sub-frames) according to the motion intensity. And to avoid the degradation of sub-frames resulting from insufficient constraints, we propose a robust state estimation method based on a tightly-coupled iterated error state Kalman smoother (ESKS) framework. Furthermore, we propose a robocentric voxel map (RC-Vox) to improve the system’s efficiency. The RC-Vox allows efficient maintenance of map points and k nearest neighbor (k-NN) queries by mapping local map points into a fixed-size, two-layer 3D array structure. Extensive experiments are conducted on 27 sequences from 4 public datasets, our own dataset and real-world scenes. The results show that our system can achieve stable tracking in aggressive motion scenes (angular velocity up to 21.8 rad/s) that cannot be handled by other state-of-the-art methods, while our system can achieve competitive performance with these methods in general scenes. Furthermore, thanks to the RC-Vox, our system is much faster than the most efficient LIO system currently published.

This study presents a hierarchical background modelling and subtraction approach for real‐time detection of moving objects. At the first level, a novel pixel‐wise background modelling method is proposed for coarse detection. The method can dynamically assign the optimal number of components for each pixel with the borrow–lend strategy. And a flexible learning rate which is variable and different for each component is presented to adapt to scene changes. Additionally, a new mechanism using a framework of finite state machine is introduced to maintain and update the background models. At the second level, in order to deal with sudden illumination changes, a block‐wise foreground validation approach is adopted for refined detection. The authors compare the proposed approach with state‐of‐the‐art methods and experimental results under various scenes demonstrate the robustness and effectiveness of the proposed approach.

Detection based on LiDAR and camera fusion is increasingly popular for researchers in the autonomous driving domain. Compared to the camera-only and LiDAR-only methods, the fusion-based methods indeed improve the detection accuracy on public-available datasets. However, due to the complexity of the projection or fusion mechanism, few of these methods can run in real time even on an advanced desktop GPU. Thus, in this paper, we propose a new fusion detection model G-Fusion with a light and fast image view-transform module. According to our receptive field analysis of image feature maps, we directly project image features to only one voxel layer located on the ground, then fuse the LiDAR and image features by concatenation and convolution. With this delicately designed module, G-Fusion greatly boosts the state-of-the-art speed performance on the nuScenes dataset, achieving a good balance with the competitive detection scores. Meanwhile, since the precision of sensor extrinsic parameters is important for most fusion-based methods, we also deeply dig into our model’s calibration error tolerance ability and discover the failure noise condition.