3D object detection is an important part of autonomous driving systems, which tasks are to accurately recognize 3D objects around the vehicle, such as cars, pedestrians, and bicycles. Current research is primarily focused on successfully integrating data from camera and LiDAR sensors to enhance detection accuracy and reliability, while overcoming the limits of using a single sensor. In this paper, we provide a review of 3D object detection methods for multi-sensor fusion. First, we introduce common camera and LiDAR sensors and their data processing methods. Subsequently, we classify the fusion algorithms into three categories: input fusion, feature fusion, and late fusion, based on different fusion strategies, and conduct an in-depth survey and discussion on them to analyze their respective advantages and disadvantages. In addition, we provide an overview of public datasets commonly used in 3D object detection. Finally, we provide an outlook on the future direction of multi-sensor fusion 3D object detection technology.
Visual simultaneous localization and mapping (SLAM) in dynamic scenarios is vulnerable to dynamic object factors, which can cause inaccurate pose estimation and limited robustness. To tackle this problem, we present a SLAM algorithm named SDG-SLAM, which combines improved visual semantic and geometric constraints to reject outliers and enhance performance. Firstly, the algorithm utilizes the neural network YOLOv5 (the fifth version of You Only Look Once) to obtain object’s prior semantics in the environment. In the meantime, the ORB feature points of keyframes are acquired, and the sparse optical flow tracking and matching are performed on non-key frames Secondly, adaptive initial dynamic point detection is implemented based on the epipolar constraints. Subsequently, the method of motion constraint posterior combined with semantic prior was used to screen and eliminate dynamic feature points. Finally, all static feature points from the environment are utilized, and these feature points are participated in the camera’s pose estimation. Substantial experiments performed on challenging public dynamic scenario datasets and outdoor dynamic scenarios demonstrate that SDG-SLAM significantly improves pose accuracy in scenes containing objects with different motion states. Specifically, compared to ORB-SLAM3, SDG-SLAM achieves more than 87% improvement in localization accuracy for high dynamic datasets and over 85% improvement for outdoor dynamic environments.
SLAM (Simultaneous Localization and Mapping) is a key technology for mobile robots to accomplish autonomous perception tasks. However, the performance of a single sensor limits the application of different scenarios, therefore, this paper proposes a SLAM algorithm for the fusion of vision and inertia: VI Fusion (Visual Inertia Fusion), which can effectively improve the stability and accuracy of the system in complex environments. First, the improved visual features are combined with the joint initialization of IMU; second, a minimization function for the joint optimization of vision and inertia is designed; and finally, the EuRoC dataset and the real outdoor scene are selected for the comparison experiments. The results show that the algorithmic trajectory error of this paper is reduced by 55.5% on average, and the algorithmic trajectory of this paper has better consistency and continuity with the reference trajectory generated by RTK compared with similar SLAM algorithms in the outdoor scene test.
Simultaneous localization and mapping (SLAM) is a significant challenge for autonomous mobile robotics, as it necessitates the reconstruction of an unfamiliar environment while simultaneously determining the robot's position using a map. The mobile robot employed a diverse range of sensors in the SLAM process to gather and interpret a map representation. Geometric model-based strategies have historically been employed for addressing SLAM problems. However, these techniques are prone to errors in tough environments, and single-sensor SLAM systems have degradation and adverse effects on localization and mapping outcomes in extreme environments characterized by high dynamics or coefficient characteristics. Recently, there has been a significant increase in the development of multi-sensor fusion SLAM techniques aimed at achieving more stable and resilient systems. This paper focuses on the development process and contemporary research conducted on multi-sensor fusion SLAM. This paper initially presents the overall structure of SLAM. It then provides a detailed explanation of the functions of front-end odometry, back-end optimization, loopback detection, and map building modules. Additionally, it provides a summary of the algorithms employed in SLAM. Furthermore, it describes and summarizes the classical representative open source algorithms based on the types of sensors to be combined. It also introduces commonly utilized public datasets, along with accuracy evaluation indexes and measurement tools. In conclusion, this paper presents a comprehensive examination of the development process, along with a summary of prominent research studies on the fusion of many sensors in SLAM.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.