• Journal of Broadcast Engineering
• /
• v.24 no.4
• /
• pp.580-591
• /
• 2019

This paper focuses on to improving object detection performance using the camera and LiDAR on autonomous vehicle platforms by fusing detected objects from individual sensors through a late fusion approach. In the case of object detection using camera sensor, YOLOv3 model was employed as a one-stage detection process. Furthermore, the distance estimation of the detected objects is based on the formulations of Perspective matrix. On the other hand, the object detection using LiDAR is based on K-means clustering method. The camera and LiDAR calibration was carried out by PnP-Ransac in order to calculate the rotation and translation matrix between two sensors. For Sensor fusion, intersection over union(IoU) on the image plane with respective to the distance and angle on world coordinate were estimated. Additionally, all the three attributes i.e; IoU, distance and angle were fused using logistic regression. The performance evaluation in the sensor fusion scenario has shown an effective 5% improvement in object detection performance compared to the usage of single sensor.

• Journal of Electrical Engineering and Technology
• /
• v.13 no.6
• /
• pp.2530-2544
• /
• 2018

In this paper, we propose an enhanced LiDAR-camera calibration method that extracts the marker plane from 3D point cloud information. In previous work, we estimated the straight line of each board to obtain the vertex. However, the errors in the point information in relation to the z axis were not considered. These errors are caused by the effects of user selection on the board border. Because of the nature of LiDAR, the point information is separated in the horizontal direction, causing the approximated model of the straight line to be erroneous. In the proposed work, we obtain each vertex by estimating a rectangle from a plane rather than obtaining a point from each straight line in order to obtain a vertex more precisely than the previous study. The advantage of using planes is that it is easier to select the area, and the most point information on the board is available. We demonstrated through experiments that the proposed method could be used to obtain more accurate results compared to the performance of the previous method.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.33 no.6
• /
• pp.1099-1110
• /
• 2023

The recognition system of autonomous driving and robot navigation performs vision work such as object recognition, tracking, and lane detection after multi-sensor fusion to improve performance. Currently, research on a deep learning model based on the fusion of a camera and a lidar sensor is being actively conducted. However, deep learning models are vulnerable to adversarial attacks through modulation of input data. Attacks on the existing multi-sensor-based autonomous driving recognition system are focused on inducing obstacle detection by lowering the confidence score of the object recognition model.However, there is a limitation that an attack is possible only in the target model. In the case of attacks on the sensor fusion stage, errors in vision work after fusion can be cascaded, and this risk needs to be considered. In addition, an attack on LIDAR's point cloud data, which is difficult to judge visually, makes it difficult to determine whether it is an attack. In this study, image scaling-based camera-lidar We propose an attack method that reduces the accuracy of LCCNet, a fusion model (camera-LiDAR calibration model). The proposed method is to perform a scaling attack on the point of the input lidar. As a result of conducting an attack performance experiment by size with a scaling algorithm, an average of more than 77% of fusion errors were caused.

• Korean Journal of Remote Sensing
• /
• v.35 no.5_2
• /
• pp.841-850
• /
• 2019

The purpose of this study is to strengthen the capability of rapid mapping for disaster through improving the positioning accuracy of mapping and fusion of multi-sensing point cloud data derived from Unmanned Aerial Vehicles (UAVs) and disaster investigation vehicle. The positioning accuracy was evaluated for two procedures of drone mapping with Agisoft PhotoScan: 1) general geo-referencing by self-calibration, 2) proposed geo-referencing with optimized camera model by using fixed accurate Interior Orientation Parameters (IOPs) derived from indoor camera calibration test and bundle adjustment. The analysis result of positioning accuracy showed that positioning RMS error was improved 2~3 m to 0.11~0.28 m in horizontal and 2.85 m to 0.45 m in vertical accuracy, respectively. In addition, proposed data fusion approach of multi-sensing point cloud with the constraints of the height showed that the point matching error was greatly reduced under about 0.07 m. Accordingly, our proposed data fusion approach will enable us to generate effectively and timelinessly ortho-imagery and high-resolution three dimensional geographic data for national disaster management in the future.

• Journal of the Society of Naval Architects of Korea
• /
• v.61 no.3
• /
• pp.143-151
• /
• 2024

In this study, we present a mapping framework for 3D spatial reconstruction of digital twin model using navigation and perception sensors mounted on an Autonomous Surface Vehicle (ASV). For improving the level of realism of digital twin models, 3D spatial information should be reconstructed as a digitalized spatial model and integrated with the components and system models of the ASV. In particular, for the 3D spatial reconstruction, color and 3D point cloud data which acquired from a camera and a LiDAR sensors corresponding to the navigation information at the specific time are required to map without minimizing the noise. To ensure clear and accurate reconstruction of the acquired data in the proposed mapping framework, a image preprocessing was designed to enhance the brightness of low-light images, and a preprocessing for 3D point cloud data was included to filter out unnecessary data. Subsequently, a point matching process between consecutive 3D point cloud data was conducted using the Generalized Iterative Closest Point (G-ICP) approach, and the color information was mapped with the matched 3D point cloud data. The feasibility of the proposed mapping framework was validated through a field data set acquired from field experiments in a inland water environment, and its results were described.