arXiv subjects

Kangcheng Liu

Publications and source records attributed to Kangcheng Liu.

24 records · Page 2Linked to original sources

arXiv2022

An Integrated Visual System for Unmanned Aerial Vehicles Tracking and Landing on the Ground Vehicles

The vision of unmanned aerial vehicles is very significant for UAV-related applications such as search and rescue, landing on a moving platform, etc. In this work, we have developed an integrated system for the UAV landing on the moving platform, and the UAV object detection with tracking in the complicated environment. Firstly, we have proposed a robust LoG-based deep neural network for object detection and tracking, which has great advantages in robustness to object scale and illuminations compared with typical deep network-based approaches. Then, we have also improved based on the original Kalman filter and designed an iterative multi-model-based filter to tackle the problem of unknown dynamics in real circumstances of motion estimations. Next, we implemented the whole system and do ROS Gazebo-based testing in two complicated circumstances to verify the effectiveness of our design. Finally, we have deployed the proposed detection, tracking, and motion estimation strategies into real applications to do UAV tracking of a pillar and obstacle avoidance. It is demonstrated that our system shows great accuracy and robustness in real applications.

cs.RO↗

arXiv2022

An Enhanced LiDAR-Inertial SLAM System for Robotics Localization and Mapping

The LiDAR and inertial sensors based localization and mapping are of great significance for Unmanned Ground Vehicle related applications. In this work, we have developed an improved LiDAR-inertial localization and mapping system for unmanned ground vehicles, which is appropriate for versatile search and rescue applications. Compared with existing LiDAR-based localization and mapping systems such as LOAM, we have two major contributions: the first is the improvement of the robustness of particle swarm filter-based LiDAR SLAM, while the second is the loop closure methods developed for global optimization to improve the localization accuracy of the whole system. We demonstrate by experiments that the accuracy and robustness of the LiDAR SLAM system are both improved. Finally, we have done systematic experimental tests at the Hong Kong science park as well as other indoor or outdoor real complicated testing circumstances, which demonstrates the effectiveness and efficiency of our approach. It is demonstrated that our system has high accuracy, robustness, as well as efficiency. Our system is of great importance to the localization and mapping of the unmanned ground vehicle in an unknown environment.

cs.RO↗

arXiv2022

A Light-Weight LiDAR-Inertial SLAM System with Loop Closing

In this work, we propose a lightweight integrated LiDAR-Inertial SLAM system with high efficiency and a great loop closure capacity. We found that the current State-of-the-art LiDAR-Inertial SLAM system has poor performance in loop closure. The LiDAR-Inertial SLAM system often fails with the large drifting and suffers from limited efficiency when faced with large-scale circumstances. In this work, firstly, to improve the speed of the whole LiDAR-Inertial SLAM system, we have proposed a new data structure of the sparse voxel-hashing to enhance the efficiency of the LiDAR-Inertial SLAM system. Secondly, to improve the point cloud-based localization performance, we have integrated the loop closure algorithms to improve the localization performance. Extensive experiments on the real-scene large-scale complicated circumstances demonstrate the great effectiveness and robustness of the proposed LiDAR-Inertial SLAM system.

cs.RO↗

arXiv2022

An Integrated LiDAR-SLAM System for Complex Environment with Noisy Point Clouds

The current LiDAR SLAM (Simultaneous Localization and Mapping) system suffers greatly from low accuracy and limited robustness when faced with complicated circumstances. From our experiments, we find that current LiDAR SLAM systems have limited performance when the noise level in the obtained point clouds is large. Therefore, in this work, we propose a general framework to tackle the problem of denoising and loop closure for LiDAR SLAM in complex environments with many noises and outliers caused by reflective materials. Current approaches for point clouds denoising are mainly designed for small-scale point clouds and can not be extended to large-scale point clouds scenes. In this work, we firstly proposed a lightweight network for large-scale point clouds denoising. Subsequently, we have also designed an efficient loop closure network for place recognition in global optimization to improve the localization accuracy of the whole system. Finally, we have demonstrated by extensive experiments and benchmark studies that our method can have a significant boost on the localization accuracy of the LiDAR SLAM system when faced with noisy point clouds, with a marginal increase in computational cost.

cs.RO↗

arXiv2022

Semi-Supervised Confidence-Level-based Contrastive Discrimination for Class-Imbalanced Semantic Segmentation

To overcome the data-hungry challenge, we have proposed a semi-supervised contrastive learning framework for the task of class-imbalanced semantic segmentation. First and foremost, to make the model operate in a semi-supervised manner, we proposed the confidence-level-based contrastive learning to achieve instance discrimination in an explicit manner, and make the low-confidence low-quality features align with the high-confidence counterparts. Moreover, to tackle the problem of class imbalance in crack segmentation and road components extraction, we proposed the data imbalance loss to replace the traditional cross entropy loss in pixel-level semantic segmentation. Finally, we have also proposed an effective multi-stage fusion network architecture to improve semantic segmentation performance. Extensive experiments on the real industrial crack segmentation and the road segmentation demonstrate the superior effectiveness of the proposed framework. Our proposed method can provide satisfactory segmentation results with even merely 3.5% labeled data.

cs.CV↗

arXiv2021

FG-Net: Fast Large-Scale LiDAR Point Clouds Understanding Network Leveraging Correlated Feature Mining and Geometric-Aware Modelling

This work presents FG-Net, a general deep learning framework for large-scale point clouds understanding without voxelizations, which achieves accurate and real-time performance with a single NVIDIA GTX 1080 GPU. First, a novel noise and outlier filtering method is designed to facilitate subsequent high-level tasks. For effective understanding purpose, we propose a deep convolutional neural network leveraging correlated feature mining and deformable convolution based geometric-aware modelling, in which the local feature relationships and geometric patterns can be fully exploited. For the efficiency issue, we put forward an inverse density sampling operation and a feature pyramid based residual learning strategy to save the computational cost and memory consumption respectively. Extensive experiments on real-world challenging datasets demonstrated that our approaches outperform state-of-the-art approaches in terms of accuracy and efficiency. Moreover, weakly supervised transfer learning is also conducted to demonstrate the generalization capacity of our method.

cs.CV↗