In large-scale long-term dynamic environments, high-frequency dynamic objects inevitably lead to significant changes in the appearance of the scene at the same location at different times, which is catastrophic for place recognition (PR). Therefore, how to eliminate the influence of dynamic objects to achieve robust PR has universal practical value for mobile robots and autonomous vehicles. To this end, we suggest a novel semantically consistent LiDAR PR method based on chained cascade network, called SC_LPR, which mainly consists of a LiDAR semantic image inpainting network (LSI-Net) and a semantic pyramid Transformer-based PR network (SPT-Net). Specifically, LSI-Net is a coarse-to-fine generative adversarial network (GAN) with a gated convolutional autoencoder as the backbone. To effectively address the challenges posed by variable-scale dynamic object masks, we integrate the updated Transformer block with mask attention and gated trident block into LSI-Net. Sequentially, in order to generate a discriminative global descriptor representing the point cloud, we design an encoder with pyramid Transformer block to efficiently encode long-range dependencies and global contexts between different categories in the inpainted semantic image, followed by an augmented NetVALD, a generalized VLAD (Vector of Locally Aggregated Descriptors) layer that adaptively aggregates salient local features. Last but not least, we first attempt to create a LiDAR semantic inpainting dataset, called LSI-Dataset, to effectively validate the proposed method. Experimental comparisons show that our method not only improves semantic inpainting performance by about 6%, but also improves PR performance in dynamic environments by about 8% compared to the representative optimal baseline. LSI-Dataset will be publicly available at https://github.KD.LPR.com/.

Light detection and ranging (LiDAR) positioning has received great attention especially when satellites fail. However, the positioning accuracy is still subjected to the following. First, the LiDAR positioning accuracy is affected by the sparsity due to LiDAR beams and environment. Second, the existing methods for suppressing cumulative errors are based on ego-vehicle sensing that requires the motion of repeated paths. Moreover, the output frequency of LiDAR is low. To solve the above problems, an enhanced-LiDAR/ultrawideband (UWB)/inertial navigation system integrated positioning methodology is proposed. First, the enhanced LiDAR odometry module is designed to improve the resolution of LiDAR beams for more accurate odometry. Then, the cooperative optimization module is proposed to introduce UWB observation to suppress the accumulated error without relying on ego-vehicle sensing. Finally, the factor graph fusion module is used to fuse multisensor information dynamically and improve the output frequency. Experimental results prove the effectiveness of our methodology.

3D semantic segmentation is a key technology of scene understanding in the self-driving field, which remains challenging problems. Recent 3D segmentation methods have achieved competitive results in indoor or typical urban traffic scenes. However, in complex and changeable scenarios where structured features are sparse and irregular, few of these methods could achieve well segmentation results, especially causing blurry and inaccurate boundary distinctions between inter-class objects, drivable areas, and backgrounds. In order to fully harvest boundary information and accurately distinguish the category of points on road and object boundaries in real-time, we present an efficient multi-level boundary-semantic-enhanced model for LiDAR semantic segmentation, which comprehensively discover boundary features in three aspects: first, boundary channels are extracted directly from LiDAR range images as the inputs of boundary-branch; second, the boundary attention module is designed to deeply fuse boundary information into the main segmentation branch; third, a modified discriminator is utilized to raise the perception of boundary information by minimizing the gap between the predicted and true boundaries. Besides, we add a semantic-enhanced module using the similar discriminator to optimize semantic segmentation results in the output end. Quantitative and qualitative evaluations are performed on both structured and unstructured real-world datasets including urban dataset SemanticKITTI, off-road dataset Rellis3D and our unstructured test set. The experimental results validate the effectiveness of the proposed methodology in improving efficiency, accuracy and scene-adaptivity.

3D semantic segmentation is a key technology for intelligent vehicles. Recently, great efforts have been made to achieve accurate and robust 3D semantic segmentation results through LiDAR-camera fusion, due to the complementary information of images like color and texture. However, current LiDAR-camera fusion methods still lag behind LiDAR-only methods in complex and varied scenarios containing sparse and similar geometric features. Three main factors limit their performance: sparse mappings between two modalities, the contribution of different modality features, and the gap between results of two modalities. Addressing that, we present a multi-phase LiDAR-camera fusion method, which comprehensively fuses information of two modalities in three phases: first, a context sampler block is proposed to fuse low-level image contextual features into LiDAR points; second, a cross dimensional fusion block is designed to implement the interaction of shared features for hierarchical supervision while retaining private features for jointly refining their predictions; third, a modality perceptual conformance loss is induced to make them learn from each other in an explicit manner by minimizing the gap between two modality predictions. Quantitative and qualitative evaluations with state-of-the-art (SOTA) methods are performed on urban dataset SemanticKITTI, off-road dataset RELLIS-3D and our unstructured test set. The experimental results validate the competitiveness of our method in improving accuracy and scene-adaptivity.

Semantic segmentation of complex marine environments based on images is crucial for the autonomous navigation of Unmanned Surface Vessels (USVs). However, existing semantic segmentation methods mainly categorize images into three coarse categories: sea surface, obstacles, and sky, with limited attention to the boundaries between significant elements. Moreover, the dynamic changes in marine environments often affect these methods, resulting in shortcomings such as blurry boundary segmentation. To address these issues, this paper proposes a boundary enhancement-driven semantic segmentation network tailored for complex marine environments, focusing primarily on the accuracy of inter-class boundaries. Specifically, 1) a boundary extraction module is proposed to extract multi-scale boundary features from the backbone network, which are fused via continuous boundary attention streams and supervised with boundary loss. 2) A boundary enhancing module is introduced, wherein the connection between channel features is strengthened using the Astous convolutional pyramid module and channel attention module, enhancing the perception of contextual boundary information and improving the accuracy of category boundaries in the segmentation results. Comprehensive comparative experiments were conducted on the MaSTr1325 dataset and MID dataset, and the proposed method was evaluated on the MODD2 dataset using the MODS evaluation method. Results demonstrate that our approach, BEMSNet, achieves clearer and more accurate boundary segmentation compared to other networks.

3D semantic segmentation is a key task of environment understanding in various outdoor scenes. Due to the sparsity and varying density of point clouds, it becomes challenging to obtain fine-gained segmentation results. Previous point-based and voxel-based methods suffer from the expensive computational cost. Recent 2D projection-based methods, including range-view, bird-eye-view and multi-view fusion methods, can run in real time, but the information loss during the projection leads to the low accuracy. Also, we find that the occlusion and interlacing problems exist in single projection-based methods and most multi-view fusion network only focus on the output level fusion. Considering the above issues, we propose a multi-level multi-view fusion network using attention modules and hierarchical transformer, which ensures the effectiveness and efficiency mainly by the following three aspects: 1) the spatial-channel attention module integrates contextual information between points and learn differences of each channel’s features. 2) the proposed geometry-based multi-projection fusion module achieves the geometric feature alignment between range view and bird eye view and fuses the features of the two views at both feature level and output level. 3) we introduce KPConv to replace KNN, which can reduce the information loss during the post-processing. Experiments are conducted on both structured and unstructured datasets including urban dataset SemanticKITTI and off-road dataset Rellis3D. Our results achieve a better performance compared to other projection-based methods and are comparable with the state-of-the-art Cylinder3D.