In outdoor environments, ground segmentation is an important pre-processing task for the local environment perception of autonomous vehicle platforms and is the basis for obstacle detection, classification, and path planning. However, existing ground segmentation algorithms primarily focus on dense point cloud data and often struggle to achieve satisfactory accuracy when applied to sparse point cloud data. In light of this challenge, we propose a novel ground segmentation method, LR-Seg, specifically tailored for sparse point cloud data. Our method aims to address the limitations observed in previous approaches and provide improved performance, particularly on sparse point cloud data. To achieve faster processing speed, the method first divides the original point cloud into sub-regions of different sizes based on the distribution characteristics of the sparse point cloud data in the XOY plane. Then, the point clouds are appropriately assigned, and most of the non-ground points in each sub-region are removed using the PCA plane fitting method. Finally, the point cloud geometric feature information within each sub-region is used to reduce over-segmentation. The experimental results show that our method can perform ground segmentation quickly on flat and slope road sections, and the average F1 score can be maintained above 90% with an average time of 3 ms, which is suitable for various road environments.

Road detection is a crucial part of the autonomous driving system, and semantic segmentation is used as the default method for this kind of task. However, the descriptive categories of agroforestry are not directly definable and constrain the semantic segmentation-based method for road detection. This paper proposes a novel road detection approach to overcome the problem mentioned above. Specifically, a novel two-stage method for road detection in an agroforestry environment, namely ARDformer. First, a transformer-based hierarchical feature aggregation network is used for semantic segmentation. After the segmentation network generates the scene mask, the edge extraction algorithm extracts the trail’s edge. It then calculates the periphery of the trail to surround the area where the trail and grass are located. The proposed method is tested on the public agroforestry dataset, and experimental results show that the intersection over union is approximately 0.82, which significantly outperforms the baseline. Moreover, ARDformer is also effective in a real agroforestry environment.

In off-road scenes, the fusion of dual-LiDAR data is crucial for ensuring the accuracy of environmental perception in autonomous vehicles. The terrain in off-road scenes is complex and filled with unstructured information. This leads to significant noise and a lack of distinct structural features in the point cloud data, making traditional point cloud registration methods difficult to apply. To address these issues posed by complex off-road scenes, we propose a novel point cloud fusion framework named Off-Fusion. We first filter and segment the ground in the input point cloud data, focusing on preserving the core features of the ground while effectively removing noise caused by the terrain. Next, we propose a robust and efficient feature point extraction method based on voxel division and curvature weighting, ensuring extracting meaningful and representative feature points from complex off-road scenes. Based on this, we use feature matching to calculate rough relative transformation pairs, providing a high-quality starting position for the Iterative Closest Point (ICP) algorithm, effectively avoiding local optima. Finally, by combining the kd-tree accelerated ICP algorithm, we achieve precise point cloud registration, successfully calculating the optimal rotation and translation matrix between the two LiDARs. The experimental results show that our method significantly improves the quality and speed of data fusion. Compared to some of the most advanced methods, it performs better in off-road scenes, achieving the best results.

The rapid development of Unmanned Aerial Vehicles (UAVs) and their unique vantage points present both new opportunities and challenges for person Re-Identification (ReID). Uncertain rotations and scale variations of targets in UAV images, coupled with complex environmental factors, hinder existing methods from extracting robust feature representations. Some methods either make minor modifications to the traditional model architecture or apply simple image rotations but still fail to effectively address the challenges of UAV person ReID. To overcome these limitations, we propose a novel Data Augmentation and Rotation Invariance (DARI) algorithm. First, rotation-invariant convolution is introduced to adaptively extract features, mitigating the uncertainty caused by target rotation. Second, a refined data augmentation correction strategy is employed to reduce noise interference by increasing the richness of global features at different stages. Additionally, considering that multiple features of the same identity should yield consistent recognition result, invariant constraints are designed to enhance the clustering effect. We conducted extensive experiments on both UAV and fixed-camera datasets. The results on PRAI-1581 demonstrate a 5.6% and 6.1% improvement in mAP and Rank-1, respectively, compared to baseline. These findings highlight the model's effectiveness in addressing the challenges of UAV ReID, demonstrating its robustness and superiority. The source code will be released at https://github.com/ZFZ314/DARI.