This paper proposes a golden sine grey wolf optimizer (GSGWO) that can be adapted to the obstacle-crossing function to solve the path planning problem of obstacle-crossable robot. GSGWO has been improved from the gray wolf optimizer (GWO), which provide slow convergence speed and easy to fall into local optimum, especially without obstacle-crossing function. Firstly, aiming at the defects of GWO, the chaotic map is introduced to enrich the initial population and improve the convergence factor curve. Then, the convergence strategy of the golden sine optimizer is introduced to improve the shortcomings of GWO, such as insufficient convergence speed in the later stage and the ease with which it falls into the local optimum. Finally, by adjusting the working environment model, path generation method and fitness function, the path-planning problem of the obstacle-crossing robot is adapted. In order to verify the feasibility of the algorithm, four standard test functions and three different scale environment models are selected for simulation experiments. The results show that in the performance test of the algorithm, the GSGWO has higher convergence speed and accuracy than the GWO under different test functions. In the path-planning experiment, the length, number and size of inflection points and stability of the path planned by the GSGWO are better than those of the GWO. The feasibility of the GSGWO is verified.

In post-disaster search and rescue scenarios, the accurate image segmentation of individuals is essential for efficient resource allocation and effective rescue operations. However, challenges such as image blur and limited resources complicate personnel segmentation. This paper introduces an enhanced, lightweight version of the Pyramid Scene Parsing Network (MHC-PSPNet). By substituting ResNet50 with the more efficient MobileNetV2 as the model backbone, the computational complexity is significantly reduced. Furthermore, replacing the ReLU6 activation function in MobileNetV2 with H-Swish enhances segmentation accuracy without increasing the parameter count. To further amplify high-level semantic features, global pooled features are fed into an attention mechanism network. The experimental results demonstrate that MHC-PSPNet performs exceptionally well on our custom dataset, achieving 97.15% accuracy, 89.21% precision, an F1 score of 94.53%, and an Intersection over Union (IoU) of 83.82%. Compared to the ResNet50 version, parameters are reduced by approximately 18.6 times, while detection accuracy improves, underscoring the efficiency and practicality of the proposed algorithm.

The safety inspection system for the underside of vehicles demands both high detection speed and accuracy, necessitating a model with small parameters and high network efficiency. To address these issues and improve the real-time performance of under-vehicle safety inspection, the YOLOv8n object-detection algorithm was enhanced, resulting in the PSP-YOLO algorithm. Firstly, the improved PSP-YOLO network introduces a P2 small-object detection head, which is particularly effective in detecting tiny objects, such as small dangerous objects under vehicles, thereby significantly enhancing the ability to detect such small targets. Secondly, the Space-to-Depth Conv (SPD-Conv) module was introduced into the backbone network, which improves the detection performance of low-quality samples and small objects while ensuring high precision and accurate localization of small targets in images. Finally, the Partial Convolution (PConv) lightweight module was also added to the feature fusion network, effectively reducing the model’s parameters and computational load, conserving computing resources, and improving detection speed and accuracy. Experimental results demonstrate that the modified model increased mAP@0.5 by 3.9% and mAP@0.5-0.95 by 3.6% on the dataset used, reduced the number of parameters to 2.53 M, and achieved a detection speed of 104.6 f/s. Therefore, the improved YOLOv8n algorithm significantly enhances the detection performance of dangerous objects, meeting the high-precision and real-time requirements for vehicle underside safety inspections.

3D point clouds are essential for representing geometric structures in various fields such as autonomous driving and virtual reality. However, real-world data often suffers from incompleteness due to occlusions and noise, and existing completion methods typically rely on paired complete–incomplete training data or are limited to recovering relatively small missing regions, which restricts their effectiveness under high missing-rate scenarios. This paper introduces a GAN-based method for completing 3D point clouds, capable of reconstructing detailed structures from partial inputs. Our end-to-end framework, consisting of an encoder, generator, and discriminator, optimizes topological accuracy and spatial continuity through a multi-term joint loss. Experimental results on the ModelNet40 dataset demonstrate superior performance over traditional and deep learning-based methods, achieving Chamfer Distance (CD = 0.085), Earth Mover’s Distance (EMD = 0.199), and F-Score (0.208). The generated high-quality point clouds support downstream tasks like path planning and robotic grasping. The source code and experimental datasets used in this work are publicly available at: DOI: https://doi.org/10.5281/zenodo.18421141.

To address the challenges of indistinct local details, difficult target boundary localization, and the concurrent occurrence of multiple hazardous behaviors in elevator shaft and car-top maintenance environments, this study proposes a high-resolution detail-enhanced hazardous behavior detection method based on D-FINE-S. First, we design a P2V2 shallow detail enhancement module that injects high-resolution P2 features into P3 after channel projection and spatial alignment, thereby supplementing local detail information without introducing an additional P2 detection head. Second, we propose a Residual-Adaptive Fine-grained Distribution Refinement (RA-FDR) mechanism, which applies bounded residual calibration to the non-uniform projection positions of the original FDR to improve the adaptability of bounding box regression to the data distribution of the target task. Overall, the final model achieves mAP@0.5 and mAP@0.5:0.95 values of 92.1% and 54.7%, respectively, representing improvements of 4.1 and 3.6 percentage points over D-FINE-S. The results indicate that the proposed method improves hazardous behavior detection and localization performance on the current dataset of elevator maintenance images.

To address the lack of traversable region awareness in conventional path planning algorithms for obstacle-crossing robots, an adaptive path planning method is proposed. First, a traversal-aware environment model is constructed by introducing graded traversable regions with associated physical traversal costs. To effectively navigate this complex model, a hybrid Ant Colony Optimization (ACO) framework integrating Jump Point Search (JPS) and the Genetic Algorithm (GA) is developed. Specifically, a JPS-inspired pruning strategy is incorporated into the state transition process to significantly reduce redundant node expansion. Crucially, genetic operators—namely crossover and mutation—are embedded within the main ACO iterative loop to dynamically sustain population diversity and effectively mitigate stagnation in local optima. Correspondingly, the pheromone initialization, state transition mechanisms, and update rules are redesigned to incorporate the robot’s obstacle traversal capabilities. The framework is further complemented by path optimization operations that reduce unnecessary turning points. Extensive simulation experiments demonstrate that the proposed method outperforms conventional ACO-based and classical path planning algorithms. In particular, it achieves an average reduction of 11.1% in path length and 65.5% in the number of waypoints, while ensuring effective coordination with the robot’s physical traversal capabilities. These results validate the superior search efficiency, robustness, and practical applicability of the proposed approach.