Ground segmentation in LiDAR point clouds is a foundational capability for autonomous systems, enabling safe navigation in applications ranging from urban self-driving vehicles to planetary exploration rovers. Reliably distinguishing traversable surfaces in geometrically irregular or sensor-sparse environments remains a critical challenge. This paper introduces a hybrid framework that synergizes multi-resolution polar discretization with sparse convolutional neural networks (SCNNs) to address these challenges. The method hierarchically partitions point clouds into adaptive sectors, leveraging PCA-derived geometric features and dynamic variance thresholds for robust terrain modeling, while a SCNN resolves ambiguities in data-sparse regions. Evaluated in structured (SemanticKITTI) and unstructured (Rellis-3D) environments, two different versions of the proposed method are studied, including a purely geometric method and a hybrid approach that exploits deep learning techniques. A comparison of the proposed method with its purely geometric version is made for the purpose of highlighting the strengths of each approach. The hybrid approach achieves state-of-the-art performance, attaining an F1-score of 95.4% in urban environments, surpassing the purely geometric (91.4%) and learning-based baselines. Conversely, in unstructured terrains, the geometric variant demonstrates superior metric balance (80.8% F1) compared to the hybrid method (75.8% F1), highlighting context-dependent trade-offs between precision and recall. The framework’s generalization is further validated on custom datasets (UMH-Gardens, Coimbra-Liv), showcasing robustness to sensor variations and environmental complexity. The code and datasets are openly available to facilitate reproducibility.

This paper presents a comparison of different image fusion methods for matching visible-spectrum images with thermal-spectrum (far-infrared) images, aimed at enhancing person detection using convolutional neural networks (CNNs). While object detection with RGB images is a well-developed area, it is still greatly limited by lighting conditions. This limitation poses a significant challenge in image detection playing a larger role in everyday technology, where illumination cannot always be controlled. Far-infrared images (which are partially invariant to lighting conditions) can serve as a valuable complement to RGB images in environments where illumination cannot be controlled and robust object detection is needed. In this work, various early and middle fusion techniques are presented and compared using different multispectral datasets, with the aim of addressing these limitations and improving detection performance.

Frente a las limitaciones que suponen las condiciones ambientales y de iluminación para el procesamiento de imágenes del espectro visible, la fusión con imágenes de otros espectros constituye una estrategia prometedora para la mejora de la percepción. Este artículo presenta un estudio comparativo de métodos de fusión por reproyección adaptativa que emplean transformaciones en el dominio de la frecuencia o de la varianza para combinar datos de imágenes RGB y térmicas. Se evalúan métodos basados en Análisis de Componentes Principales (PCA), Análisis Factorial (FA), y fusión basada en Wavelets y Curvelets, todos ellos integrados en un flujo de detección con YOLOv8. Los experimentos se realizan sobre el \textit{dataset} KAIST, con especial atención al rigor metodológico y la reproducibilidad. Los resultados se comparan con métodos previos, mostrando sus fortalezas pero también sus limitaciones y desventajas frente a métodos más clásicos. Finalmente, se discuten las implicaciones para futuras investigaciones y el valor de un diseño experimental robusto como punto indispensable en el avance del estado del arte de la fusión de imágenes multiespectrales.