This paper presents an approach for road detection based on image segmentation. This segmentation is resulted from merging 2D and 3D image processing data from a stereo vision system. The 2D layer returns a matrix containing pixel's clusters based on the Watershed transform. Whereas the 3D layer return labels, that are classified by the V-Disparity technique, to free spaces, obstacles and non-classified area. Thus, a feature's descriptor for each cluster is composed with features from both layers. The road pattern recognition was performed by an artificial neural network, trained to obtain a final result from this feature's descriptor. The proposed work reports real experiments carried out in a challenging urban environment to illustrate the validity and application of this approach.

This study investigates the optimization of tree detection in static images using YOLOv5, YOLOv8, and YOLOv11 models, leveraging a custom non-standard image bank created exclusively for this research. Objectives: To enhance tree detection in static images by comparing the performance of YOLOv5, YOLOv8, and YOLOv11 models. The comparison involved hyperparameter tuning and the application of various optimizers, aiming to improve model performance in terms of precision, recall, F1, and mean average precision (mAP). Design/Methodology/Approach: A custom image bank was utilized to train YOLOv5, YOLOv8, and YOLOv11 models. During training, the hyperparameters’ learning rate and momentum were tuned in combination with the optimizers SGD, Adam, and AdamW. Performance metrics, including precision, recall, F1, and mAP, were analyzed for each configuration. Key Results: The optimization process achieved precision values of 100% with Adam for YOLOv8 and SGD for YOLOv11, and recall of 91.5% with AdamW on YOLOv8. Additionally, mAP values reached 95.6% for AdamW on YOLOv8 and 95.2% for SGD on YOLOv11. Convergence times for mAP were also significantly reduced, demonstrating faster training and enhanced overall model performance. Originality/Research gap: This study addresses a gap in tree detection using YOLO models trained on non-standard image banks, a topic that is less commonly explored in the literature. The exclusive development of a custom image bank further adds novelty to the research. Practical Implications: The findings underscore the effectiveness of model optimization in tree detection tasks using custom datasets. This methodology could be extended to other applications requiring object detection in non-standard image banks. Limitations of the investigation: This study is limited to tree detection within a single custom dataset and does not evaluate the generalizability of these optimizations to other datasets or object detection tasks.

The navigation of an autonomous vehicle is a highly complex task and the dynamic environment is used as a source for reasoning. Road detection is a major issue in autonomous systems and advanced driving assistance systems applied for inner-city. Uncertainty may arise in environments with unmarked or weakly marked roads or poor lightning conditions. Moreover, when a common benchmark is not used, it is hard to decide which approach performs better on the road detection problem. This paper introduces a comprehensive performance analysis of two road recognition approaches using the urban Kitti-road benchmark. The first approach makes the extraction of a feature set based on statistical measures of 2D and 3D information from each superpixel. An Artificial Neural Network is used to detect the road pattern. The second approach extracts the feature set based on a multi-normalized histogram of Textons and Disptons for each superpixel. This feature set is used as a source for a Joint Boosting algorithm to model the road pattern. The proposed work presents a detailed evaluation highliting the pros and cons of each approach.

Uncertainty from urban environment arises not only by the imprecise pose estimation and noisy information in images but also by the lack of semantic information. This paper presents an approach to improve the perception capability of intelligent vehicles in complex urban environments. The method uses the meta-knowledge acquired from a built Semantic Context image and applies it on evidential grids constructed from the Stereo Vision. In the detection of semantic information problem, Texton and Dispton maps are used as a source to model a probabilistic joint boosting classifier. The evidential grids are based on occupancy grids, an approach based on the Dempster-Shafer theory that manages different sources of uncertainty arising in dynamic temporal scenes. An additional structure is created to handle a refined set of propositions originated from the meta-knowledge available. This structure bound with evidential conflict analysis enables an accurate detection of stationary and mobile objects in the perception grid. The proposed work reports real experiments carried out in a challenging urban environment using the KITTI benchmark in which meaningful evaluation can be done to illustrate the validity and application of this approach.