The accuracy of random forest (RF) classification depends on several inputs. In this study, two primary inputs—training sample and features—are evaluated for road classification from an unmanned aerial vehicle-based point cloud. Training sample selection is a challenging step since the machine learning stage of the RF classification depends greatly on it. That is, an imbalanced training sample might dramatically decrease classification accuracy. Various criteria are defined to generate different types of training samples to evaluate the effectiveness of the training sample. There are several point features that can be used in RF classification under different circumstances. More features might increase the classification accuracy, however, in that case, the processing time is also increased. Point features such as RGB (red/green/blue), surface normals, curvature, omnivariance, planarity, linearity, surface variance, anisotropy, verticality, and ground/non-ground class are investigated in this study. Different training samples and sets of features are used in the RF to extract the road surface. The experiment is conducted on a local road without a raised curb located on a relatively steep hill. The accuracy assessment is conducted by comparing the model classification results with the manually extracted road surface point cloud. It is found that the accuracy increases up to around 4%–13%, and 95% overall accuracy was obtained when using convenient training samples and features.
Unmanned Aerial Vehicle (UAV) technology is one of the fastest-growing technologies especially used in image processing. Structure-from-Motion (SfM) based software are usually used to convert two-dimensional UAV-based images into three-dimensional (3D) data. Then, objects such as buildings, trees, and roads can be classified from the 3D data for further analysis. In this study, the road surface generated from 3D data was evaluated. There are several factors that affect the accuracy of the 3D data. In this study, two factors, namely UAV flight altitude and SfM based software, were evaluated. Two different flight altitudes, which were 35 meters and 50 meters, were used. It was found that the lower flights with closer altitudes did not make a significant difference on the results and produced similar results. Another factor is different SfM based software. Two well-known SfM based software were used in this study, which were the Agisoft Metashape and Pix4D Mapper. In this case study, it was found that the Agisoft Metashape software produced more accurate and faster results than Pix4D Mapper software.
Roads are one of the main characteristics of cities, and their data should be updated periodically. In this study, a new automatic method is proposed for extracting road surface information and road inventory from a Mobile LiDAR System-based point cloud. The proposed method consists of four steps. First, a three-dimensional point cloud is acquired using the MLS raw data. To improve the extraction accuracy, irrelevant points are removed from the point cloud. Piecewise linear models are used in the third step to classify the road surface. Road geometric characteristics such as centerline, profile, cross-section, and cross slope are extracted in the final step. The manually obtained road boundary is compared with the extracted road boundary to assess the classification results. Completeness, correctness, quality, and accuracy measures are range from 97% to 99%. When comparing these measures with previous studies, the proposed method produces one of the highest ones.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.