A USER-FRIENDLY EVALUATION TOOL FOR POINT CLOUD CLASSIFICATION AND SEGMENTATION

II. INTERNATIONAL HAZAR SCIENTIFIC RESEARCHES CONFERENCE, Baku, Azerbaycan, 10 - 12 Nisan 2021, cilt.1, ss.841-852

In recent years, point cloud data has been begun to use in many application areas such as

robotic and architecture, thanks to developments in 3D sensing technologies. Simultaneously,

researchers proposed a variety of advanced techniques for the classification and segmentation

of point clouds. However, evaluation tools have been remained scarce despite producing

metric and visual results is a time-consuming process. The main contribution of this study is

to present a user-friendly evaluation tool for point cloud classification and segmentation. We

used the QT library to build the interface and Point Cloud Library (PCL) to process point

cloud data. In the classification problem, the number of classes is known in advance, and the

points in the test sample must belong to one of these classes. We utilized this information to

produce results. On the other hand, evaluating segmentation results is problematic since the

number of segments is undetermined. The proposed tool offers two options (automatic and

manual) to determine paired segments. The evaluation tool also provides a visual result as

updating the test sample. The correct results can be stained with the same colors related to

classes or paired segments in the ground truth. The visual result can be saved as an image or a

point cloud, while the metric results can be stored in Excel datasheets. The second

contribution of this study is to utilize the set data structure to speed up the evaluation and

other functional processes. The experiments were conducted to examine the set data

structure's efficiency by comparing with the K-Nearest Neighbor (KNN) search method. The

experimental results exhibited that the set data structure significantly decreases the processing

time. The tool is publicly available for the researchers