Regarding the third experiment – the determination of the crown base height of individual trees – our study highlights that crowdsourcing can be a tool to obtain values with even higher accuracy in comparison to an automated computer-based approach. ![]() For the more complex tasks of the second Experiment 2 (multiple answer classification) the accuracy ranges from 65% to 99% depending on the label class. For our first experiment (binary classification with single answer) we obtain an accuracy of 91%, a sensitivity of 95% and a precision of 92%. Furthermore, the results of our empirical case study reveal that the accuracy, sensitivity and precision of 3D crowdsourcing are high for most information extraction problems. Our results for three different experiments with increasing complexity indicate that a single crowdsourcing task can be solved in a very short time of less than five seconds on average. ![]() We design web-based 3D micro tasks tailored to assess segmented LiDAR point clouds of urban trees and investigate the quality of the approach in an empirical user study. In this paper, we propose a method to crowdsource the task of complex three-dimensional information extraction from 3D point clouds.
0 Comments
Leave a Reply. |