Abstract
This study proposes a three-dimensional (3D) position measurement method that uses two cameras fixed at arbitrary positions. Unlike many existing techniques, the proposed method allows a direct 3D position computation from the image coordinates. The direct image-to-space relations are redundantly obtained and can be used to make efficient 3D computations. The proposed model can be easily implemented using multilayer feedforward artificial neural networks due to directness. In our test, four small neural networks were constructed using different combinations of image coordinates as the network inputs. The most accurate network was selected. As a result, we could easily train the neural networks and obtain better results compared to existing methods.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have