基于孪生Bi-LSTM网络的线状要素相似判别模型研究OA

Traditional methods of similarity discrimination between linear elements suffer from algorithmic complexity and difficulty in determining thresholds.Addressing this problem,we proposed a linear element similarity discrimination model based on twin bidirectional long short-term memory(Bi-LSTM)network.We utilized twin Bi-LSTM networks to model the input pairs of linear elements,learning the spatial geometric characteristics of these elements.After combining the element vectors,we obtained similarity judgments.Experiments demonstrate that this model achieves F1-score of 96.34%in the similarity discrimination of linear elements in river element datasets,exhibiting superior performance compared to traditional algorithms,which can offer a novel approach to geometric similarity discrimination among linear elements.