Optimizing Aircraft Trajectory Prediction: A Hybrid TCN-LSSVM Model with Image Recognition

This paper proposes a novel TCN-LSSVM model for precise trajectory prediction. An approach that integrates the image recognition and the trajectory prediction is designed, which improves the prediction process. By combining event and depth cameras, multi-dimensional images are transformed into single-dimensional data. This hybrid approach to information access can enhance the model's learning efficiency and reinforce ability to discern intricate patterns in movement data. The processed single-dimensional data is put into a trajectory prediction model for processing. The trajectory prediction phase includes causal convolution and regularization to optimize prediction accuracy, reduce errors and improve real-time performance, taking full advantage of the combined benefits of TCN and LSSVM to improve performance. Comprehensive comparisons are made with existing prediction algorithms, and the results demonstrate the superior accuracy of the designed TCN-LSSVM model, which reduces error margins and underscores the efficacy in leveraging image recognition for trajectory prediction. This innovative model is a significant contribution to the field and provides a powerful framework for future applications in relevant scenarios.