Short-time Prediction of Image Motion Trajectories Based on Convolutional Long and Short-term Memory Neural Networks

Traditional image motion trajectory prediction is based on the analysis of the observation data to make numerical predictions. This prediction method has the problems of low prediction efficiency and low accuracy when the data is large. In this paper, a prediction method based on Convolutional long and short-term memory networks model (ConvLSTM) is proposed for the short-term prediction of the movement trajectory of meteorological cloud images. And the Structural Similarity (SSIM) loss function is introduced in the model training, which effectively improves the accuracy of prediction and reduces MAE and MSE, where MAE is as low as 0.017, and MSE is as low as 0.00009. The experimental results show that the ConvLSTM-SSIM network model designed in this study can effectively extract the temporal and spatial characteristics of image sequence and realize the short-term prediction of image motion trajectory, it also provides some reference for other fields that needs short-time image sequence prediction.