STBT: 1spatio-temporal belt transformer prediction network for belt deviation

Belt conveyor is an important equipment for the transportation of raw materials in ports, wharves and industrial sites. Whether it runs stably is directly related to whether the whole production process can operate normally, and belt deviation is the most prone type of failure of belt conveyor. This paper proposes a Spatio-temporal Belt Transformer (STBT) method based on deep learning and the Encoder-Decoder network based on Transformer Block. The continuous frame belt motion images fused with temporal and spatial features are input, and the fusion information of global features and local features is extracted by the multi-head self-attention algorithm based on sliding window, and the final output of the predicted continuous frame results is simple and efficient. The experimental results show that STBT can predict the future deviation state of the belt well.