Oil spill detection method based on time sequence feature focusing network and gated cyclic convolution unit

The invention provides an oil spill detection method based on a time sequence feature focusing network and a gated cyclic convolution unit, and the method comprises the steps: constructing a time sequence feature focusing network model, calculating the attention weight of each channel, and applying the attention weight of each channel to a channel of a feature map; constructing a gating cyclic convolution model; and based on the constructed time sequence feature focusing network model and the gated cyclic convolution model, oil spill detection is realized. According to the technical scheme, the processing capacity of channel-level important information in the input feature map is enhanced based on the time sequence feature focusing network model, so that the input feature map is more suitable for processing image data with complex space attributes, and the problem that images of any size cannot be input when the image sequence problem is processed is solved through the gating cyclic convolution model; by comb