Tunnel disease detection method based on multi-scale feature pyramid

The invention discloses a tunnel disease detection method based on a multi-scale feature pyramid. The problem of low multi-scale target detection precision in a complex scene in the prior art is mainly solved. According to the implementation scheme, data in a target detection database are read, and images of the data are preprocessed; using a ResNet convolutional neural network as a backbone network to extract preprocessed image features; constructing a multi-scale feature pyramid structure composed of an efficient multi-scale attention network and a feature filtering network; inputting the preprocessed image features into a multi-scale feature pyramid structure, selecting the input image features through an efficient multi-scale attention network of the multi-scale feature pyramid structure, performing feature filtering on the features after feature selection through a feature filtering network, and repeating the process to obtain multi-scale pyramid features; and coding and decoding the multi-scale pyramid