“The research team proposed a self supervised fire trail recognition method based on spectral shape encoding, which converts multispectral reflectance into morphological feature vectors through a hybrid encoding strategy of OPE, AE, and SAE. The SimCLR framework is introduced to achieve self supervised pre training, maintaining high discriminative ability even with limited labeled samples. This provides a solution to the problem of accurate monitoring of fire trails in complex scenes and opens up new directions for constructing spectral morphology driven detection frameworks.”