Abstract: For explosives recognition problem in safety inspection field， dual-energy X-ray transmission technology， low-energy forward scattering and back scattering technology were combined to get the gray-levels of dual-energy transmission， low-energy forward scattering and back scattering images taking the feature extraction and recognition of the radiation data as the core. The eigenvalue R associated with effective atomic number and the eigenvalue L associated with density were obtained. Based on the least mistake probability， the eigenvalue R and L were synthesized to get the discriminate， decision-making plane and distinguish rule. The experiments validated that the correct rate of the discrimination rules by the Bayesian decision theory is up to 90%， and it is a more effective method for solid explosives identification that improves the detection capability of X-ray comprehensively. This may significantly contribute to recognition of solid explosives.

Key words: dual-energy transmission, low-energy scattering, explosives detection, Bayesian decision, pattern recognition