A fast classification method based on sequential Bayesian analysis for figures in irradiation images

Abstract

Radiography is often employed to analyze whether a closed container contains the targeted object in customs inspection, medical diagnosis and arms control verification. Based on the analysis of whether the irradiation image of the closed container contains the specific figure of the targeted object, one can determine the existence of the targeted object effectively. In the figure determination of the irradiation image, the classification and identification of the figures is a key technique. In many cases, considering the time cost of radiography, the existence determination of the targeted object should be done as fast as possible, and fast classification and identification of the figures in the irradiation image is required. In this paper, a fast figure classification method based on the sequential Bayesian analysis with the ability of fast classifying the figures in irradiation images is proposed. First, the theoretical framework of this method was established. Second, the feasibility of this method in analyzing positive irradiation images (the data point density of the image region where the inspected object locates is higher than those of other image regions) was proved by establishing its mathematical models. Third, through the Monte Carlo numerical simulation, the feasibility of this method in analyzing negative irradiation images (the data point density of the image region where the inspected object locates is lower than those of other image regions) was also proved. At last, the feasibility of this method in classifying the complicated figures in irradiation images by establishing and identifying figure templates was proved through the numerical simulation calculation. • A fast figure classification method based on the sequential Bayesian analysis is proposed. • This method can make fast classification of the figures in irradiation images. • The theoretical framework of this method was established. • Numerical simulation studies prove the feasibility of this method.