基于CNN的康普顿背散射图像中违禁品分割方法

Contraband Segmentation of Compton Back-Scattering Images Based on CNN

康普顿背散射(CBS)技术是一项较新的射线安检技术,可以提高复杂背景下安检设备对爆炸物等有机违禁品的探测力度,这其中一个重要环节就是图像中违禁品的分割问题.本文提出了一种基于细胞神经网络(CNN)的CBS图像滤波及分割方法,在此基础上又提出了一种基于CNN和数学形态学的孤立点滤除方法,并对这两种方法进行了详细分析,给出了实例的仿真结果,验证了方法的有效性.本文提出的方法为并行处理算法,易于大规模集成电路(VLSI)实现,满足安检设备对CBS图像处理的实时性要求.

Compton back-scattering（CBS） is a new technology for ray security inspection,that can enhance accuracy of explosive detection,such as organic contraband.For this task,contraband segmentation of image is very important.One new approach for Compton back-scattering（CBS） image filtering and segmentation using CNN is proposed in this paper.Furthermore,one new way for eliminating isolated point based on CNN and morphologic method is also described.Some detailed analyses and practical results are presented,which demonstrates the successful operation of the proposed algorithm.This new approach is very affordable to parallelism and analog VLSI implementation,which allowing the security inspection of CBS image processing to be performed in real-time.