一种基于X射线图像和特征曲线的危险品检测方法

Dangerous goods detection based on X-ray image and characteristic curve

目前,车站行李安检系统检测危险品主要依靠人眼观察安检仪监视器识别危险品,容易出现漏检、错检以及视觉疲劳等问题。为此,提出一种基于X射线图像和特征曲线的危险品检测方法,该方法依据X射线安检仪获得的平均有效原子序数将物体分为两大类:当物体的平均有效原子序数大于等于10时,对X射线图像采用小波将其分解为近似、水平、垂直和对角线方向的4个子图像,并分别采用不同的边缘检测算子进行边缘提取,然后进行小波重构和形态学后处理以提取完整目标轮廓,通过目标与标准危险品图像进行轮廓曲率角点匹配,实现危险品的形状检测;当平均有效原子序数小于10时,获取被检物体的高低能灰度值进行曲线拟合,并与标准危险品的特征曲线进行比较以判别是否存在危险品。实验结果表明,利用图像形状和高低能灰度特征曲线相结合的方式对旅客携带物品进行检测,可以较全面地检测出危险品。

Station luggage screening systems detecting dangerous goods mainly rely on the human eyes to observe the monitor to identify dangerous goods. But there may be false or missed detection in the observa- tion process. A method for dangerous goods detection based on X-ray image and characteristic curve is proposed, where goods of passengers are divided into two categories according to the average effective atomic number of the scanned objects. Shape image detection is employed to detect the dan when the object average effective atomic number is ≥ 10, in this case, X-ray image is firstly into four sub-images by wavelet transform, and edges ge d rous goods ecomposed of each sub-image are detected by Canny, Sobel and Roberts algorithms, respectively. Then the edge image is reconstructed from the four edge sub-images by wavelet reconstruction transform. To remove small regular details and fill the holes of the objects of the bi- nary image, and finally contours curvature corner detection and matching are employed to detect danger- ous objects. When the average effective atomic number is 〈 10, a characteristic curve matching is used to detect dangerous goods. Experiments show that most dangerous goods carried by tourists or travelers can be detected by the combination of shape and characteristic curve matching.