2022
09/22
相关创新主体

创新背景

仅用传统的x射线很难发现一些爆炸物,而这种新方法可能会彻底改变毒品、非法野生动物和爆炸物等非法物品的检测方式。研究结果对安全部门具有重要意义,有可能支持医疗保健和工业。

 

创新过程

在这篇发表在《自然通讯》上的论文中,工程师们将一种新的x射线测量技术与人工智能机器学习结合起来,并在定制的x射线安全扫描仪中测试了它,使用了含有隐藏爆炸材料的物体和安全物体。

 

 

研究负责人奥利沃教授在早期的研究中发现,物体的微观变化或不规则会导致x射线光束在穿过物体时弯曲。他目前的工作是通过他的新兴技术讲座探索这一现象的潜在应用,该讲座由英国皇家工程院资助。这种新方法依赖于测量光束穿过不同纹理物体时的微小弯曲。

资深作者Sandro Olivo教授(伦敦大学学院医学物理学和生物医学工程)解释,这是一种通过分析纹理来检测材料和物体的完全不同的方法,让研究人员有了一种检测非法材料的新方法。x射线中的微小弯曲一直都存在,但它们在传统的x射线系统中是看不到的,因此这让他们能够获得大量以前未开发的信息。

研究人员已经证明它在探测爆炸物方面非常有效,但它也可以用于任何依赖x射线的应用,如医疗成像或检测工业部件的弱点。

 

 

x射线光束的微小偏差发生在微放射线的角度,微放射线的角度大约是一度的2万倍。该团队将这些角度的测量(称为微放射线散射)与人工智能结合起来,通过它们的纹理准确识别物体和材料。在炸药上测试时,检出率为100%。

该团队通过在扫描仪中引入x射线掩模来检测散射来测试这项技术。这产生了非常详细的图像,每个像素都显示了物体的微观不规则程度。图像因其微观结构不同而不同,这在传统的x射线图像上是看不到的。该团队能够通过分析微观上的不规则性来区分危险和良性物质。

 

创新关键点

工程师们将一种新的x射线测量技术与人工智能机器学习结合起来,并在定制的x射线安全扫描仪中测试了它,使用了含有隐藏爆炸材料的物体和安全物体。

 

创新价值

该创新方法所取得的成就不仅有可能增强探测爆炸物和武器的安全应用,而且将他们的技术应用于其他材料,如非法毒品,还可以积极影响其他终端用户群体,如海关。

 

Combining new X-ray measurements with AI could improve explosives detection

In the paper, published in Nature Communications, engineers combined a new X-ray measurement technique with AI machine learning and tested it in custom X-ray security scanners, using objects containing hidden explosive materials and security objects.

Research leader Professor Olivo found in earlier work that microscopic changes or irregularities in an object can cause X-ray beams to bend as they pass through it. His current work explores the potential applications of this phenomenon through his lecture on Emerging Technologies, which is funded by the Royal Academy of Engineering. The new method relies on measuring the tiny bends of light beams as they pass through objects of different textures.

Senior author Professor Sandro Olivo (UCL Medical Physics and Biomedical Engineering) explains that this is a completely different approach to detecting materials and objects by analyzing their texture, giving researchers a new way to detect illicit materials. Tiny bends in X-rays have always been present, but they are not visible in conventional X-ray systems, so this allows them to access a wealth of previously untapped information.

Researchers have shown it to be very effective at detecting explosives, but it can also be used in any application that relies on X-rays, such as medical imaging or detecting weaknesses in industrial components.

The tiny deviation in the X-ray beam occurs at the Angle of the microbeam, which is about 20,000 times greater than one degree. The team combined measurements of these angles, called microradiation scattering, with artificial intelligence to accurately identify objects and materials by their texture. When tested on explosives, the detection rate is 100%.

The team tested the technique by introducing an X-ray mask into the scanner to detect scattering. This produces very detailed images, with each pixel showing the degree of microscopic irregularity of the object. The image is different because of its microstructure, which is not seen in traditional X-ray images. The team was able to distinguish dangerous from benign substances by analyzing microscopic irregularities.

智能推荐

  • 通过重新实现和集成分页在边缘设备上训练神经网络

    2022-07-26

    UC 伯克利的几位研究者提出了一种用于深度神经网络的图形级编译器——PORT,它重写了大型模型的训练 DAG以适应边缘设备的内存限制,同时保持高能效,通过重新实现和分页,达到了以最小的能耗扩展有效的内存容量的效果。

    涉及学科
    涉及领域
    关键词 
  • 创新使用无人机技术简化排雷过程

    2022-09-05

    创新将先进的无人机技术用于排雷工作,检测覆盖传统排雷方法无法覆盖的范围,简化排雷工作,增加排雷操作员的技术安全性。

    涉及学科
    涉及领域
    关键词 
  • 创新3D打印项圈帮助神经系统疾病患者

    2022-08-03

    一位UCL临床研究人员开发了一种创新的3D打印颈环,帮助改变了患有严重神经系统疾病的人的生活。

    涉及学科
    涉及领域
    关键词 
  • 结合视觉传感器和神经网络的新系统使机器人具有塑造弹性物体的能力

    2022-08-17

    研究团队建立的RoboCraft新系统,可以通过视觉传感器“看见”,然后利用基于图形的神经网络进行学习,再通过算法规划机器人的行为。这个新系统可以让机器人处理类似于橡皮泥的柔软材质,为机器人学习如何实现复杂的家庭或工业任务奠定了基础。

    涉及学科
    涉及领域
    关键词