基于监控摄像头CCD/CMOS传感器的智能核辐射监测技术研究

核安全是国家安全的重要组成部分，核辐射监测是核安全保障和核应急的必要手段。建立一个全面覆盖的核辐射监测网至关重要，但由于专用设备价格昂贵，目前国控点布局数量有限。本课题旨在研究基于监控摄像头的CCD/CMOS传感器的核辐射探测方法，通过蒙特卡罗模拟和测试实验相结合的研究方法，实现可见光和核辐射(X/γ射线)的同时、实时、高效探测。针对不遮光动态场景通过图像预处理及深度学习方法研究开发核辐射亮斑实时提取算法，最大化提取核辐射事件并降低误判率；硬件上设计并优化一种新型核辐射放大器，并应用滤光技术提高探测灵敏度；研究运动人员轨迹提取算法，并进一步基于信息融合技术实现多摄像头信息关联，实现放射源快速识别和定位。本研究成果预期可应用到现有视频监控网络，成为一种低成本高覆盖的核辐射监测手段。

Nuclear safety is a significant component of national security. Nuclear radiation monitoring is an essential means for nuclear security and nuclear emergency. It is crucial to establish a comprehensive coverage of nuclear radiation monitoring network, however due to the high price of dedicated devices, the number of national control points is currently limited. This project aims to study the method of nuclear radiation detection based on the CCD/CMOS sensor of surveillance cameras to realize simultaneous, real-time and high-efficiency detection of visible light and nuclear radiation (X/γ-ray). A combination of Monte Carlo simulations and experiments will be used. The nuclear radiation spot recognition algorithm will be developed based on image preprocessing and deep learning technology to eliminate visible light interference, maximize the extraction of nuclear radiation events and reduce the false positive rate. Improved methods of adding a nuclear radiation amplifier and a filter on the hardware are proposed to further improve the detection sensitivity. The motion trajectory extraction algorithm and the information fusion technology will be studied to realize the rapid identification and localization of the radiation source for the surveillance camera network. The research results can be employed to the existing video surveillance network and become a low-cost and high-coverage intelligent nuclear radiation monitoring method.