Tracking a moving source based on a computer vision system: Improving detection using data correlation

Abstract

Data fusion between different sensors can improve the detection of nuclear threats by extracting more reliable and effective information. In this study, tracking a moving radioactive hotspot source using a combination of a radioactive detector (NaI) and a surveillance camera is addressed. For this purpose, three mobile robots were used, and a radioactive source was placed on one of these robots. An algorithm was developed to correlate the radioactive and camera data, so the robot with the highest correlation was selected as the moving source quickly. By increasing the acquisition time from 5 to 125 seconds, the algorithm's success rate in detecting the moving radioactive source increases from 42.7% to 98.3%. In addition, the moving source's detection speed and the detection's precision over different times were studied. The results have presented a model that can be scaled up by equipping surveillance cameras with radioactive detectors to provide a network, and this network can continuously monitor and control a vast area or even a city to detect and track suspicious items.

Keywords

• data fusion
• computer vision
• moving source
• correlation
• radioactive material out of regulatory control