Abstract: Fire video image analysis is an important part of fire investigation and one of the key methods to accurately identify the fire sites and cause. In video image of the fire, we can accurately calibrate the location of the fire points through the information of light and shadow, which can observably improve the efficiency of fire investigation. Based on a large amount of video data from actual fire cases, and combined with light and shadow formation rules, monitoring camera imaging rules and actual fire cases, we summarize the methods of position calibration of light source or fire sites in typical fire video images, such as direct reference calibration method, light shadow inference method, simulation experiment comparison method, and photogrammetric calculation method. The study can improve the accuracy and reliability of position calibration of fire investigators in the process of video image analysis.

Key words: position calibration, fire investigation, video analysis, light shadow inference, photogrammetry