Tracing technology of personnel in mine video images based on random features

In order to solve a low iluminance, inhomogeneous iluminance,serious varied iluminance and lack of color information in the underground mine,due to the s imilar visual apparent and background of the personnel in underground mine would bring a difficulty to the technology development of the personnel positioning in the unde rground mine based on the visual sense,the application of the random projection technology proposed a simple,novel and effective method to trace and position the person nel in the mine video images based on the compressed sensing theory. A very sparse random projection matrix was applied to extract the target features from the multi dim ension texture feature space of the images to form the target model,the Naive Bayes classification with the identification method was applied to determine the location of th e tracing target and the random features were applied to the suitable update of the target model. The operation results of the underground video collected from Daliuta Min e of Shendong Group showed that the calculation method would have high robustness to the target shielding,rotary and uneven environment iluminance and iluminance s erious variation. The average tracing frame speed could be 50 frame per s and could provide the references to the positioning technology of the underground personnel ba sed on the visual sense.