Abstract:

In order to prevent accidents caused by irrelevant personnel entering the operation area of gas transmission station, the early prediction of personnel location was necessary. Therefore, a method integrating personnel moving intention and trajectory was proposed and the intelligent risk warning model of personnel at gas transmission station using"intention + trajectory"based on machine vision was built. With the personnel walking images collected, features of personnel head were extracted through the Histogram of Oriented Gradient(HOG). Then the Support Vector Machine(SVM) classifier was used to classify and recognize the head direction. Thus the intention of moving direction (normal straight ahead, straight ahead under observation, and intent to turn) was preliminarily predicted based on the head direction. The situation where people intend to turn was tracked continuously and the Kalman filtering algorithm was used to predict moving trajectory. Therefore, different risk scenarios were predicted and layered early warning was achieved. The research results show that the accuracy rate of moving intention identification is 90.79%, and the correlation coefficients between the predicted curves and actual trajectory ones are r₁=0.994 84, r₂=0.993 43. It indicates that the predicted curves and actual trajectory ones are highly correlated and the accuracy of prediction is high. The method can achieve accurate early warning of the intrusion of irrelevant personnel at the gas transmission stations.

Key words: gas transmission station, moving intention, trajectory prediction, risk early warning, machine vision