| 摘要 |
The present invention relates to a method for recognizing and tracking a fire image using a Kalman filter. The present invention comprises the steps of predicting an estimation value of a fire location and an error covariance of the estimation value; calculating a Kalman filter gain of minimizing an error of the estimation value of the fire location; measuring the fire location of an fire image, estimating a fire statue by a measurement value, and calculating the estimation value of the fire location to be outputted as an output value of the fire location; calculating an estimation error covariance to identify accuracy of the calculated estimation value of the fire location; and predicting a new estimation value and an error covariance of the new estimation value by compensating the measurement value and an prediction value by the calculated estimation error covariance. The present invention is able to reduce an error generated due to a system noise and a measurement noise, thereby providing the method for more exactly tracking the location of the fire relatively to the existing method for recognizing and tracking a fire image. |
