In order to achieve high precise height value of laser proximity fuze, a kind of Kalman filter method was provided which estimates the characteristics of measurement noise periodically. This method can filter proximity fuze height finding noise whose statistic characters vary with ground surface characters. The statistic relationship between height finding value difference serial and measurement noise serial was analyzed. Unbiased estimation of measurement noise variance was derived, which applied the height finding value. Non-parameter hypothesis test was applied to identify the type of noise, and then the variance of measurement noise was estimated periodically. Against the background of ARM blast at preconceived height, simulation computation was implemented. The result indicates that this method could properly solve the Kalman filter problem that measurement noise statistic characteristics vary unpredictably.