Abstract: This article focuses on the method of objection from hash background. Doppler radar is a kind of moving object detection radar by means of the Doppler domino effect. Among all kinds of the interferes,the wind and the movement of leaves and branches are the serious interferes.Which,in some cases,will make the radar find nothing.People have made many kinds of methods to overcome the effect of those interferes,but few can match the request of radar object auto-recognition.On the other hand,the range of the frequency and the time can be hardly known when thoses occure.In order to match the request of data for the program of object recognition, the authors put forward a method of adaptive filter designing by means of wavelet packet analysis. Firstly, it transform the data by means of FFT, from the graph of the frequency.It can be found that the peak of those interferes is higher than the peak of the object when those interferes occur. Secondly, by means of non-linear adding power of the frequency, it can elevate the peak of the object, which is useful for the computer to calculate the real frequency of the object. Thirdly, by means of adaptive wavelet packet analysis, it can find the frequency range of those interfere automatically. When it re-conducts the signal, it can get rid of those interferes.