Application Of Wavelet-Transform To Improve The Precision Of Vehicle Platform Attitude

In the close shot target three-dimension survey system, the position of objects is obtained by the attitude angle based on GPS platform position, and the attitude parameters are computed by the fixed position relation of GPS flat roof, LS and CCD. Finally, based on the relative observation data of LS and CCD, the position of objects in local coordinate system can be obtained. Therefore, we can draw a conclusion that the precision of GPS attitude angles has a closed relation to the precision of the position of objects. The crucial problem is how to improve the precision of GPS platform attitude angles.In order to improve the precision of attitude angles, the common method is the least -square curve imitated arithmetic after attitude angles are determined .The precision of attitude determined is usually dissatisfied using this method. In order to improve the precision of attitude angles, a new method called wavelet denoising is introduced in this paper.In this paper, in consideration of improving the precision of GPS platform attitude angles, some researches based on the observation data of vehicle platform in the close shot target three-dimension survey system have been done. This paper introduces at length the theory of a wavelet transform denoising algorithm based on modulus maximum, spatially selective noise filtration, the denoising method of wavelet-threshold, the denoising method based on translation invariant and the denoising method based on wavelet packet.Then with the experimental data, this paper analyses the advantage of those denoising methods in the same condition.Compared with the signal to noise ratios (snr) of denoising results acquired by the least-square curve imitated method merely, the results acquired by the wavelet denoising methods which have been mentioned above are better. In the denoising method based on wavelet-threshold, when rigrsure threshold selection scheme and sln or mln of threshold rescaling method are chosen, the snr is better except the dissatisfied vision effect of signal. In order to conquer this shortcoming, a new method, using the least-square curve imitated method for the denoised data is proposed in this paper. Compared with rigrsure threshold selection scheme, the snr with this new method is smaller but the denoising result is betterthan the result with the least -square curve imitated method, and the vision effect of signal isalso better.