Abstract: To the multi-sensor detection system,its measurement information is non-linear and spatiotemporal variation,etc.so the basic requirement of information fusion method is robust and parallel processing capabilities.In this paper,the wavelet analysis theory is applied to denoise the original data,which makes the data after denoising is better to reflect the nature of soil and its Variation Law.Robust least squares estimation technique can be integrated with different sensors data,to remove redundancy,to overcome the ambiguity,so it can obtain more comprehensive,more accurate and reliable information than any single sensor.Considering highly uncertain and empirical of the natural environment in farmland soil,the information fusion estimation method based on wavelet denoising and robust least squares is proposed.The analysis of experimental results shows that the method is feasible and worth studying.