Abstract: Many sensors usually are installed in a landslide body.So,their comprehension information can be extracted.Only one key monitoring sensor instead of the others is usually used in the prediction model at the same time in the landslide at present.The present prediction model can't make a full use of the monitoring data from the many sensors.The errors from the sensors cannot be considered in the model too.In order to overcome the above shortcomings,multi-sensor valuation fusion theory has been developed and used.The fusion efficiency of multi-sensor is theoretically proved to be better than that of any single sensor.Analysis of landslide is carried out by using a pixel-level fusion method.The result identifies that the landslide had experienced slowly deformation stage form July 2006 to September 2007and uniform motion deformation stage form September 2007 to September 2008and accelerated deformation stage form September 2008 to August 2009and rapidly deformation stage form August 2009 to February 2010.Such result is consistent with the law of nature.Two different periods of transitional,happened in August and September,which reflected the rain is the main factors caused landslide sliding.And the feasibility and effectiveness of the method is also proved at the same time.