Abstract: The ground raindrop spectrum observation method is studied. The traditional filter paper sampling method is improved through enlarging the sample area, and increasing the methods of auxiliary observation. The image automatic identification software is designed and put into application in the processing of the raindrop spectrum data. Not only the sample member to obtain stable raindrop spectra data is enlarged by these ways, but also the mutual coordination and the analysis of the data are benefited. It is tested and result error analysis on its image position variation, the shape change, the drop spectrum overlaps and so on is made. The pattern recognition results are not influenced by the location and shape changes of single drop image in the data of the raindrop spectra data. The automatic image procedure is stable and reliable. It is very sensible when the separated drop spectra data are dealt with by the procedure. When many scattered drops formed from big drops, a large member of independent small raindrops will be formed, and it manifests that the splash problem can't be solved. The intelligence of the procedure is not high on the problem of raindrop overlapping each other together and it can't be recognized if the phenomenon of overlap really exists, the overlapped raindrops can't be separated and the overlapped drops can't be taken as one drop, so the member of the observing drops is small. From the error analysis of the raindrop spectra data, the relative error of spot mark diameter is smaller than 6% from 3 mm to 18 mm and the error is controlled within 6% for drops smaller than 4 mm, the error is even smaller when handling the small droplets. The result is obtained that this method is of great accuracy, stable performance, and very practical in the ground raindrop spectrum observation and the data processing. A new thought for the raindrop spectrum data processing analysis is provided and can be applied to the practical work.