0.13%～2.12%。In order to improve the precision of soil moisture testing under natural causes and nonlinear system, the fuzzy controller and signal pickup assembly based on DSP(TMS320VC5402) were constructed by neural networks technique, and the signal processing and optimized method were raised. The BP learning algorithm for fuzzy control rule and amending rule were determined, moreover, the experiment and verification were done with the system. The results of experiment show the errors of soil moisture test by neural networks technique are 0.13%～2.12% lower than by routine method.