Aiming at the ambiguity and uncertainty between nutrient deficiency and color characteristic of plant canopy image, a classification decision model based on regularized adaptive fuzzy neural network was set up to diagnose plant nutrition by using the complete rules of inference of fuzzy logic and adaptive of neural network. The “if-then” rules was fully used by the model, and the adaptive selection of law-level nodes and back propagation learning algorithm were given, meanwhile, network inference construction was perfected. The result of diagnosing soybean nutrient deficiency showed that the accuracy can be reach to 100%, meanwhile, the model has many advantages such as fast speed, stable, high precision, good robustness, as well as good adaptability and practical applicability.