The automatic monitoring timely and accurately of crop diseases has become an important research field in precision agriculture. Aiming at the low application rate of crop disease identification models in real field environment, a disease identification method based on multi-feature fusion and improved deep belief network was proposed. We Obtained representative samples in field conditions, then we augmented the data set. K-means clustering segmentation and morphological corrosion processing were utilized to obtain segmentation maps with clear boundaries and low noises. Then color features, shape features and textures of disease images were extracted respectively and they were fused to normalize as input data. A corn disease recognition model based on deep belief network was designed, using labeled and unlabeled dual hidden layer network structure to investigate the DBN hidden layer node combination mode. We obtained the optimal hidden layer node number combination method for disease classification: [26,85,29,4]. The accuracy of optimal DBN was 92.79%. On this basis, the deep belief network recognition model was optimized by particle swarm optimization algorithm for further performance enhancement. The experiment indicated that recognition effect using multi-feature fusion as input vectors was better than a single feature. The updated PSO-DBN reached the accuracy of 96.65%, which had a faster convergence speed and higher recognition accuracy of 3.86% than the standard DBN. Compared with state-of-the-art methods including SVM, ANN and CNN models, the proposed method can effectively dig deep digital features of disease areas or lesions and has the best performance, which could meet the needs of intelligent identification of field diseases.