Research on crop diseases has become a hot topic of the application of artificial intelligence technology in smart agriculture. Existing studies are restricted by the factors such as difficulties in collecting data，high cost of technology implementation，and complex crop disease trends. Plant electronic medical records（PEMRs）formed by Beijing Plant Clinic provides a new idea for the diagnosis and prevention of crop diseases and pests. PEMRs are stored in the form of multi-modal data，containing a wealth of plant information，disease information，and environmental information. Therefore，how to mine PEMRs information and utilize it to assist follow-up research is an urgent problem to be solved. In view of the information representation ability of knowledge map，the mining ability of machine learning and the feature extraction ability of deep learning，structured data is used to construct the knowledge graph of crop diseases and pests，unstructured data and domain knowledge are used for knowledge enhancement according to the characteristics of PEMRs. Further，Neo4j graph database and graph data science（GDS）combined with machine learning algorithm was employed to conduct association mining from three dimensions of popular point discovery，link discovery and similar disease discovery. At the same time，text feature extraction and disease diagnosis were realized by using unstructured text data based on the bidirectional encoder representation from transformers （BERT） with convolutional neural network （CNN） model，and intelligent service was realized by simulating plant doctor. The comprehensive accuracy of 20 common diseases was up to 93. 13%. This study can provide theoretical support for timely diagnosis，symptomatic control，scientific medication guide and assistant decision-making of crop diseases and pests，and innovate a new model and new business of social service of agricultural science and technology.

Rapid and accurate detection of pathogen spores is an important step to achieve early diagnosis of diseases in precision agriculture. Traditional detection methods are time-consuming, laborious, and subjective, and image processing methods mainly rely on manually designed features that are difficult to cope with pathogen spore detection in complex scenes. Therefore, an MG-YOLO detection algorithm (Multi-head self-attention and Ghost-optimized YOLO) is proposed to detect gray mold spores rapidly. Firstly, Multi-head self-attention is introduced in the backbone to capture the global information of the pathogen spores. Secondly, we combine weighted Bidirectional Feature Pyramid Network (BiFPN) to fuse multiscale features of different layers. Then, a lightweight network is used to construct GhostCSP to optimize the neck part. Cucumber gray mold spores are used as the study object. The experimental results show that the improved MG-YOLO model achieves an accuracy of 0.983 for detecting gray mold spores and takes 0.009 s per image, which is significantly better than the state-of-the-art model. The visualization of the detection results shows that MG-YOLO effectively solves the detection of spores in blurred, small targets, multimorphology, and high-density scenes. Meanwhile, compared with the YOLOv5 model, the detection accuracy of the improved model is improved by 6.8%. It can meet the demand for high-precision detection of spores and provides a novel method to enhance the objectivity of pathogen spore detection.