To address the problems of low spore capture efficiency and insufficient automation in wheat stem rust monitoring, this study developed an integrated spore detection system featuring automatic capture, identification, and remote monitoring. A microfluidic chip with a four-stage micro-separation structure was designed and optimized via numerical simulation and fabricated using soft lithography with PDMS. An aerosol generator was used to simulate spore diffusion, a microscopic imaging system was applied to obtain spore morphological features, and an intelligent spore recognition model was established based on the YOLOv8 algorithm. Finally, a spore capture and analyzer integrated with environmental monitoring and remote communication functions was developed. The results showed that the simulated enrichment efficiencies of the microfluidic chip for spores of 100 μm, 50 μm, 25 μm, and 10 μm were 89.3%, 92.6%, 95.1%, and 90.8%, respectively, with an actual enrichment efficiency of 86.67%. The mAP@50 of the spore recognition model reached 96.7%, with precision and recall of 93.8% and 89.5%, respectively. The system established in this study enables high-efficiency capture, accurate identification, and full-process unattended operation of Puccinia graminis f. sp. tritici spores, providing technical support for the intelligent monitoring of airborne crop diseases.