During feather production, after cleaning and drying the feathers, mixed-colored feathers often appear. Existing sorting methods are inefficient, and manual screening is costly, which limits the quality of the produced feathers. To address this issue, experiments were conducted to compare and analyze technologies such as ResNet-18, YOLOv5, and OpenCV. Based on the results, a machine vision-based feather sorting system was designed and implemented. The design and implementation process of both the hardware and software of the system is described in detail. The hardware design includes the microcontroller, relays, solenoid valves, and air pumps. The software modules include HSV extraction, motion detection, color recognition, data communication, air pump control, and remote monitoring. By integrating image processing and machine vision algorithms, an automated feather sorting system based on air pump control was designed and implemented.